beaufort sea mesoscale meteorology study - IARC Research

1 downloads 69 Views 20MB Size Report
Summary of the Beaufort Sea Mesoscale Meteorology Modeling Study Phase I..... .............2. 2.1. Tasks Overview .
PHASE I FINAL REPORT AND PHASE II STUDY PLAN for the

BEAUFORT SEA MESOSCALE METEOROLOGY STUDY

Prepared for

U.S. Department of the Interior Minerals Management Service Alaska Outer Continental Shelf Region Anchorage, Alaska Prepared by

Arctic Region Supercomputing Center, Geophysical Institute & International Arctic Research Center at the University of Alaska Fairbanks

Contract 0106CT39787

PHASE I FINAL REPORT AND PHASE II STUDY PLAN for the Beaufort Sea Mesoscale Meteorology Study Jing Zhang

Jeremy Krieger

Arctic Region Supercomputing Center, University of Alaska Fairbanks, Fairbanks, Alaska Martha Shulski Geophysical Institute, University of Alaska Fairbanks, Fairbanks, Alaska Xiangdong Zhang

David Atkinson

International Arctic Research Center, University of Alaska Fairbanks, Fairbanks, Alaska

September 2008 Acknowledgment This study was funded by the U.S. Department of the Interior, Minerals Management Service (MMS), Alaska Outer Continental Shelf Region, Anchorage, Alaska, under Contract No. 0106CT39787 as part of the MMS Alaska Environmental Studies Program. Disclaimer The opinions, findings, conclusions, or recommendations expressed in this report or product are those of the authors and do not necessarily reflect the views of the U.S. Department of the Interior, nor does mention of trade names or commercial products constitute endorsement or recommendation for use by the Federal Government.

TABLE OF CONTENTS List of Tables ................................................................................................................................ i List of Figures .............................................................................................................................. ii List of Units ................................................................................................................................ iv Abstract .........................................................................................................................................v Task Requirements – Restatement ............................................................................................. vi 1. Introduction ............................................................................................................................1 2. Summary of the Beaufort Sea Mesoscale Meteorology Modeling Study Phase I..................2 2.1. Tasks Overview ............................................................................................................. 2 2.2. Achievements..................................................................................................................5 2.2.1. Data Collection and Analysis..............................................................................5 2.2.2. Model Selection ................................................................................................10 2.2.3. Model Sensitivity Analysis ...............................................................................12 2.2.4. Data Assimilation with QuikSCAT Winds.......................................................21 2.2.5. Wind Field Simulations ....................................................................................23 2.3. Conclusions and Discussions........................................................................................29 3. Study Plan for the Beaufort Sea Mesoscale Meteorology Modeling Study Phase II ...........31 3.1. Introduction ..................................................................................................................31 3.2. Data Collection, Bias Correction, and Error Analysis..................................................33 3.2.1. In Situ Observational Data Collection ..............................................................33 3.2.2. Satellite Data Collection ...................................................................................35 3.2.3. Data Bias Correction and Quality Control........................................................37 3.2.4. Satellite Data Error Analysis ............................................................................38 3.2.5. Climatological Analysis with Collected Data...................................................38 3.3. Potential Field Work .....................................................................................................39 3.4. Optimization and Improvements of Model Physics......................................................40 3.5. Optimization of Model Configurations: Nudging and Assimilation Strategy ..............43 3.6. Production Simulation and Result Validation...............................................................45 3.7. Climatology, Interannual Variability, and Long-Term Change of Beaufort Sea Surface Winds..............................................................................................................49 3.7.1. High Resolution Climatological Features of Beaufort Sea Surface Winds .......49 3.7.2. Interannual Variability and Long-Term Change of Beaufort Sea Surface Winds .................................................................................................................51 3.7.3. Physics for Shaping the Wind Field Climatology, Variability, and Long-Term Change ...............................................................................................................52 4. References.............................................................................................................................55

List of Tables Table 2.1. List of stations compiled for this study..................................................................6 Table 2.2. Physics schemes used for MM5/WRF comparison tests .....................................15

i

List of Figures Figure 2.1. Wind direction frequency distribution (%) for August at Cottle Island, surface (left) and 850 hPa (right). ...............................................................................8 Figure 2.2. Surface and 850 hPa wind scatter plots for Betty Pingo (left) and Galbraith Lake (right) for August. ..............................................................................................9 Figure 2.3. Surface wind speed and vectors difference at 00 UTC August 11, 2000 in (48 hr MM5 forecast – NARR) (top); 48 hr WRF forecast – NARR (middle); WRF – MM5 (bottom). Red colors signify positive differences, blue signify negative. NARR SLP (black contours) is shown for reference. ...............................11 Figure 2.4. Topography and bathymetry of the study area; the modeling domain used for much of this study is marked with a red box.......................................................12 Figure 2.5. Average RMSE and bias for 10-m wind speed (left) and direction (right) for the ERA-40 (red) and NARR (blue) runs, averaged over all available stations. ......14 Figure 2.6. Average RMSE and bias for 10-m wind speed (left) and direction (right) for the 5 km (red) and 10 km (blue) runs, averaged over all available stations. ............15 Figure 2.7a. Comparisons of the downward longwave (LW) (upper panel) and shortwave (SW) (lower panel) radiation between SHEBA observations and the WRF simulations with the Thompson microphysics (M8), the Kain-Fritsch cumulus (C1) and varied radiations (R1 for RRTM LW and Dudhia SW; R3 for CAM LM and SW; R12 for RRTM LW and Goddard SW; R31 for CAM LW and Dudhia SW; R32 for CAM LW and Goddard SW) from May 9–13, 1998. ......17 Figure 2.7b. Same as Figure 2.7a but for the WRF simulations with the Lin microphysics (M2). ...................................................................................................18 Figure 2.7c. Same as Figure 2.7a but for the WRF simulations with the WSM 6-class microphysics (M6). ...................................................................................................18 Figure 2.8. RMSE of 10-m wind speed (left) and direction (right) for a series of sensitivity tests with WRF v3.0. ...............................................................................19 Figure 2.9a. Station wind speed RMSE for three runs: Nudg, Nudgall, and Nonudg..........20 Figure 2.9b. Station wind direction RMSE for three runs: Nudg, Nudgall, and Nonudg.....20 Figure 2.10a. The RMSE of the 10-m wind speed (left) & direction (right) grouped by forecast hour, for the QuikSCAT assimilation (yellow), control run (red), and the NARR analyses (blue) vs. QuikSCAT data. .......................................................22 Figure 2.10b. RMSE (left column) and bias (right column) of the 10-m wind speed (upper panel), wind direction (middle panel), and SLP (lower panel), grouped by the forecast hour for the QuikSCAT assimilation run (yellow), control run (red), and NARR analyses (blue) vs. station observations. ......................................22 Figure 2.10c. RMSE of the 18-hour forecasted 10-m wind speed (left), direction (middle), and SLP (right) at individual stations for the QuikSCAT assimilation (yellow), control run (red), and NARR analyses (blue). Stations 1–9 are the coastal stations from west to east, and 10–14 are the inland stations from north to south......................................................................................................................23 Figure 2.11. WRF simulated surface wind directions vs. 700 hPa geostrophic wind direction 10 km north of Lonely for June–September 2004. ....................................24 Figure 2.12a. Comparisons of the WRF modeled surface temperatures over sea (50 km north of Lonely) and land (50 km south of Lonely) for June–September 2004. ......25

ii

Figure 2.12b. Comparisons of the WRF modeled 700 hPa geostrophic wind direction and surface wind direction 10 km north of Lonely for June–September 2004.........25 Figure 2.13a. WRF simulated temperature (color), potential temperature (contour), and wind field circulation (vector) along the south-north cross section centered at Lonely (red stars mark the location of Lonely) at 16 UTC, 22 UTC July 20, and 06 UTC July 21.........................................................................................................26 Figure 2.13b. WRF simulated horizontal winds parallel to the south-north cross section centered at Lonely at 00 UTC Jul 21. The red star marks the location of Lonely....26 Figure 2.14. Differences in the 10-m wind speed (color) and wind vectors between Noter and Ter runs at 08 UTC Jan. 23, 2008, when a cyclone moved over the Beaufort Sea..............................................................................................................27 Figure 2.15. Differences of the 10-m wind speed (color) and wind vectors between Noter and Ter runs at 18 UTC Jan. 08, 2008, when a high pressure system was located over the Chukchi-Beaufort seas. ..................................................................27 Figure 2.16. Surface wind direction (WD) in the No-ter run vs. WD in the Ter run at 10 km (upper panel) and 150 km (lower panel) north of Barter Island (left panel), Cottle Island (middle panel), and Lonely (right panel) for January 2008. ...............28 Figure 2.17. Comparison of onshore and offshore geostrosphic (left panel), ageostrosphic (middle panel), and actual winds (right panel) at Barrow during June–September 2004. ..............................................................................................29 Figure 3.1. Model-simulated 3-dimensional ocean circulation and salinity distribution in the Beaufort Sea, when the Arctic Oscillation (AO) is in its (a) positive; and (b) negative phase. The solid and dashed contours show the zonal velocity and the arrows show the meridional and vertical velocities. The eastward coastal currents and upwelling demonstrate a sensitive response to the surface wind forcing . .....................................................................................................................32 Figure 3.2. Flowchart of the step-by-step procedures for producing Beaufort Sea mesoscale meteorology database . ............................................................................47 Figure 3.3. The PDF showing the changes in the frequency distribution of maximum storm intensity over the northern hemisphere in winter (December–February) between the 20th and 21st centuries, suggesting that there will be fewer weak storms in a future warming climate . ........................................................................52 Figure 3.4. The year-by-year positions of the Beaufort high in January when the AO is in its positive (a), and negative phase (b). ................................................................53

iii

List of Units Sea Level Pressure (SLP)....................................................................................................hPa Wind Speed.........................................................................................................................m/s Wind Direction...............................................................................................................degree Temperature ........................................................................................................................ °C Potential temperature ......................................................................................................... °C Shortwave Radiation ....................................................................................................... w/m2 Longwave Radiation ....................................................................................................... w/m2 Seaice Concentration ............................................................................................................ % Snowcover............................................................................................................................. m Friction Velocity ................................................................................................................m/s

iv

Abstract This report, as one of the project tasks, is planned to be delivered to the Minerals Management Service (MMS) of the U.S Department of the Interior (DOI) by the end of the Phase I study. The aim of this report is to summarize the research efforts we conducted during Phase I and develop a further research plan for Phase II of the Beaufort Sea Mesoscale Meteorology Modeling Study. The Phase I study began with a literature review with the aim to summarize past and present research efforts concerning the mesoscale meteorological models that would best support MMS objectives in modeling the meteorology of the Beaufort Sea region. Based on the review, the Weather Research and Forecasting (WRF) model was selected as the preferred model to be used in order to best achieve this goal. Initial sensitivity tests with the selected model were conducted, which included the analysis of sensitivity to the forcing data and modeling configuration. QuikSCAT ocean surface wind speed and direction data were also assimilated into the WRF model to improve the simulation of the Beaufort Sea surface wind field. The model’s performance in simulating the wind field was analyzed, with emphasis placed on evaluating the capabilities of WRF to simulate the sea breeze and topographic effects. Overall, the WRF model performed reasonably well in estimating the surface winds, as well as capturing the timing of the wind shifts and the magnitude of the surface wind speed. The capability of WRF in simulating the sea breeze-influenced surface wind fields for the Beaufort Sea coast was confirmed. The topographic effects of the Brooks Range were found to exhibit complicated impacts under different types of weather systems. Interactions between the sea breeze circulation and the Brooks Range were also uncovered. All of these are worth further investigation. Based on the preparatory studies conducted in Phase I, we have developed a research plan for the follow-up Phase II study that will meet MMS objectives for the accurate specification of the surface wind fields in the Beaufort Sea region, which in turn will ultimately be used for oil spill modeling. The major components for the Phase II study plan include: 1) further data collection, including potential field work, for the offshore open water areas in particular, which will be used for further validation of model performance and improvement of the simulation of Beaufort Sea mesoscale meteorological conditions; 2) further modeling studies for the optimization of the model physics and configuration, including the coupling of a sea ice model, in order to establish a well-tuned Beaufort Sea mesoscale model; 3) a 5-year experimental simulation for validating the modeling results and for use in driving a test simulation with the oil spill or wave model; 4) a 30-year (1978–2008) production simulation, along with an uncertainty assessment of the modeled surface wind fields; and 5) an analysis of the climatology, interannual variability, and long-term change in the features of the modeled Beaufort Sea surface winds.

v

Task Requirements – Restatement: For the purpose of clearly documenting the research efforts made in Phase I and in producing a detailed research plan for Phase II that meets MMS objectives for the modeling of the mesoscale meteorology of the Beaufort Sea region, the task requirements from the MMS contract are restated here: C.3.6 – Task (F) Produce a Phase II plan to develop a Mesoscale Meteorological Model for the Beaufort Sea to meet MMS objectives Write a detailed plan and budget for Phase II which shall be used to direct improvements to the mesoscale meteorological model for the subject study area and produce a superior model that provides substantial resolution enhancements that shall fulfill or exceed the MMS stated objectives. The plan shall address the following: •









Define prioritized research and data acquisition tasks for the development of a high resolution mesoscale meteorology model for the Beaufort Sea that will achieve MMS objectives. This may include, but not limited to model code development, additional satellite and field data collection efforts, user interface modifications, user output etc… Develop a plan of action that shall accomplish these research tasks. The plan shall contain information on the required project time frame, tasks, personnel assigned to each task, personnel qualifications to complete each task, personnel hours for each task, equipment, deliverables, time frame for deliverables, proposed data collection efforts, and a proposed budget for each task that shall be needed to achieve the development and deployment of a mesoscale meteorological model for the Beaufort Sea that shall meet or exceed MMS objectives. The plan shall outline those parties that are willing to provide cost, time sharing capabilities, based upon the outcome of Task E. The MMS shall require if determined feasible from Task D that the mesoscale meteorological model can run successfully on MMS PC based computers and detailed documentation is provided that describes the loading and operation of the model. The contactor shall assist with the loading of the software and data on MMS and SRB machines if necessary. If feasible, the contactor shall provide model output as ESRI ARCGIS grids or shapefiles that can be used on MMS computers. The contractor shall also discuss the flexibility of the model to generate other types of output that would be helpful towards other MMS applications as stated in Section C.2. The phase I final plan shall guide the implementation of the mesoscale meteorological model under Phase II. The MMS shall make the final determination on how to proceed with the final development of the mesoscale meteorological model in Phase II based upon the fulfillment of all tasks by the contractor within Phase I, the proposed project plan and budget submitted to MMS, and from the guidance of the SRB.

Phase I Final Products: The final report of Phase I includes model evaluation, sensitivity analysis of model runs, recommendation for model selection, and Phase II implementation plan. The implementation plan for Phase II shall document the tasks, personnel, and proposed budget that are required to complete the development of a mesoscale meteorological model for the Beaufort Sea and training MMS staff to run the developed model and load the model and data on MMS computers.

vi

1. Introduction Prudhoe Bay, located on Alaska’s Beaufort Sea coast, is one of the largest oil fields in the world. New development opportunities exist along the Beaufort Sea coast and in the Chukchi Sea, as evidenced by the 2008 Oil and Gas lease sale that generated more than $2.6 billion in revenue to the U.S. Oil development, however, is always accompanied by the potential threat of oil spills. In particular, the Beaufort Sea, with its coastal areas, is a vulnerable and fragile region with an ecosystem and environment that is sensitive to human impact. It is therefore of critical importance to be able to predict dispersal and movement of oil spills, and to assess the potential impacts on the environment if a spill should occur. Surface wind, primarily determined by prevailing local weather patterns and prominent underlying geographic features, is a crucial parameter for assessing and predicting dispersal and movement of oil spills. As such, the U.S. Department of the Interior, Minerals Management Service (DOI/MMS) has initiated and sponsored an environmental study of the Beaufort Sea mesoscale meteorology, specifically the surface wind and stress fields, aiming to ensure accurate simulation and prediction of ocean and sea ice circulation and correct assessment of oil spill risk. Sponsored by MMS, we have conducted an initial study (Phase I) of Beaufort Sea mesoscale meteorology over the past two years (Zhang et al. 2007, 2008). We began the Phase I study with a literature review with the aim to summarize past and present research efforts concerning the mesoscale meteorological models that would best support MMS objectives in the development of the Beaufort Sea mesoscale meteorology model. A ProCite database containing over 400 article titles and abstracts has been compiled. The models’ availability, capability, performance, and potential for fulfillment of MMS objectives have been documented. Based on the review, the Weather Research and Forecasting (WRF) model (Skamarock et al. 2005) has been selected as the preferred model to be used as the foundation for future development of the Beaufort Sea Mesoscale Meteorological Model. WRF represents the state of the art in mesoscale meteorological modeling, and, just as important, it appears to represent the direction that mesoscale modeling will take in the near future. In support of model verification and data assimilation, data from in situ observations and satellite retrievals have been collected in the Phase I study. Three principal agencies were utilized for the compilation of hourly in situ observational data with a total of 40 stations over the North Slope and the Beaufort Sea coast: the National Climatic Data Center (NCDC), the Water and Environmental Research Center (WERC) at the Institute of Northern Engineering, University of Alaska Fairbanks, and the MMS. Surface Heat Budget of the Arctic (SHEBA) field measurements and Quick Scatterometer (QuikSCAT) SeaWinds over the Beaufort Sea were also collected. The QuikSCAT SeaWinds were used for assimilation into the model, and the others were used for model verification. Initial sensitivity tests with the selected model were conducted in the Phase I study, which include an analysis of sensitivity to forcing data and to the modeling configuration, including the horizontal resolution, model physics, and the nudging technique. We also assimilated the QuikSCAT ocean surface wind speed and direction data into the WRF model to improve the simulation of the Beaufort Sea surface wind field. The model’s performance in simulating the wind field was analyzed, with emphasis placed on evaluating the capabilities of WRF to

1

accurately simulate the sea breeze and topographic effects. Overall, the model performed reasonably well in estimating the surface winds as well as capturing the timing of the wind shifts and the magnitude of the surface wind speed. The capability of the WRF model in simulating the sea breeze-influenced surface wind fields for the Beaufort Sea coast was confirmed. The topographic effects of the Brooks Range were found to exhibit complicated impacts under different types of weather systems. Interactions between the sea breeze circulation and the Brooks Range were also uncovered. All of these are worth further investigation. From the preparatory studies conducted in Phase I, we learned that an extensive collection of observational data, including both in situ observations and satellite retrievals, is extremely important for our study region, which contains a sparse observational network. The collected data will not only help to calibrate the model’s performance, in order to better tune it for the Beaufort Sea region, but will also be used to improve the simulations themselves through data assimilation. Another essential effort for the Beaufort Sea mesoscale modeling study is the improvement of the model through tuning and optimization of various built-in model physics and configuration options, as well as the implementation of more advanced model physics. Based on the Phase I study, we developed the following research plan for the follow-up Phase II study: 1) further data collection, including potential field work, for the offshore open water areas in particular, which will be used for further validation of model performance and improvement of the simulation of Beaufort Sea mesoscale meteorological conditions; 2) further modeling studies for the optimization of the model physics and configuration, including the coupling of a sea ice model, in order to establish a well-tuned Beaufort Sea mesoscale model; 3) conducting a 5-year experimental simulation for validating the modeling results and for use in driving a test simulation with the oil spill or wave model; 4) conducting a 30-year (1978–2008) production simulation, along with an uncertainty assessment of the modeled surface wind fields; 5) perform an analysis of the climatology, interannual variability, and long-term change in the modeled Beaufort Sea surface winds. We will arrange this report as follows: the major achievements made in the Phase I study are summarized in Section 2, and the details of the Phase II study plan are given in Section 3. The personnel and budget requested for conducting the Phase II study are included in Sections 4 and 5, respectively. 2. Summary of the Beaufort Sea Mesoscale Meteorology Model Study Phase I 2.1. Tasks Overview A total of seven tasks (Tasks A–G) were defined in the contract for the Phase I study. Upon receiving the award, we immediately began the project with Task A (Formation of Scientific Review Board (SRB)). Three experts were invited to serve as our SRB members. They helped in polishing the study plan, reviewing the study reports, and, most importantly, providing invaluable insights throughout the course of the project. Following this, project managers from MMS, investigators, and SRB members gathered in Fairbanks in November 2006 for the initial project meeting and discussed the study plan as defined in Task B (Conduct Initial Project Meeting and Submittal of Draft and Final Study Plan). After the meeting, the final study plan was submitted to MMS and SRB members. At the end of the first year of Phase I, a ProCite 2

bibliographic database, which contained over 400 relevant sources, and a model report documenting important mesoscale meteorological model development and associated data assimilation efforts were submitted to MMS and SRB members for review – the final product of Task C. Six months later, the final product of Task D: Model Sensitivity Analysis was submitted to MMS and SRB members, who then provided valuable and constructive comments on the modeling report. A revised modeling report was submitted in June 2008. Throughout the course of conducting the Phase I study, we developed a partnership with a NOAA-sponsored project entitled “Social Vulnerability to Climate Change and Extreme Weather of Alaskan Coastal Communities”, in which two of the investigators were involved by performing wind field downscaling in Alaska under current and future climates. The PI of this NOAA project will join our team for the Phase II study. With MMS’s help, a partnership with the EPA’s meteorological data collection program over the Beaufort Sea has been developed; the EPA will share the collected meteorological data with us. An effort to establish a partnership with the National Weather Service in Anchorage is being developed. The improved WRF model for the simulation of Beaufort Sea wind fields may benefit the wave and ice forecasting that they perform and their feedback will also assist our model development. The report presented herein is to fulfill Task F (Develop Phase II Study Plan). Further details regarding the performed tasks are given below. Task A: Formation of Scientific Review Board We invited Drs. David Atkinson, Peter Olsson, and Tom Weingartner to serve as the members of the Scientific Review Board (SRB) for this phase of the project, based on their combined experience in studying the Alaskan atmospheric and oceanic environment, as well as their familiarity with Alaskan mesoscale modeling. They accepted our invitation, and have served to provide both MMS and the project team with invaluable insights throughout the course of the project. Task B: Conduct Initial Project Meeting and Submittal of Draft and Final Study Plan Prior to the initial project meeting, we submitted a draft study plan to MMS, detailing our proposed course of action in carrying out the tasks associated with this project. The initial project meeting was subsequently held in Fairbanks on November 13, 2006. Participants included all group and SRB members in addition to Warren Horowitz, Caryn Smith, Ron Lai, Olivia Adrian, and Walter Johnson of MMS and Kevin Engle of GINA. The group presented the details of our proposed study plan at this meeting, and all attendees participated in a discussion about our proposal. Afterwards, a meeting summary was sent to all participants. Based on the many valuable comments and suggestions that were given at the meeting, we amended our draft study plan and submitted the final study plan to MMS. Task C: Provide a ProCite bibliographic database of mesoscale meteorology model development efforts for the project study area The group conducted a search of the available literature related to research conducted with mesoscale meteorology models, paying particular attention to those studies that focused on polar and coastal environments. In addition, we investigated sources of both conventional and satellite observations that could potentially be useful in our study of the Beaufort Sea region. Based on

3

our literature review, we compiled a ProCite bibliographic database, which contained over 400 relevant sources, and submitted it to MMS. Task C1: Document important mesoscale meteorological model development and associated data assimilation efforts With a combination of the information obtained from our literature review and the sum of our own experiences with mesoscale modeling, we selected the WRF model as our focus for this study, and the platform around which the Beaufort Sea Mesoscale Meteorological Model would ultimately be built. We put the results of our efforts toward this task into a document entitled “Report on Mesoscale Model Development and Data Assimilation Efforts”, which was submitted to MMS. In this report we compared the relative merits of the three potential mesoscale models under consideration (MM5, WRF, and RAMS), examining the capabilities of the models to satisfactorily meet MMS’s objectives. In doing so, we investigated the variety of physics options that are available in each model, the data assimilation techniques that each model employs, the ability of the models to simulate necessary Arctic processes (e.g., sea ice), and the spatial resolutions that each model is capable of utilizing, among other factors. In addition to our own evaluation of the models’ capabilities based on the documentation and code, we performed a comprehensive review of published studies which used the different models for a variety of applications, including physics sensitivity tests and model comparison tests, in order to better gauge the relative success of the models in performing mesoscale simulations. In the end, we settled on WRF as the preferred model with which to continue our modeling efforts, and documented its basic components and usage in the report. In addition to our model examination and selection, we researched additional observational data sources and types (aside from the conventional radiosonde and surface station data) that could potentially be useful in the Beaufort Sea region, both as data assimilated into the model and for model verification. We found three sources of surface station data (NCDC, WERC, and MMS) and five different types of satellite-derived data that had the potential to assist our modeling efforts. Of these latter types, the QuikSCAT ocean-surface winds were already capable of being assimilated via the WRF-Var package and so were used as part of our subsequent sensitivity analysis. Task D: Conduct Initial Sensitivity Analysis for Model Evaluation After making our model selection in Task C, we proceeded to perform a sensitivity analysis using the chosen model (WRF). Through the simulation of several high-wind case studies, we conducted tests comparing WRF results to those generated by MM5, tested the results obtained from the use of two different reanalysis datasets (NARR and ERA-40) for initial and lateral boundary conditions, tested the model’s sensitivity to horizontal grid spacing (10 km vs. 5 km), and examined the impacts of assimilating both conventional and QuikSCAT satellite-derived observations. We also performed a battery of sensitivity tests examining the performance of the model using many combinations of various physics options, including different shortwave and longwave radiation, microphysics, and cumulus parameterizations, and verified the results against observations made during the SHEBA experiment, which provided us with more types of

4

observational data than are available from conventional surface stations. In so doing, we were able to determine a physics package that would perform well in the region of interest. In addition to these short-term case studies, we also performed several longer-term runs of one month or more in order to gauge how well the model was able to capture the important nearsurface circulation patterns along the Beaufort Sea coast, namely the summertime sea breeze and terrain-influenced flow impacted by the Brooks Range. We conducted simulations for both summer and winter months, both with and without terrain, in order to isolate these effects. The full results of all the above sensitivity tests were submitted to MMS in a report entitled “Beaufort Sea Mesoscale Meteorology Model Evaluation: Initial Sensitivity Analysis”. Task G: Submission of Journal Article A manuscript entitled “Assimilating QuikSCAT SeaWinds with the Weather Research and Forecasting Model for Beaufort Sea Surface Wind Simulation” is currently being prepared. This manuscript details the group’s efforts in using the WRF-Var software package to assimilate QuikSCAT ocean-surface wind data into WRF and the performing of two case studies comprising a total of 8 independent simulations. We examined the effects of the assimilation on the initial analyses, and performed verification of the modeling results, both with and without assimilation, against both surface station and QuikSCAT observations in order to gauge its success. We found that the assimilation of QuikSCAT data had an overall positive effect on the simulations, which has encouraged us to expand our use of this data source for future applications. Two additional manuscripts entitled “Beaufort Sea Wind Regime Study with Model Simulations and Observations” and “Verifying WRF Simulations for the Beaufort Sea” are under preparation to summarize the research results and findings of the Phase I study, for which a total of eight presentations at various conferences have also been presented. These have involved all aspects of our work, including both research done with the model as well as observational studies of data across the Beaufort Sea region, and have allowed us to present our work to the wider research community and foster communication with other groups who are focused on similar applications. 2.2. Achievements Following the requirements as delineated in the Phase I contract, we conducted all of the tasks as described in Section 2.1. While conducting the tasks A–G, a number of achievements were made which will be highlighted in this section. 2.2.1. Data Collection and Analysis One of the prime challenges related to atmospheric modeling in the Beaufort Sea region is the relative lack of observational data throughout the area. Not only is data sparse on land, but there are also little to no observations across much of the ocean. As such, one of the primary goals of this project was to identify whatever sources are available and collect them for use in assimilation into the model, as well as for model verification and analysis of the observations themselves.

5

In Situ Observations A preceding MMS-funded project, “Nearshore Beaufort Sea Meteorological Monitoring and Data Synthesis Project”, compiled an observational database containing much of the conventional surface data available in the region, which we used as the starting point for our investigation. This database contains observations from MMS, NCDC, and WERC stations, in addition to other more minor sources. We performed quality control checks on these data in order to ascertain the existence of any problems related to instrument icing, and found that some of the WERC data were indeed faulty. We removed these erroneous data from the database, and did not use them further. Ultimately, we compiled observations from a total of 40 surface stations from these three sources (Table 2.1), which were subsequently used for analysis and model verification. Table 2.1. List of stations compiled for this study Location

Longitude (deg) 149.417 145.567 156.763 -156.65 -143.577 -143.633 -143.583 -166.133 -154.75 -148.483 -149.583 -150.933 -148.683 -149.483 -148.817 -155.9 -149.583 -153.25 -150.983 -149.883 -149.9 -162.917 -163.017 -163.017 -148.517 -152.133 -152.133 -159.995 -159.85 -159.85

Elevation (m) 1017.1 636.0 4.0 6.1 1.5 11.9 2.0 3.0 3.0 585.0 982.0 5.0 108.0 814.0 286.0 27.0 20.0 9.1 17.0 4.9 5.0 6.0 6.0 6.1 5.0 81.1 81.1 9.1 27.1 27.1

Agency

Atigun Arctic Village Barrow Airport Barrow - Point Barrow Barter Island AP Barter Island DEW Barter Island AWOS Cape Lisburne AWOS Cape Simpson Pow Chandalar Lake Chandalar Shelf DOT Deadhorse Franklin Bluffs Galbraith Lake Happy Valley Iguigig Kuparuk Lonely Nuiqsut Oliktok Pow Oliktok DEW/AWOS Point Lay DEW Point Lay DEW/AWOS Point Lay Liz 2 Prudhoe Bay Umiat Umiat AP Wainwright AP Wainwright DEW Wainwright Liz 3

Latitude (deg) 68.183 68.117 71.287 71.333 70.134 70.133 70.133 68.883 71.05 67.5 68.067 70.333 69.717 68.483 69.15 59.3 70.317 70.917 70.217 70.5 70.5 69.817 69.733 69.733 70.4 69.367 69.367 70.639 70.617 70.617

Betty Pingo Franklin Bluffs

70.2795 69.8922

-148.8957 -148.768

11.6 77.7

WERC WERC

6

NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC NCDC

Sagwon West Dock West Kuparuk Badami Cottle Island Endicott Milne Point Northstar

69.4243 70.3806 69.4262

-148.6959 148.5608 150.3404

299.0 7.6 158.0

WERC WERC WERC

70.136 70.499 70.323 70.507 70.49

-147.009 -149.093 -147.865 -149.662 -148.698

17.0 5.0 5.0 7.0 21.0

MMS MMS MMS MMS MMS

QuikSCAT Observations While the use of conventional, in situ observations is ideal, due to the sparsity of such data in the Beaufort Sea region it is necessary to investigate the availability of remotely sensed data as well. We identified five different satellite data sources that have the potential to be useful for mesoscale modeling and the simulation of surface winds in this region: the Synthetic Aperture Radar (SAR)-derived surface wind over water (Pichel et al. 2005), the MODIS (Moderate Resolution Imaging Spectroradiometer) polar wind product (Key et al. 2003), the SeaWinds ocean wind data from the QuikSCAT satellite (http://podaac.jpl.nasa.gov/quikscat), the ocean surface wind derived from SSM/I (Special Sensor Microwave/Imager) radiometer measurements, and the ocean surface wind derived from ERS-2 (Earth Resource Satellite) wind Scatterometer (http://manati.orbit.nesdis.noaa.gov/doc/oceanwinds1.html). Of these, only QuikSCAT data could be directly assimilated into WRF with the WRF-Var package without any modifications to the source code, and so we focused our efforts on using this source for the first phase of the project. The QuikSCAT SeaWinds product is derived through the measurement of the reflection of emitted radar signals off of the surface of the open ocean by a satellite-borne scatterometer, a specialized type of microwave radar sensor. The radar is used to emit two separate beams to the earth’s surface, the scatter of which is then used to determine the surface wind speed and direction over liquid water surfaces. The result is a ~12 km resolution surface wind product that is (only) available over the open ocean, coming to within 20–30 km of the coast. SHEBA Observations Another data set that we utilized in our study was that produced by the Surface Heat Budget of the Arctic Ocean (SHEBA) experiment. SHEBA was a field experiment that took place in the Arctic Ocean from 1997–98, consisting of a research ship that was frozen into the pack ice and allowed to drift for 13 months. This platform collected many more types of measurements than are typically made by standard meteorological stations, including solar radiance and surface energy fluxes, which are valuable for validating model performance. This, plus the fact that this represented data from an offshore station location far removed from land, unlike all of our other collected in situ data sources, encouraged us to utilize SHEBA observations in the course of our model sensitivity tests. This allowed us to better evaluate the performance of various model physical schemes, while at the same time gave us a better idea of how well the model performed over the ice-covered ocean, which is a dominant feature of the Beaufort Sea region.

7

Data Analysis of In Situ Observations Wind speed and direction climatologies performed for the collected stations show typically high wind speeds for this Arctic environment (>10 m/s). There are slightly higher wind speeds during the cold seasons and slightly lower wind speeds during the short summer season, a result also found by the preceding MMS project. In addition, wind direction frequencies illustrate the typical bimodal distribution with the highest frequencies coming from the ENE direction and a lower secondary mode from the WSW (not shown). Previous work by Kozo (1982, 1986) has shown that there is a high frequency of onshore flow events in the Beaufort Sea coastal region of Alaska during the short summer season. With the recent decline of sea ice in this sector of the Arctic, and an earlier disappearance of the snowpack in spring, it is likely that such sea breeze events have become more prevalent. Wind speed and direction observations from the NCDC, WERC, and MMS sites were compared with the North American Regional Reanalysis (NARR) gridded data. NARR wind information at seven levels of the lower atmosphere, from 1000 hPa to 850 hPa at 50 hPa increments, was investigated to check for the occurrence and strength of onshore vs. offshore flow along the coast and at inland locations. Vertical profiles of wind direction in the lower atmosphere showed a transition from ENE prevailing winds at the surface to westerly winds at 850 hPa. This pattern of variability was strongest for July and August, the time of year with a generally snow-free landscape and the greatest potential for the setup of thermally-driven circulations. An example is shown for August at Cottle Island (Fig. 2.1). In addition, as shown in Fig. 2.2, there is a discontinuity in the wind direction between the surface and 850 hPa for a near-coast location (Betty Pingo), as compared to a site that is far inland (Galbraith Lake).

Figure 2.1. Wind direction frequency distribution (%) for August at Cottle Island, surface (left) and 850 hPa (right).

8

Figure 2.2. Surface and 850 hPa wind scatter plots for Betty Pingo (left) and Galbraith Lake (right) for August. Specific cases of clear and cloudy sky conditions were also tested for the occurrence and strength of the sea breeze. Incoming solar radiation data from the MMS network of stations allowed for the separation of such cases where the thermal imbalance would be highest (clear) and lowest (overcast). Onshore winds at the surface and offshore winds at 850 hPa occur for clear sky conditions whereas the directions are the same in the case of overcast skies. Error Analysis of QuikSCAT Winds Prior to assimilating observational data into a model, an estimate of the measurement error must first be made. Verification of QuikSCAT surface winds against ocean buoy data have been conducted for the Indian, Pacific, and Atlantic Oceans, but not for the Arctic Ocean due to the lack of buoys in this region. Most such verification studies (e.g., Ebuchi et al. 2002; Pickett et al. 2003) show that QuikSCAT surface winds are basically reliable, but that the quality does vary over different regions (Satheesan et al. 2007). Thus, before assimilating QuikSCAT surface winds into our simulations, we performed a rudimentary analysis of the quality of the QuikSCAT data in the Beaufort Sea region by comparing it with nearby (within 30 km) Beaufort Sea coastal station observations. Because QuikSCAT data is only available over the ocean, it was not possible to make direct comparisons with station data; there were very few data points within even a distance of 15 km. The true observational errors are thus undoubtedly less than those we calculated, and when QuikSCAT data is used to verify model results there is at most 5 km between the observation locations and model grid points. Overall, it was found that the QuikSCAT winds are generally very similar to the station observations, lending credibility to the accuracy of the product. Although our analysis showed a slight positive bias in the QuikSCAT wind speeds, the separation between QuikSCAT and station locations most likely explains this. Moreover, the QuikSCAT measurements are made over the open ocean, where terrain is of course absent and the surface friction is lower than on land, which undoubtedly leads to somewhat higher resultant surface wind speeds. The wind directions were also found to be generally similar between the two data sources, with relatively few large discrepancies between the QuikSCAT and station wind directions. Those that do exist are also no doubt partially a result of the distance between the QuikSCAT and station locations, as wind direction can change significantly over relatively small distances in a coastal environment.

9

These comparisons are very encouraging and give us confidence in using the QuikSCAT winds for this study. Also, Chen (2007) found an observational error of 1.42 (m/s)2 for this product, which is reasonable for use in the Beaufort Sea region. 2.2.2. Model Selection Upon commencing the Phase I study, one of the first things that had to be determined was the mesoscale model that we would be focusing our efforts on. As a result of our literature review (Zhang et al. 2007), three mesoscale models were initially selected: the Regional Atmospheric Modeling System (RAMS), the Fifth-Generation Penn State University (PSU)/National Center for Atmospheric Research (NCAR) Mesoscale Model (MM5) (Grell et al. 1994), and the WRF model. The RAMS model was immediately ruled out because it contains no treatment of sea ice, one of the most important features in the Beaufort Sea area. In addition to this severe limitation, which has sharply restricted the number of studies that have used RAMS in the polar region, there exists no broad user community to insure upgrades, support and documentation, in sharp contrast to the community model development of MM5 and WRF. Because WRF was developed as a successor to MM5, the MM5 and WRF models are related in many ways, sharing many of the same physical schemes. Considering the similarity between MM5 and WRF, we performed a simulation test to gauge the relative performance of MM5 and WRF (v2.2) over the region of interest, including the Beaufort and Chukchi Seas and the North Slope of Alaska. In the modeling configuration, we instituted a suite of physics options that were common to both models as listed in Table 2.2. A grid spacing of 10 km was used. The comparisons (Fig. 2.3) show that, when using the same physics schemes, the two models behaved remarkably similarly in their development of the circulation patterns and near-surface winds. Even at forecast hour 96 (not shown), after four days of free forecast, there exist relatively few notable differences between the pressure and wind fields of the two models. It is encouraging to know that even at its current early developmental stage, WRF is equal to the much more mature MM5 when using the same physics. Given that MM5’s development has been ceased by NCAR in favor of WRF and WRF is continuing to undergo rapid development, already possessing more advanced physics options than are available in MM5, particularly in its latest release (v3.0), the results of this comparison further encouraged us to choose WRF as the preferred model to be used as the foundation for future development of the Beaufort Sea Mesoscale Meteorological Model. Table 2.2. Physics schemes used for MM5/WRF comparison tests Microphysics: Longwave Radiation: Shortwave Radiation: Surface Layer: Land-surface: Boundary Layer: Cumulus:

WRF: NCEP 5-class RRTM Dudhia Monin-Obukhov NOAH MRF Kain-Fritsch (new Eta) 10

MM5: Reisner 1 (5-class mixed phase) “ “ “ “ “ “

300 km

300 km

300 km

Figure 2.3. Surface wind speed and vector difference at 00 UTC August 11, 2000 (48 hr MM5 forecast – NARR) (top); 48 hr WRF forecast – NARR (middle); WRF – MM5 (bottom). Red colors signify positive differences, blue signify negative. NARR SLP (black contours) is shown for reference.

11

2.2.3. Model Sensitivity Analysis With the selected WRF model (Section 2.2.2), we performed initial sensitivity tests to evaluate the model’s performance in the Beaufort Sea region. The modeling domain was defined over the Beaufort Sea region as shown in Fig. 2.4, including the Beaufort and Chukchi Seas, the North Slope and Brooks Range, along with the northwestern Yukon and the eastern tip of Russia. This domain configuration should allow the model to produce high-resolution simulations of the entire region that MMS desires surface wind data, as well as providing an adequate representation of the overall synoptic environment which impacts the study area. 600 km

Figure 2.4. Topography and bathymetry of the study area; the modeling domain used for much of this study is marked with a red box. In a regional model simulation, there is always a need to provide the model with initial and boundary conditions. Reanalysis data, such as from the European Centre for Medium-Range Weather Forecasts (ECMWF) Forty-Year Reanalysis (ERA-40), the National Centers for Environmental Prediction (NCEP)/NCAR Reanalysis Project (NNRP), and the North American Regional Reanalysis (NARR), are commonly used to provide the initial and boundary conditions in the WRF modeling system. In this study, we first conducted a set of sensitivity tests to compare the performance of these reanalysis data in forcing the regional modeling of the Beaufort Sea region. In addition, it is important to note that the Beaufort Sea region is a prominent geographical feature that is largely covered by sea ice on a seasonal basis and exhibits 12

great variability in sea ice, atmospheric, and oceanic conditions. Due to this geographical complexity, it is both important and non-trivial to determine the optimal model configuration for the WRF model’s application in this region. To this end, simulations testing sensitivity to the resolution, nudging technique, and various model physics options were also performed. In this section, we will summarize the major achievements made from these sensitivity tests. Forcing Data Analysis As mentioned above, there are three long-term sources of reanalysis data that cover the Beaufort Sea region: the NNRP, an ongoing reanalysis produced with a horizontal grid spacing of ~209 km with 17 pressure levels, available every 6 hours since 1948; the ERA-40, another global reanalysis product produced every 6 hours for the period 1957–2002 with a grid spacing of 1.125° (~125 km) and 23 pressure levels; and the NARR, an ongoing regional reanalysis covering the North American continent and surrounding areas, produced at a grid spacing of just 32 km with 29 pressure levels, available every 3 hours since 1979. Due to the relatively low resolution of the NNRP, which, in addition to suffering in accuracy, would necessitate several levels of nesting in the model in order to get down to the ~10 km scale that is desired for this project, we decided to forgo consideration of that dataset and instead focus our efforts on a comparison of ERA-40 and NARR. As NARR has a horizontal resolution nearly four times greater than that of ERA-40, it would logically be the preferred choice. The modeling domain constructed for this study (Fig. 2.4) fits within the northwest corner of the NARR domain. However, NARR only extends back to 1979, and just barely encompasses the study region, so it is worthwhile to examine the performance of WRF when using ERA-40 as well in order to be prepared should there be a need to either expand the study domain or simulate the pre-1979 period. Thus, two forcing datasets, ERA-40 and NARR, were compared for a simulation using a grid spacing of 10 km. The simulated results were verified against all available surface observational data collected in this project (Section 2.2.1), which for this case was a total of 14 stations. When verifying the simulations, we focused on the surface winds, as this is the primary objective of the current study. All model results were interpolated to each station location, and the root-mean-square error (RMSE) and mean bias were calculated across all available observations for each hour. Fig. 2.5 depicts the time series of the error statistics averaged over all available stations. It appears that the model run driven by the lower-resolution ERA-40 reanalysis actually slightly outperformed that using the NARR in simulating the surface wind speeds. While the statistics for the two runs are relatively similar in time, there is a slight but persistent edge to the ERA-40 run in lowering RMSE. This is a surprise, given the higher horizontal, vertical, and temporal resolution of the NARR data. The statistics for wind direction are more mixed, with a notable advantage held by the NARR simulation late in the period (after hour 63), when it greatly outperformed the ERA-40 run in predicting the timing of the wind shift that occurred at nearly every station around that time, associated with the passage of the low pressure system in the Beaufort Sea (not shown).

13

Figure 2.5. Average RMSE and bias for 10-m wind speed (left) and direction (right) for the ERA-40 (red) and NARR (blue) runs, averaged over all available stations. While a single case is obviously insufficient to be able to draw any definitive conclusions about the relative merits of the two reanalysis datasets, this performance certainly suggests that the ERA-40 would be a satisfactory source of data for driving WRF in the Beaufort Sea region. However, for the remainder of the sensitivity tests conducted in this study we continued to use NARR instead of ERA-40, due to the latter only being available through mid-2002, predating much of the observational data that we had access to via the MMS Beaufort Sea observational database. The result from this comparison is encouraging in the event that future work will require the model domain to be expanded, or to simulate a time period that predates the NARR. Grid Spacing Analysis Another consideration that needs to be addressed in producing a mesoscale model tuned for the Beaufort Sea region is the question of horizontal grid spacing. While it is generally accepted that higher model resolutions produce better, more realistic results, there is always the matter of the tradeoff between increasing resolution and the increased computational requirements that necessarily follow (e.g., Mass et al. 2002; Krieger and Zhang 2005). A grid spacing of 10 km is at the upper end of the range of MMS’s stated requirements for the model to be developed. While it is true that 10 km spacing might be too coarse to accurately simulate the dynamics within steep, rugged terrain, where the highly variable surface features may vary significantly over such a length scale, the same is not necessarily true of the region of interest, which is 14

largely composed of the ocean bordered on its south by the relatively smooth and flat North Slope. In order to address this concern, we designed two simulation runs, one with a grid spacing of 10 km and other with 5 km. All else in the modeling configurations of the two runs was identical. The simulated results were verified against all available surface observational data collected in this project (Section 2.2.1). Again, when verifying the simulations, we focused on the surface winds. All modeled winds were interpolated to each station location, and the RMSE and mean bias were calculated across all available observations for each hour. Fig. 2.6 depicts the time series of the error statistics averaged over all available stations. It is apparent that, for this case at least, there is little to no gain from the increase in resolution. Overall, the results are nearly identical for the wind speed, while a bit of difference emerges late in the simulation for the wind direction. However, these differences are as likely to benefit the 10 km run as they are the 5 km. Examining the comparisons between the simulations and observations at individual stations (not shown), the greatest differences were seen at the inland stations. This is unsurprising, since one of the major benefits of using higher resolution is a more accurate representation of the land surface, particularly of terrain.

Figure 2.6. Average RMSE and bias for 10-m wind speed (left) and direction (right) for the 5 km (red) and 10 km (blue) runs, averaged over all available stations. In the end, this comparison suggests that increasing the horizontal grid spacing beyond 10 km may have few, if any, benefits to the simulation of surface winds along the Beaufort Sea coast. Given the significant computational costs associated with doing so (taking 40 hours to complete a 96-hour run in this instance), it is very questionable as to whether there is sufficient value in 15

higher-resolution modeling in this region. Especially in light of the fact that the ultimate aim of this project will be to produce surface wind data over the ocean, where the benefit of having increased land-surface resolution is obviously irrelevant, and given the relative smoothness of the terrain on the North Slope bordering the Beaufort Sea, the need for additional effort toward this end would need to be further substantiated. As such, and for the sake of computational efficiency, we reverted back to using 10 km grid spacing for the remainder of this study. Model Physics Analysis Application of a regional model to a particular area is largely a process of selecting appropriate model physical parameterizations. This can be one of the most complicated tasks in establishing a model configuration as WRF possesses a large set of parameterization options for each physical process. As such, an extensive validation of different parameterizations was conducted to test the efficacy of various physical schemes and their combinations for the Beaufort Sea region. Three model physical processes – microphysics, radiation, and cumulus – were tested by modeling the Arctic storm that occurred from May 9–13, 1998, when the SHEBA observation platform was near the Beaufort Sea. Modeling results were validated against the SHEBA measurements collected in this project (Section 2.2.1), for which a wide range of observation types are present and thus a more thorough evaluation of the modeling results could be achieved. Various comparisons of the modeled variables (sea level pressure (SLP), 2-m temperature, and 10-m wind fields) with the SHEBA measurements show that the WRF model captured the basic evolution of the modeled Arctic storm (not shown). A relatively large discrepancy between the simulations and observations occurred after the 48th simulation hour, which is likely unavoidable forecast error related to the length of time that the simulation was run without nudging. It was also found that the choice of cumulus parameterization had a very minor impact on the simulations over our study area, an environment that is not favorable to convective activity. As such, the Kain-Fritsch cumulus scheme (C1) (Kain and Fritsch 1993) was used for the remainder of the analysis. Variable performance was found with different radiation schemes (Fig. 2.7). The Dudhia shortwave radiation scheme (Dudhia 1989) significantly underestimated downward shortwave radiation, especially under cloudy sky; both CAM (NCAR CCSM radiation) and Goddard (Chou and Suarez 1994) shortwave radiation schemes did a superior job. The performance of the longwave radiation schemes (both CAM and RRTM (Rapid Radiative Transfer Model) (Mlawer et al. 1997)) was similar when the Purdue Lin microphysics (Lin et al. 1983; Chen and Sun 2002) and the Thompson microphysics (Thompson et al. 2004) were used. However, when the WSM (WRF Single Moment) 6-class microphysics (Hong et al. 2004) was used, the RRTM scheme produced slightly stronger longwave radiation than the CAM scheme, especially under clear sky. The Thompson microphysics tended to overestimate the cloud relative to the Lin and WSM 6-class schemes. Cloud microphysics interacts with radiation; the overestimated cloud generated by Thompson resulted in overestimated longwave radiation and underestimated shortwave radiation.

16

None of these model physics schemes (microphysics, radiation, and cumulus) significantly impacted the surface winds in this Arctic storm simulation, however, which was likely due to the dominating influence of the strong synoptic system.

05/09/98

05/09/98

05/10/98

05/10/98

05/11/98

05/11/98

05/12/98

05/12/98

05/13/98

05/13/98

Figure 2.7a. Comparisons of the downward longwave (LW) (upper panel) and shortwave (SW) (lower panel) radiation between SHEBA observations and the WRF simulations with Thompson microphysics (M8), Kain-Fritsch cumulus (C1), and varied radiations (R1 for RRTM LW and Dudhia SW; R3 for CAM LM and SW; R12 for RRTM LW and Goddard SW; R31 for CAM LW and Dudhia SW; R32 for CAM LW and Goddard SW) from May 9–13, 1998.

17

05/09/98

05/10/98

05/11/98

05/12/98

05/09/98

05/13/98

05/10/98

05/11/98

05/12/98

05/13/98

Figure 2.7b. Same as Fig. 2.7a but for the simulations with the Lin microphysics (M2).

05/09/98

05/10/98

05/11/98

05/12/98

05/13/98

05/09/98

05/10/98

05/11/98

05/12/98

05/13/98

Figure 2.7c. Same as Fig. 2.7a but for the simulations with the WSM 6-class microphysics (M6). The other two model physics types, planetary boundary layer (PBL) and land-surface processes, which have a closer relationship with the simulation of surface wind, were later tested with the newly released WRF v3.0, in which more options are available for these parameterization schemes. For these tests, a series of 4.5-day simulations were conducted, which spanned the entire month of September 2004. The first 12 hours of each simulation (which overlapped the final 12 hours of the previous simulation) were used for spinning up the model, while the remaining 4 days were used for verification. Fig. 2.8 shows the RMSE of the 10-m wind speed and direction for these sensitivity tests as verified against all available station data collected in this study. In this figure, the ctrl run represents the best combination of physics options as taken from the above SHEBA tests, along with the Yonsei University (YSU) PBL scheme (Hong et al. 2006) and the NOAH land-surface model (Chen and Dudhia 2001; Mitchell et al. 2002; Ek et al. 2003). Sea surface temperature (SST) and sea ice were also varied in time according to the analysis data, taking advantage of one of the new capabilities in WRF v3.0. The nosst runs were performed with constant SST and sea ice fields (using the data at the initial time of each of the individual 4.5 day runs) in the same fashion as is done in WRF v2.2; sfc3 and sfc7 used the

18

Rapid Update Cycle (RUC) land surface (Smirnova et al. 1997, 2000) and Pleim-Xu (PX) land surface (Pleim and Xiu 1995; Xiu and Pleim 2001) schemes, respectively; bl2, bl7, and bl99 used, in order, the Mellor-Yamada-Janjic (MYJ, also known as Eta) PBL scheme (Mellor and Yamada 1982; Janjic 1990, 1994, 1996, 2002), the ACM2 (Asymmetric Convective Model version 2) PBL scheme (Pleim and Chang 1992; Pleim 2007a,b), and the Medium Range Forecast (MRF) PBL scheme (Troen and Mahrt 1986; Hong and Pan 1996); allpx used the PX land surface and ACM2 PBL schemes, which were developed by the same authors. In each test only the stated option was varied from the control run. We can see that, first of all, the ability to use time-varying SST and sea ice had a beneficial effect, particularly for the wind direction, showing the importance of converting our efforts to the new version of the model as we proceed to Phase II. The PX land-surface model (unavailable in WRF v2.2) showed promise as well, especially in reducing wind speed error. The results for the different PBL schemes were more mixed, as the Eta PBL produced reduced wind speed error but larger directional error, while the ACM2 PBL did just the opposite. However, when the PX land surface and ACM2 PBL schemes were combined, the results were far superior for both speed and direction, surpassing what each was able to do individually, indicating the necessity of analyzing the combined performance of different schemes. These preliminary efforts have demonstrated the desirability of moving our focus to the new version of WRF and highlight the importance of continuing to perform additional sensitivity tests for longer time periods, using a broader array of observations for verification, and utilizing the newly available physical schemes.

Figure 2.8. RMSE of 10-m wind speed (left) and direction (right) for a series of sensitivity tests with WRF v3.0. WRF-simulated sea ice surface temperatures were much warmer than in the SHEBA observations for every tested radiation scheme (not shown), suggesting that a comprehensive treatment for sea ice might be necessary within the WRF model. This necessity was further demonstrated by comparing the surface temperatures simulated by the Arctic MM5 (Zhang and Zhang 2004) and the PSU/NCAR (unmodified) MM5. The Arctic MM5 is based on the PSU/NCAR MM5, with the addition of coupled thermodynamic sea ice (Zhang and Zhang 2001) and mixed layer ocean models (Kantha and Clayson 1994). Comparisons showed that a comprehensive sea ice treatment did help to improve the surface temperature simulation, and the timing of the increase in the downward longwave radiation by the Arctic MM5 matched the SHEBA observations very well (not shown).

19

Nudging Technique Analysis In performing long-term simulations, the use of a nudging technique is essential to ensure that the model does not deviate significantly from reality. There are three different degrees of nudging available in the WRF model: one can nudge everything, including the boundary layer; one can nudge only above a specified model level; or one can limit nudging to the region above the model-defined boundary layer. The latter is beneficial in maintaining an accurate synoptic environment over the course of a long-term simulation, while still allowing the high-resolution WRF model to develop the boundary layer according to its own physics. To assess the sensitivity of a simulation to different nudging levels, three simulation runs were designed: Nudg, which nudges everything above the boundary layer; Nudgall, which nudges everything, including the boundary layer; and Nonudg, which doesn’t nudge anything at all. The results of these simulations were verified against the station observations collected in this project (Section 2.2.1) (Fig. 2.9). The run that included no nudging at all performed significantly more poorly than both of the other two, indicating the necessity of doing at least some nudging to an existing reanalysis. The run that nudged everything performed better, though still somewhat worse than the run that excluded the boundary layer from nudging. This result is very encouraging, and displays the ability of the WRF model to improve upon the NARR reanalysis in producing high-resolution near-surface winds over the Beaufort Sea region.

Figure 2.9a. Station wind speed RMSE for three runs: Nudg, Nudgall, and Nonudg.

Figure 2.9b. Station wind direction RMSE for three runs: Nudg, Nudgall, and Nonudg. 20

2.2.4. Data Assimilation with QuikSCAT Winds Data assimilation is a very useful tool that has the potential to significantly improve modeling results. Data that can be assimilated include conventional surface and upper-air sounding observations, as well as satellite retrievals. The 3DVAR (three-dimensional variational) assimilation approach was used in this study via the use of the software package WRF-Var (WRF Variational data assimilation system). As the region of interest, including both the Beaufort and Chukchi seas, is largely covered by ice/water, a large portion of it is devoid of any kind of in situ measurements, including a lack of buoys due to the coverage of sea ice throughout most of the year. During the open water period, satellite retrievals of the QuikSCAT SeaWinds are available over the ocean, which provides surface wind speed and direction data. Therefore, QuikSCAT ocean-surface winds provide an exciting possibility for filling in the large areas of the domain that are lacking in conventional observations and thus have the potential for improving the simulation of surface winds in the region. In this section, we summarize the assimilation results by comparing simulations using a 10 km grid spacing with and without the assimilation of QuikSCAT winds to quantitatively check to what extent the wind field simulations over the Beaufort Sea region can be improved with the use of data assimilation. Case studies were selected based on an analysis of the ice-free area in the Beaufort Sea and the prevailing wind field. A large ice-free area ensures a good availability of QuikSCAT winds, which are only available over open water. Two types of model runs were conducted: one that assimilated QuikSCAT winds, and one that didn’t (control run). A total of two high wind case studies were simulated, each five days long. Each case was split into four 30-hour periods, resulting in a total of eight independent cases over the two five-day periods. The first 6 hours of each period was used for assimilation of the QuikSCAT data (grouped into one-hour windows), followed by a 24-hour free forecast period, which was then used for verification. Prior to these sensitivity tests, a six-month model run was first conducted using WRF, with the same modeling configuration as the later assimilation runs, to generate a customized background error (BE) file as required by the WRF-Var system. Both the NARR forcing data and the control and QuikSCAT assimilation model runs were verified against both QuikSCAT winds and station observations. The average RMSE and bias for the surface variables, such as wind speed and direction and SLP, were calculated. As shown in Fig. 2.10a, the modeled wind speeds and directions from the QuikSCAT run were improved for the first 12 hours of the free forecast period, suggesting that assimilation of QuikSCAT data has a positive impact on the simulation of wind speed and direction over the open ocean.

21

Figure 2.10a. The RMSE of the 10-m wind speed (left) and direction (right) grouped by forecast hour, for the QuikSCAT assimilation (yellow), control run (red), and the NARR analyses (blue) vs. QuikSCAT data.

Figure 2.10b. RMSE (left column) and bias (right column) of the 10-m wind speed (upper panel), wind direction (middle panel), and SLP (lower panel), grouped by the forecast hour for the QuikSCAT assimilation run (yellow), control run (red), and NARR analyses (blue) vs. station observations.

22

In addition to the domain-averaged station verification, the model forecasts were also verified against the observations at individual stations to examine the geographical distribution of the assimilation impacts. As shown in Fig. 2.10c, wind speed was generally improved in the QuikSCAT assimilation runs relative to the control runs for stations along the Beaufort Sea coast (stations 1–9).

Figure 2.10c. RMSE of the 18-hour forecasted 10-m wind speed (left), direction (middle), and SLP (right) at individual stations for the QuikSCAT assimilation (yellow), control run (red), and NARR analyses (blue). Stations 1–9 are coastal stations from west to east, and 10–14 are inland stations from north to south. Part of the inconsistency between the verification results when comparing with QuikSCAT vs. station data is probably due to the difficulty in accurately interpolating from a 10 km grid over a region with variable topography and/or variability associated with the coastal environment. Over the open ocean, these issues are less applicable, and the interpolated forecast is therefore more representative of the skill of the model. Thus, it is very encouraging to see the consistent improvements in the forecasting of surface winds when verified against QuikSCAT data, which suggests that the analyses are being reliably improved through the assimilation of this data type, at least as far as winds are concerned. 2.2.5. Wind Field Simulations The Beaufort Sea and surrounding areas comprise a complex geographical region, largely covered by sea ice on a seasonal basis over the ocean and bounded by the Brooks Range to the south. Arctic sea breeze effects, due to land/ocean thermal contrasts, and complex orographic effects significantly complicate mesoscale weather systems and the associated surface winds in this region. One of the goals of this study is to improve understanding of the mesoscale weather patterns and associated surface wind features in the Beaufort Sea areas through numerical model simulations. Thus, two sets of month(s)-long numerical simulations with a grid spacing of 10 km were conducted with the WRF model over the domain encompassing the Beaufort Sea region to evaluate the capabilities of the model to simulate the sea breeze-influenced surface wind regime (with a summer case) and the topographic impacts on the surface winds (with a winter case). In addition, throughout the course of a year the Beaufort Sea can be completely or partially covered by ice, or completely ice free. How significantly do these changing surface conditions impact the

23

surface winds? Various surface conditions and their frictional effects on the simulation of surface winds were also investigated to answer this question. Sea Breeze Effects Following the method of Kozo (1979), in order to isolate the sea breeze effects that influence the wind field along the Beaufort Sea coast, we compared the surface 10-m wind direction and 700 hPa geostrophic wind direction for the entire simulation period (June–September 2004). To isolate the sea breeze effects, we focused on coastal stations far away from the Brooks Range, such as Lonely as shown in Fig. 2.11. According to Kozo (1979), the sea breeze can offset geostrophic winds in the boundary layer to produce an average turning of 120° counterclockwise (CCW) from the free stream level to the surface, which increases the frequency of easterly surface winds. The scattered points in the upper left quadrant of Fig. 2.11 represent an approximate 120° CCW turning of the surface winds, representative of the sea breeze effects. Note that the absence of points in the opposing quadrant is due to the lack of a land breeze. This phenomenon is caused by the constant influx of solar radiation during the Arctic summer, which results in a constant positive thermal contrast between land and sea along the Alaskan Beaufort Sea coast.

Figure 2.11. WRF-simulated surface wind directions vs. 700 hPa geostrophic wind directions 10 km north of Lonely for June–September 2004. The land-ocean thermal contrast is a necessary condition for the development of a sea breeze circulation. As such, the surface temperatures over land and ocean were compared at 10 km and 50 km north and south of Lonely (Fig. 2.12a). For the entire simulation period, excepting the beginning of June and most of September, the land was much warmer than the ocean, which provided excellent thermodynamic conditions for the generation of sea breeze circulations. Also, the 120° CCW turning of the wind direction (Fig. 2.12b) strongly correlates to the temperature contrast. During most of September, however, the land-sea temperature contrast is very small due to the onset of snow cover on land. As a result, the surface wind directions are very close to those of the 700 hPa geostrophic winds at that time. A significant turning, though not 120° CCW, occurred in the beginning of June, which might be related to a strong inversion associated with the cold sea ice surface temperatures at the time. 24

Figure 2.12a. Comparisons of the WRF modeled surface temperatures over sea (50 km north of Lonely) and land (50 km south of Lonely) for June–September 2004.

Figure 2.12b. Comparisons of the WRF modeled 700 hPa geostrophic wind direction and surface wind direction 10 km north of Lonely for June–September 2004. Fig. 2.13a depicts the dynamical development of a sea breeze from July 20–21, 2004 along the Lonely coast in the WRF simulation, representative of this common type of circulation. At 16 UTC July 20, 2004 (8 am LDT), a north wind prevails along the south-north cross section from the surface to the higher atmosphere (up to 5 km). As the sun ascends in the sky, the air over the land warms up quickly and significantly, which causes warmed air to begin to rise. On the other side, the air over the nearby seawater (or sea ice) is still very cold and begins to descend. This heating contrast reaches its maximum in the late afternoon, and a sea breeze circulation develops with onshore flow at the surface and offshore flow aloft. Further interaction with flows affected by the Brooks Range can significantly enhance the “sea breeze” circulation and cause it to last longer. The extent of the enhanced onshore flow can be 200 km in both onshore and offshore directions as shown in the modeled horizontal winds parallel to the south-north cross section centered at Lonely, in which positive values represent southerly winds and negative values northerly (Fig. 2.13b).

25

Figure 2.13a. WRF-simulated temperature (color), potential temperature (contour), and wind field circulation (vector) along the south-north cross section centered at Lonely (red stars mark the location of Lonely) at 16 UTC, 22 UTC July 20, and 06 UTC July 21.

Figure 2.13b. WRF-simulated horizontal winds parallel to the south-north cross section centered at Lonely at 00 UTC Jul 21. The red star marks the location of Lonely. Overall, the capability of the WRF model in simulating the sea breeze-influenced surface wind fields for the Beaufort Sea region was confirmed, however much additional quantitative analysis regarding the frequency and extent of the sea breeze circulation and its interactions with the Brooks Range is still required. Topographic Effects To isolate the terrain impacts on the wind flows, two sets of simulation runs done with and without terrain, named Ter and No-ter, respectively, were conducted for a winter case (January 2008), during which both relatively strong cyclonic and high pressure systems moved into the Beaufort Sea region. In the simulation run without the Brooks Range (No-ter), all terrain elevation values were set to zero and the surface pressure was set equal to the SLP. The extrapolated pressure-level data in NARR (which exists down to 1000 hPa) was used to fill in the below-surface atmospheric variables.

26

When a strong cyclone moved over the Beaufort Sea from the south, the presence of terrain had little impact on the cyclone intensity; its central pressure was 982 hPa in No-ter and 983 hPa in Ter (not shown). However, the Brooks Range did weaken the cyclonic winds to the north of the mountains, reducing both the northeast winds behind the cyclone and the southwest winds ahead of it (Fig. 2.14).

300 km

Figure 2.14. Differences in the 10-m wind speed (color) and wind vectors between the No-ter and Ter runs at 08 UTC Jan. 23, 2008, when a cyclone moved over the Beaufort Sea. Another significant weather system that frequently occurs in the Beaufort Sea region during winter is the high pressure system. When a high pressure system was located over the ChukchiBeaufort Seas, the Brooks Range significantly weakened the northeastern winds ahead of the high system, which affected the wind flows to the north of the Brooks Range and the Beaufort Sea coast (Fig. 2.15). Due to the presence of the terrain, wind speeds in the northern Brooks Range and along the Beaufort Sea coast decreased by about 8 m/s compared to those seen without its presence (Fig. 2.15). 300 km

Figure 2.15. Differences of the 10-m wind speed (color) and wind vectors between the No-ter and Ter runs at 18 UTC Jan. 08, 2008, when a high pressure system was located over the Chukchi-Beaufort Seas. The overall impacts of the Brooks Range on the wind fields along the Beaufort Sea coast are further illustrated in Fig. 2.16, in which the modeled surface wind directions from Ter and No-ter 27

offshore from three coastal stations for the entire simulation period of January 2008 are compared. Among the three stations, Lonely is the farthest from the Brooks Range, while Barter Island is the closest. As a result of this, the modeled wind directions in both Ter and No-ter offshore from Lonely are essentially the same, indicating that the Brooks Range has little impact on the wind fields over the western Beaufort Sea coast. However, the coastal winds to the north of Barter Island behave very differently in the two runs. Either easterly or westerly winds dominate in Ter, while there is little directional preference in No-ter, both 10 km and 150 km north of Barter Island, suggesting that the Brooks Range significantly reduces the southerly and northerly winds from near shore out to 150 km offshore. The extent of the impacts of the Brooks Range on the coastal winds north of Cottle Island is weaker than that seen for Barter Island; the wind directions are relatively close between Ter and No-ter 150 km north of Cottle Island.

Figure 2.16. Surface wind direction (WD) in the No-ter run vs. WD in the Ter run at 10 km (upper panel) and 150 km (lower panel) north of Barter Island (left panel), Cottle Island (middle panel), and Lonely (right panel) for January 2008. Surface Frictional Effects Surface conditions also impact the simulation of surface winds, especially the large contrast in the surface roughness between onshore (land) and offshore (ice/water) points. Modeled geostrophic, ageostrophic, and actual winds at Barrow were compared between onshore (10 km south) and offshore (10 km north) points for the entire simulation period (June–September 2004) (Fig. 2.17). There is no systematic difference in the geostrophic winds between onshore and offshore (left panel in Fig. 2.17) points. However, ageostrophic winds over land are generally larger than those over the ocean (middle panel in Fig. 2.17). The maximum ageostrophic wind over land is 20 m/s, while only 10 m/s offshore; this is no doubt due to the relatively larger surface roughness and associated friction velocity of the onshore land points. As a result, the

28

actual wind onshore is generally weaker than that over the offshore point (right panel in Fig. 2.17).

Figure 2.17. Comparison of simulated geostrosphic (left panel), ageostrosphic (middle panel), and actual winds (right panel) at grid points 10 km north (“off shore”) and 10 km south (“on shore”) of Barrow during June–September 2004. 2.3. Conclusions and Discussion The MMS-sponsored Phase I study of Beaufort Sea mesoscale meteorology modeling was conducted, and a summary of the significant results presented herein. Based on a literature review and a model comparison test, the WRF model was selected as the preferred model to be used as the foundation for future development of the Beaufort Sea mesoscale meteorological model. A variety of initial sensitivity experiments testing the selected mesoscale model were subsequently performed. The goal for conducting these sensitivity tests was to gauge the model’s performance and identify the model configuration that will produce the best possible results in simulating atmospheric conditions, and in particular near-surface winds, in the Beaufort Sea coastal region. Among the sensitivity tests conducted with the WRF model, the analysis of sensitivity to forcing data was first performed in order to compare the use of NARR and ERA-40 reanalyses. When these modeling results were verified against station observations, we found that the simulated wind speed driven by ERA-40 slightly outperformed that driven by NARR, especially for the coastal areas. This is a bit of a surprise given that ERA-40 has only ¼ of the spatial and ½ of the temporal resolution of NARR. On the other hand, the simulated wind direction driven by NARR was superior for a longer forecast period. Certainly, a single case study is insufficient to draw any definite conclusions; more case studies are needed for making a better comparison of the performance of these two forcing datasets when used in WRF simulations. The ultimate aim of this study is to produce high-resolution surface forcing for oil spill modeling throughout the Beaufort Sea coastal area. Thus, another factor that needs to be addressed in the configuration of the model is the horizontal grid spacing. In order to test this, we compared two simulations with grid spacings of 5 and 10 km. In verifying the modeling results against station observations, we found that the 5 km simulations were nearly identical to those achieved with a 10 km grid spacing, which suggests that a 10 km spacing is sufficient for successfully simulating surface wind conditions along the Beaufort Sea coast, most likely due to the relatively smooth topographic conditions in the region. Also, though there was no significant improvement in the simulations done with doubled resolution, there naturally were significant computational costs, 29

which make the use of such a high grid resolution difficult to justify as we move on to production runs in the near future. An extensive validation of different parameterizations was conducted to test the efficacy of various physical schemes and their combinations for the Beaufort Sea region. Modeling results were validated against the SHEBA measurements and we found that the choice of cumulus parameterization had a minor impact on the simulations over our study area, an environment that is not favorable to convective activity. Varied performance was found with different radiation schemes and strong interactions were found to exist between the cloud microphysics and radiation. None of these model physics (microphysics, radiation, and cumulus) significantly impacted the surface winds in this Arctic storm simulation, however, which was likely due to the dominating influence of the strong synoptic system. Other simulation periods without strong synoptic systems might help us to further investigate the impacts of these model physics on the simulation of surface wind. Planetary boundary layer (PBL) parameterizations and land surface processes were tested with the newly released WRF v3.0. The ability to use time-varying SST and sea ice (albeit with flagged values (all or none) instead of partial ice concentrations) had a beneficial effect, particularly for the wind direction, suggesting the importance of a realistic description of sea ice and ocean in the model. The unpopular Pleim-Xu land-surface model showed promising results, especially in reducing wind speed error, further demonstrating the importance of model physics tests for application in a particular region. The results for the different PBL schemes were mixed, indicating additional sensitivity tests for longer time periods, using a broader array of observations for verification, are necessary. Additional simulations for the Arctic storm case were also performed with the Arctic MM5 and the standard MM5 (no sea ice and ocean model coupled, as in the WRF model). Comparisons showed that a comprehensive sea ice treatment did help to improve the simulation of surface temperature, as well as of longwave radiation, further suggesting that it is necessary to improve the sea ice and ocean treatment in the WRF model. Use of a nudging technique is essential for performing accurate long-term simulations with the WRF model. Long-term runs with two different nudging levels (nudging above the boundary layer (BL) and nudging everything) were compared. The simulation that only nudged above the BL performed better than the simulation that nudged everything, which demonstrates the ability of the WRF model to improve upon the NARR reanalysis in producing high-resolution nearsurface winds over the Beaufort Sea region. However, how the nudging above the model-defined BL interacts with the simulation of the boundary layer, such as for sea breeze circulations and topographic effects, remains unknown. This is an important consideration for Beaufort Sea mesoscale modeling and needs further study. In addition to the model configuration, another aspect that helps to improve model performance is the assimilation of additional data sources to improve on the existing forcing data. A lack of in situ observations in the Beaufort Sea might hamper this effort. However, QuikSCAT SeaWinds provide an exciting opportunity for improving the simulation of surface winds in the region. Simulations performed when assimilating these data demonstrated very promising effects. While

30

the modeled winds at land-based locations were moderately improved, the simulation of surface winds over the ocean (which is, after all, the ultimate aim of this study) showed a consistent and significant improvement over the first 18 forecast hours. These results are very encouraging and suggest that assimilation of QuickSCAT data should continue to be investigated and has great potential to introduce significant gains to the simulation of surface winds across the Beaufort Sea. The performance of WRF in simulating the overall Beaufort Sea wind field was analyzed, with emphasis placed on evaluating the capabilities of the model to simulate the sea breeze and topographic effects. Overall, WRF performed reasonably well in estimating the surface winds, as well as capturing the timing of the wind shifts and the magnitude of the surface wind speed with minimal significant errors. The capability of the model in simulating the sea breeze-influenced surface wind fields for the Beaufort Sea coast was confirmed. However, much additional quantitative analysis regarding the frequency and extent of the sea breeze circulation and its interactions with the Brooks Range remains to be done, and the comparison of sea breeze circulations between onshore and offshore points and its impacts on the surface wind field are all worth further investigation. In addition, it was found that the Brooks Range exhibits complicated impacts on different weather systems, with significant effects extending out to 150 km offshore, which further emphasizes the important role the Brooks Range plays in Beaufort Sea mesoscale meteorology. In support of model verification and data assimilation, data from in situ observations and satellite retrievals have been collected in the Phase I study. A database including a total of 40 stations over the North Slope and the Beaufort Sea coast, containing hourly in situ observations, was compiled by utilizing data sources from the NCDC, WERC, and MMS. SHEBA field measurements and QuikSCAT SeaWinds over the Beaufort Sea were also collected. QuikSCAT SeaWinds were used for assimilation into the model and the others were used for model verification. However, except for SHEBA measurements, we have little in situ observations offshore, which hampered the verification of wind field simulations over the Beaufort Sea. In addition, QuikSCAT surface winds were also not verified against any observations for the Beaufort Sea region due to the lack of buoys in this area. Even though most verification studies (e.g., Ebuchi et al. 2002; Pickett et al. 2003) show that QuikSCAT surface winds are reliable, the quality does vary among different regions (Satheesan et al. 2007). Thus, further verifying the quality of the QuikSCAT data in the Beaufort Sea region against in situ observations is needed, which requires more effort in collecting offshore in situ observation data from various potential sources. 3. Study Plan for the Beaufort Sea Mesoscale Meteorology Model Study Phase II 3.1. Introduction The preparatory studies carried out in Phase I, as documented above, built a solid foundation for conducting the follow-up Phase II study, which aims to achieve accurate simulation of the Beaufort Sea surface wind and associated mesoscale meteorology and ensure correct assessment of oil spill risk in the Beaufort Sea. The final product from the Phase II study will be a 30-year hindcast simulation of the Beaufort Sea mesoscale meteorology at high spatial (10 km) and temporal (hourly) resolution. 31

There is both scientific and practical significance in producing a dynamically and thermodynamically consistent high-resolution meteorological dataset for the Beaufort Sea region. First of all, in situ observations are extremely sparse spatially and temporally over the Arctic Ocean, including the Beaufort Sea, leaving considerable observational gaps; secondly, while satellite data provides improved coverage, it is only available over a relatively short period of time, and also needs to be validated against other in situ observations to be effective; and finally, the existing state-of-the-art reanalysis datasets generally have coarse resolutions and lack accuracy over the Arctic Ocean (Bromwich et al. 2007). All of these factors have obviously hampered a thorough understanding of the mesoscale meteorology of the Beaufort Sea region, as well as of associated hazardous environmental events, such as extreme wind events, wave surges, coastal flooding, and erosion. The newly-created high-resolution model data produced during the Phase II study will provide an unprecedented opportunity for a detailed analysis of the Beaufort Sea meteorological features. In addition, Beaufort Sea ocean currents also demonstrate complex fine-scale features (e.g., Carmack and Chapman, 2003), consisting of narrow eastward coastal currents, with their associated embedded eddies, and the prominent anticyclonic Beaufort gyre, which are formed by competitive interactions between density gradients and forcing by the surface wind. Small changes or errors in the surface wind forcing can accordingly result in significantly different modeled ocean currents. Figure 3.1 shows an example of the sensitivity of the Beaufort Sea ocean circulation to different wind patterns associated with the Arctic Oscillation (AO). In the AO positive phase, both strong eastward coastal currents, extending from the surface to deep layers, and adjacent westward currents in the subsurface occur. However, when the AO transitions to its negative phase, these currents are substantially weakened (Zhang, 2008). In practice, accurate prediction of oil spill motion, pollutant transport, and wave surges strongly depends upon the ocean circulation state. Considering the high sensitivity of ocean currents to surface wind forcing, there is therefore a pressing need for high quality high-resolution model simulations of the meteorology of the Beaufort Sea.

(a)

(b)

Figure 3.1. Model-simulated 3-dimensional ocean circulation and salinity distribution in the Beaufort Sea, when the Arctic Oscillation (AO) is in its (a) positive; and (b) negative phase. The solid and dashed contours show the zonal velocity and the arrows show the meridional and vertical velocities. The eastward coastal currents and upwelling demonstrate a sensitive response to the surface wind forcing.

32

From the Phase I study, we learned that there are two critical efforts needed for a high quality production simulation of the Beaufort Sea mesoscale meteorology. The first of these is the collection of a more extensive set observational data, including both in-situ observations and satellite retrievals. This is extremely important for our study region, which contains a sparse observational network. The collected data will not only help to calibrate the model’s performance, in order to better tune it for the Beaufort Sea region, but will also be used to improve the simulations themselves through data assimilation. The second essential effort is the improvement of the model through tuning and optimization of various built-in model physics and configuration options, as well as the implementation of more advanced model physics. Certainly, these two efforts are linked as model improvement cannot be undertaken without the collection of observational data. As a result of the preparatory studies conducted in Phase I, we have developed a scientific plan for the follow-up Phase II study that will meet MMS objectives for the accurate specification of the surface wind fields in the Beaufort Sea region, which in turn will ultimately be used for oil spill modeling. The major components of the Phase II study plan include: 1) further data collection, including potential field work, for the offshore open water areas in particular, which will be used for further validation of model performance and improvement of the simulation of Beaufort Sea mesoscale meteorological conditions; 2) further modeling studies for the optimization of the model physics and configuration, including the coupling of a sea ice model, in order to establish a well-tuned Beaufort Sea mesoscale model; 3) a 5-year experimental simulation for validation of modeling results and for use in an oil spill (or wave) test simulation; 4) a 30-year (1978–2008) production simulation, along with an uncertainty assessment of the modeled surface wind fields; and 5) an analysis of the climatology, interannual variability, and long-term change in the features of the modeled Beaufort Sea surface winds. We will discuss the details of each of these components in Sections 3.2–3.7. 3.2. Data Collection, Bias Correction, and Error Analysis Efforts to validate mesoscale model performance and improve oil spill simulation for the Beaufort Sea have been hampered by the lack of sufficient in situ measurements in the region. To better understand the Beaufort Sea mesoscale meteorological environment for improving the simulation of oil spills, further data collection efforts are needed, especially for the offshore open water areas. In Phase II, the following data collection efforts will be conducted to ensure accurate evaluation of the model simulations over the Beaufort Sea region. 3.2.1. In Situ Observational Data Collection In Phase I, we acquired a good deal of land-based station observations along the Beaufort Sea coast, comprising hourly meteorological data compiled from three networks of in situ observational data: MMS, WERC, and NCDC. The periods of record are variable for the available stations and include the parameters of air temperature, humidity, wind speed and direction, precipitation, atmospheric pressure, and solar radiation. This data then served as the basis for the verification of many model sensitivity tests.

33

We also acquired data from the SHEBA field experiment, which provided us with a limited number of observations offshore, and were also used for model verification. However, the SHEBA site was located far from the Alaskan coastline. As a significant part of this project is to study the effects of the sea breeze and the Brooks Range along the Beaufort Sea coast, it is imperative to be able to obtain observations in this region with which to verify the model. We attempted to address this deficiency by collecting QuikSCAT satellite data, which is available over the open ocean (though only to within ~30 km of the coast), for model verification. This is an imperfect solution, however, as satellite data is certainly not equal in quality to in situ observations, and is generally poorer nearer to the more turbulent atmospheric and oceanic circulation patterns associated with coastal regions. It is therefore of great importance to obtain more observational data, especially buoy data, near the Beaufort Sea coast. Additional sources of meteorological data will be sought as part of Phase II and will include the following: (1) Buoy and ship (various transient locations) (2) Radiosonde (vertical profiles from atmospheric soundings – Barrow and Barter Island) (3) CMAN (Coastal Marine Automated Network – Prudhoe Bay) (4) HADS (Hydrometeorological Automated Data System – 7 inland locations) (5) CRN (Climate Reference Network – Barrow) (6) ARM (Atmospheric Radiation Measurement program – Barrow, Atqasuk) (7) Other specific field observations in the Beaufort Sea (1) Buoy and ship data represent transient and short-term data for offshore regions. Buoy data archived by the National Data Buoy Center only extend as far north as the Bering Sea region. The International Arctic Buoy Program (IABP) maintains an archive of circum-Arctic buoy data. These consist of drifting stations deployed in various years, though they are some distance from Alaska’s Arctic coast and it is not clear how usable these data will be for Phase II needs. (2) Data from radiosondes are available for Barrow (1947–present) and Barter Island (1953– 1989). These data are vertical atmospheric profiles measured twice daily (00 and 12 UTC) and include the following parameters: temperature, humidity, wind speed and direction, and pressure with height. The National Climatic Data Center archives these data in the Integrated Global Radiosonde Archive (IGRA) and they have gone through a series of quality control checks. (3) One CMAN station, at Prudhoe Bay, exists within the domain of this study region. The variables of temperature, pressure, and wind speed and direction are observed at this site that has been in operation for a period of two years. Data are archived at the National Data Buoy Center (NOAA). (4) There are seven HADS stations located at various places inland from the Beaufort Sea coast. This network is designed for the hydrological monitoring of rivers and streams, and air temperature is observed at all locations. However, two of these locations observe wind speed and direction and four stations observe precipitation. Most of the sites have data spanning just over one year. (5) NOAA maintains and operates four CRN sites in Alaska; one is located within our study domain (Barrow 4 ENE). This site has been in operation since 2002 and observations include

34

temperature, precipitation, humidity, wind speed and direction, atmospheric pressure, and solar radiation. (6) Two ARM sites are located on Alaska’s North Slope at the western end of the study domain – Barrow and Atqasuk. These stations produce extensive, high frequency, and research-quality meteorological data, though the primary focus is on radiative measurements. In addition, the ARM site at Barrow has a radar wind profiler that can measure horizontal wind, vertical velocity, and virtual temperature using Doppler shift technology. In this regard, a three-dimensional perspective of the atmosphere can be obtained. (7) There are some other ongoing and planned offshore observational programs in the Beaufort Sea. The Environmental Protection Agency (EPA) is planning an offshore meteorological data collection program in Camden Bay, and we have established a partnership with them in order to share access to their collected data. The International Arctic Research Center (IARC) at UAF has an ongoing field observational project named SEDNA (Sea ice Experiment – Dynamic Nature of the Arctic), in which meteorological data are collected. We will also investigate IPY (International Polar Year) field observations made during 2007–2009 in order to identify additional potential meteorological data offshore. 3.2.2. Satellite Data Collection In the Phase I study, we identified several sources of satellite-based observational data (Section 2.2.1) that could potentially be useful for assimilation into the Beaufort Sea mesoscale model that would directly impact wind fields, as well as for verification of model output across this relatively data-sparse region. Among them, only the QuikSCAT SeaWinds data could be readily assimilated by the WRF-Var package, and so this was the only product tested during Phase I. Results of the assimilation sensitivity tests were encouraging, and so as we go forward into Phase II a significant part of our work will include the continued investigation into the use of QuikSCAT data. In a similar manner, other ocean surface wind products, such as those from the SAR, SSM/I, and ERS-2 instruments, as well as the upper-air MODIS polar wind product, will be closely looked into during the Phase II study. In addition to these, the following satellite products will be included in our investigation for potential assimilation into the model: COSMIC soundings COSMIC soundings are satellite-derived atmospheric soundings produced by the Constellation Observing System for Meteorology, Ionosphere and Climate / Formosa Satellite Mission #3 program (COSMIC/FORMOSAT-3) (http://www.cosmic.ucar.edu), a joint Taiwan-U.S. project that was launched in 2006. This program consists of a constellation of six satellites that uses radio occultation techniques via interaction with GPS satellites to measure atmospheric profiles of temperature, pressure, and water vapor across the entire globe, including the data sparse Beaufort Sea region, at very high vertical resolution and with roughly a dozen daily soundings across the study area at a horizontal spacing of ~300 km. This is in contrast to profiles from polar orbiting satellites, which produce data at a higher horizontal but much lower vertical resolution, which in turn result in their lack of ability to sufficiently resolve many vertical atmospheric structures, particularly in the lower atmosphere. The WRF-Var package already contains the

35

capability to ingest COSMIC data into the model, and so this is a particularly intriguing dataset that we plan to pursue in Phase II. TOVS/ATOVS retrievals The (Advanced) TOVS (TIROS [Television Infrared Operational Satellite] Operational Vertical Sounder) system is a set of instruments onboard the NOAA Polar Operational Environmental Satellites (POES) that consists of a High Resolution Infrared Radiation Sounder (HIRS) and two Advanced Microwave Sounding Units (AMSU). Measurements from these instruments are combined to produce global temperature and moisture profiles, among other observation types such as radiation, clouds, and precipitable water, with a horizontal spacing of ~60 km. These data exist from April 1999 to the present. Prior to the current incarnation of this system (ATOVS), the original TOVS used previous-generation instruments onboard older NOAA satellites to generate similar atmospheric profile data from 1979 to 2004. Krieger has previously worked with this data, modifying the MM5 3DVAR system to assimilate it as part of the Arctic Reanalysis Project, and though Fan et al. (2008) found that its assimilation generated mediocre results for various atmospheric parameters, it did produce positive results for surface winds over a pan-Arctic domain. Given the lengthy availability of this dataset, as well as its ability to provide upper-air data for the Beaufort Sea region, which is greatly lacking in radiosonde stations, and its positive impacts on Arctic surface winds in Fan et al. (2008), we plan to devote additional effort to investigating this data source in Phase II and adapting it for use with WRF. SSM/I sea ice The Beaufort Sea is partially covered by multiyear ice near the Canadian Basin, and by first-year ice and land-fast ice seasonally in the shelf regions. The sea ice coverage is highly variable, both spatially and temporally, and therefore an accurate description of sea ice conditions is important for a long-term model simulation. For years we have used the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC) sea ice concentration dataset, which combines Scanning Multichannel Microwave Radiometer (SMMR) and SSM/I observations at a grid cell size of 25 x 25 km, covers the entire Arctic Ocean and adjacent seas, and includes both daily and monthly averages (Cavalieri et al. 2005). This dataset begins in October 1978 and is continuously updated at the National Snow and Ice Data Center (NSIDC; http://nsidc.org/data/nsidc-0051.html). In addition, the Submarine Upward Looking Sonar Ice Draft Profile data archived at NSIDC (http://www.nsidc.org/data/g01360.html) covers a period from 1975 to 2000 and may provide information about ice thickness. We will continue to use these sources of sea ice data as initial and lower boundary conditions in the long-term production simulation. SSM/I and MODIS snow Another important element to be considered in the Beaufort Sea region is the state of snow cover, which can have significant impacts on the surface energy budget, including masking effects of the underlying surface albedo and insulating effects between the ground and atmosphere. In addition, the sea breeze circulation can only develop when the coastal land areas are snow-free. It is thus desirable to have the most accurate snow coverage possible at any given

36

time for use in initializing and updating the model. A high-resolution dataset would be preferable in order to take advantage of the high-resolution land surface present in the model. However, the snow cover fields in the NARR reanalysis, used for forcing our model, are taken from a dataset with a resolution of ~70 km before being interpolated to the 30 km NARR grid, and so these are obviously lacking in fine-scale precision. To address this deficiency, we plan to collect and use three different higher-resolution snow cover datasets produced by the NSIDC, all of which use satellite measurements as their basis. The first one is an SSM/I-measured product, produced in near real-time with a resolution of 25 km (as is the sea ice dataset discussed above), available from May 1995 to the present. The second dataset is the Interactive Multisensor Snow and Ice Mapping System (IMS) Northern Hemisphere snow and ice analysis, which combines observations from three different instruments to produce a human-assisted analysis at 24 km resolution since February 1997, and at 4 km resolution since February 2004. The third one is derived from MODIS measurements onboard the Terra and Aqua satellites, has a resolution of 0.05 degrees (~5 km), and is available from February 2000 onwards. As we generate the 30-year Beaufort Sea mesoscale meteorology dataset, we will use the highest-resolution satellite dataset available at any given time, which should hopefully result in improved simulations for those years when satellite data is available. 3.2.3. Data Bias Correction and Quality Control A critical component to the development of a high value meteorological database is the quality assurance and quality control (QA/QC) procedures performed. Since various observation networks will be used in this project, the data will obviously be subject to different QA/QC techniques, with some datasets simply being raw data (such as we found for the WERC data in Phase I). Metadata describing these techniques exist for each network at the various points of archival. For example, meteorological data from hourly observations archived at the National Climatic Data Center (NCDC, Integrated Surface Hourly Database) have all been subject to a series of standardized QA/QC procedures. These include several dozen algorithms designed to check for data validity, extreme values, internal consistency, and temporal continuity, which are outlined in documentation by the NCDC (Lott). Furthermore, a separate quality control routine specifically designed for hourly wind observations found the quality of these NCDC data to be virtually error free (DeGaetano, 1997; 1998). Error checks also exist which relate mean wind speed and peak gust that have been utilized in additional methods of analysis (Graybeal, 2006). Since wind observations are of specific importance for this study, these documented QA/QC procedures are expected to result in the highest possible data quality. One specific check lacking in the NCDC procedures is the neighbor (or ‘buddy’) check for spatial consistency. This type of check can be quite tricky for stations located in heterogeneous and complex terrain that produces distinct microclimates, during the passage of a weather system, or with discontinuous snow and ice cover. All of these instances represent realistic and physically-based mechanisms for spatial variability. Techniques do exist, however, to check for instances of spatial inconsistency. Data accessible through MADIS (Meteorological Assimilation Data Ingest System, http://madis.noaa.gov) are subject to a spatial consistency check using optimal interpolation, a technique developed by Belousov et al. (1968). This technique calculates the difference between an observation and its nearest neighbors, and then flags those in which

37

the differences are large (greater than a predetermined threshold). It is anticipated that these ‘buddy’ checks will be performed for stations in relatively close proximity, such as the dense network of onshore and offshore sites in the immediate vicinity of Prudhoe Bay. For any newly-collected dataset, we will perform certain QA/QC procedures among those discussed above to ensure a high quality meteorological database for the Beaufort Sea region. 3.2.4. Satellite Data Error Analysis When assimilating any kind of data via the 3DVAR process, it is essential to be able to specify the observational error associated with a given observation type. While QuikSCAT wind data has been verified against ocean buoy data by other groups for other regions of the world, up to this point no one has verified it in the Beaufort Sea region, no doubt in large part due to the lack of any permanent buoys in the area, and so in our initial tests we have used error data calculated for other regions. This may be a good first guess, but it would be beneficial to derive a better error estimate through a comparison between QuikSCAT and buoy data in the Beaufort Sea itself. Assuming that we can acquire buoy data from the EPA or other sources that is sufficiently far enough from shore to be co-located with QuikSCAT observations, which are generally a minimum of 30 km from land, we will therefore perform an error analysis of this data by directly comparing the QuikSCAT to the buoy observations. This estimate of the observational error will then be applied to the 3DVAR assimilation of such data, hopefully serving to improve the quality of the assimilation even further. We will, as for the QuikSCAT data, download a greater amount of data and perform an error analysis on the observations as compared to the ocean buoy data acquired from the Beaufort Sea. In a similar fashion, we will perform an error analysis on any other satellite data products used for assimilation as well. For those that contain surface data, the satellite observations will be directly compared to any available co-located in situ data, whether from land-based surface stations, ocean buoys, or ships. Care must be taken when verifying satellite winds against buoys, as the latter generally measure the winds at 2 meters above the surface, rather than the 10 meters that is typical for other types of measurements. When doing so, the buoy wind data will be extrapolated to the 10-meter level using standard techniques. Upper-air satellite data, such as the COSMIC soundings, will be compared to available radiosonde data in order to determine an error estimate. For all satellite data analyzed, we will calculate the root-mean-square error (RMSE) for all variables to be assimilated and use these error estimates for assimilation via WRF-Var. 3.2.5. Climatological Analysis with Collected Data Climatological characteristics of the wind regime in the Beaufort Sea region will also be determined from the in situ station data once they are quality controlled and compiled in a database; this will be similar to the analysis done for the gridded model data (Section 3.7). Features such as interannual variability, diurnal variability, and the frequency of extreme events will be investigated. Statistics to document these features will include: frequency distributions of the monthly and annual wind speed and direction, frequency and variability of extreme events, and warm season diurnal variability.

38

Basic wind speed and direction climatologies will be generated for each station that contains at least one full year of data, resulting in histograms that display the frequency of occurrence for specific wind speed and direction combinations. As with the gridded datasets, station wind speed data will be viewed in terms of percentile rankings (e.g. W10, W25, W50, W75, W90, Wmax). This will be done both monthly and annually to illustrate seasonal differences in the wind characteristics. For those stations with a longer record (more than 10 years), the changes over time for these histograms will be investigated to look for trends in the wind field. Some emphasis will be placed on cases of extreme high winds and their change over time. One feature that is of particular interest to this project is the sea breeze impact, which was shown to occur in both station and gridded data in Phase I. Further investigation into this phenomenon will be done for Phase II through the creation of climatologies of diurnal variability in wind direction and the corresponding wind speed changes. Special emphasis will of course be given to the snow-free period (mid-June to mid-September), when the maximum thermal gradient between ocean and land is expected. 3.3. Potential Field Work Extensive data collection efforts will be involved in the Phase II study as described in Section 3.2. Unfortunately, however, there are very few resources for collected offshore observations in the Beaufort Sea. The IABP is one source for Arctic buoy observations, but those that exist are relatively far away from the Beaufort Sea coast and are therefore of limited utility in verifying model output, particularly for land-influenced circulations such as the sea breeze and flows impacted by the Brooks Range. SEDNA, which has produced meteorological observations at an ice camp situated on the sea ice in the Beaufort Sea from April 2–16 2007, is another source for offshore observations. While these will be very valuable for calibrating model simulations, they have not collected any measurements during the warm season, when sea breeze circulations occur. There is thus a strong and pressing need to conduct field work for this study, in order to provide a means of satisfactorily validating the model simulations against offshore observations and thereby determine the success of the model in replicating the complete circulation patterns. A design for the deployment of buoys in the Beaufort Sea could take advantage of the dense meteorological network in the Prudhoe Bay area by expanding it outward in a perpendicular transect from the coastline at 30 to 50 km intervals. With the existing Prudhoe Bay network relating near-shore conditions, and the stations farther inland along the Dalton Highway measuring more continental conditions, we could extend this north-south transect into the Beaufort Sea for the purposes of investigating sea breeze and topographic effects on the offshore wind field. The variables we plan to measure include surface temperature, humidity, wind speed and direction, and pressure. Obviously, the more buoys that could be deployed and the more data collected, the more useful the resulting dataset would be. This is simply a preliminary proposal; the ultimate plan would depend in large part on the level of funding that MMS would be willing to contribute to this exercise. Given the utility of having access to such a unique dataset off of the north coast of Alaska, even for just a few winter months, to support the validation of our modeling efforts and

39

to give greater insight into the shoreline circulation patterns, we feel that this is a worthwhile investment for MMS to make in support of this project. 3.4. Optimization and Improvements of Model Physics By the end of the Phase I study we have applied the WRF model (v.3.0) to the Beaufort Sea region. It is the latest incarnation of a community mesoscale model developed by NCAR, allowing for the simulation of atmospheric conditions at a wide range of scales, from grid resolutions of tens of kilometers down to cloud-resolving scales of less than 1 km. It possesses a large suite of physics options that treat various atmospheric and land-surface processes, including microphysics, radiation, subgrid-scale convection, boundary and surface layer processes, and land-surface processes, in addition to having available a simple mixed-layer ocean model. Sea ice coverage (with flag values of 0 or 1) can be updated during a simulation to correspond to available analyses. Included with the model is a data assimilation package that uses 3DVAR techniques to incorporate both conventional and remotely-sensed observations into available analyses in order to produce more accurate initial and lateral boundary conditions, which in turn lead to improved simulations. Sensitivity experiments performed in Phase I (Section 2.2.3) indicated that the choice of cumulus parameterization had a very minor impact on the simulations over the Beaufort Sea region, an environment that doesn’t favor convective activity. Varied performance was found with the use of different radiation schemes, and strong interactions with the cloud microphysics were demonstrated. In general, CAM radiation, while interacting with either the Purdue Lin or WSM 6-class microphysics, produced the best simulations of downward shortwave and longwave radiation. The tests examining sensitivity to the PBL turbulence and land surface parameterizations, two classes of model physics important for modeling atmospheric properties critical for the simulation of surface wind (such as low-level atmospheric stability, vertical wind shear, and surface friction), demonstrated that the PX land surface scheme produced superior simulations of both wind speed and direction, while the Eta PBL simulated the best wind speeds and the ACM2 PBL produced the best wind directions when combined with the NOAH land surface scheme. However, the very best results for both wind speed and direction were found when the ACM2 PBL was combined with the PX LSM, both of which were implemented by the same developers, which suggests that an optimal combination of various model physics schemes does exist when applying the model in a particular region, and that the process of determining this combination is an important effort towards achieving the best possible model performance. As such, the CAM radiation, Purdue Lin and WSM 6-class microphysics, Eta and ACM2 PBL, and PX and NOAH land surface schemes will be the primary candidates considered in Phase II for the optimization of existing WRF model physics in the Beaufort Sea region. A general description of these parameterizations can be found at the end of this subsection. In addition to the physics schemes currently available in WRF, there is a need to implement additional physical packages specifically designed for the unique Arctic environment into the WRF modeling system. To accomplish this, we will first incorporate some efforts (of particular utility for use in the Beaufort Sea region) made by the Polar Meteorology Group at Ohio State University, which primarily include: • Revised cloud/radiation interaction for separating the treatments of the radiative properties of 40

liquid and ice phase cloud particles. • Adapted mixed-phase microphysics developed especially for Arctic stratus clouds. Sensitivity tests with time-varying SST and sea ice conducted in Phase I demonstrated a beneficial effect on the simulation of the surface wind field (Section 2.2.3), suggesting the importance of a realistic description of sea ice and oceanic conditions in the WRF model. However, sea ice is treated as a type of land cover in WRF, which resulted in significant error in the prediction of surface temperature (Zhang et al. 2008). Thus, the second effort that we will conduct is to couple a thermodynamic sea ice model (Zhang and Zhang) to WRF, which will help to improve the prediction of surface temperatures, as well as of the sea ice concentration. In this model, the treatment of sea ice thermodynamics follows Hibler (1979) and Parkinson and Washington (1979), where the sea ice thickness h and concentration A at a grid cell are described by the following equations: ∂h =F (1) h ∂t ∂A = FA (2) ∂t and Fh and FA are the thermodynamic source/sink functions, parameterized following Hibler (1979): (∂h ∂t )0therm F (3) (1 − A) FA = h A + 2h h0 1

0

 ∂h   ∂h  (1 − A) (4) Fh =   A+   ∂t therm  ∂t therm where superscript “0” represents new ice forming over open water while superscript “1” represents growth of existing sea ice, (∂h ∂t )therm is the local rate of sea ice growth or melt and is

determined from the sea ice surface energy budget H n and the turbulent heat flux H w between the ocean and the sea ice: 1  ∂h  (5) = (H w − H n )    ∂t  therm q 0 Sea ice surface temperature Ti is calculated with the Newton/Raphson iterative scheme by linearizing the sea ice surface energy balance. The changes in sea ice thickness and concentration, sea ice temperature, and the flux exchanges between the atmosphere and the surface can thus be predicted with this thermodynamic sea ice model. While determining the optimal model physics and improving the sea ice treatment, the simulation of atmospheric properties, such as low-level atmospheric stability, vertical wind profile, and surface friction, critical for the simulation of surface wind over the Beaufort Sea region, will be our focus in the Phase II study. Additional quantitative analysis regarding the frequency and extent of the sea breeze circulation and its interactions with the Brooks Range will be conducted. The comparison of sea breeze and Brooks Range effects between onshore and offshore points and their impacts on the surface wind field will also be investigated further. Parameterization candidates to be used in the Phase II study 41

The CAM radiation scheme is a spectral-band scheme used in the NCAR Community Atmosphere Model (CAM 3.0) (Collins et al. 2004). The interactions of shortwave and longwave radiation with resolved and fractional clouds are included. The CAM radiation scheme also includes the radiative effects of several trace gases. The Purdue Lin microphysics scheme is a relatively sophisticated microphysics scheme taken from the Purdue cloud model (Chen and Sun 2002). A total of six classes of hydrometeors are included: water vapor, cloud water, rain, cloud ice, snow, and graupel. All parameterization production terms are based on Lin et al. (1983) and Rutledge and Hobbs (1984) with some modifications, including saturation adjustment following Tao et al. (1989) and ice sedimentation. The WSM 6-class microphysics scheme, as does the Purdue Lin microphysics, predicts six categories of hydrometeors: water vapor, cloud water, rain, cloud ice, snow, and graupel (Hong et al. 2004). A diagnostic relation is used for calculating the ice number concentration that is based on ice mass content rather than temperature. The freezing/melting processes are computed during the fall-term sub-steps to increase accuracy in the vertical heating profile of these processes. The saturation adjustment follows Dudhia (1989) and Hong et al. (1998) in separately treating ice and water saturation processes, rather than using a combined saturation as in the Purdue Lin microphysics. The mixed-phase particle fall speeds for the snow and graupel particles are represented by a single fall speed weighted by the mixing ratios for both sedimentation and accretion processes (Dudhia et al. 2008). The PX land surface model is a 2-layer force-restore soil temperature and moisture model (Pleim and Xiu 1995; Xiu and Pleim 2001). The top layer is taken to be 1 cm thick, and the lower layer 99 cm. There are three pathways for moisture fluxes: evapotranspiration, soil evaporation, and evaporation from wet canopies. Evapotranspiration is controlled by bulk stomatal resistance. Grid aggregate vegetation and soil parameters are derived from the fractional coverages of land use categories and soil texture types. There are two indirect nudging schemes that correct biases in 2-m air temperature and relative humidity by dynamic adjustment of soil moisture and deep soil temperature. The NOAH land surface model is a 4-layer soil temperature and moisture model with layer thickness of 10, 30, 60 and 100 cm (adding up to 2 meters) from the top down (Chen and Dudhia 2001; Mitchell et al. 2002; Ek et al. 2003). It includes root zone, evapotranspiration, soil drainage, and runoff, taking into account vegetation categories, monthly vegetation fraction, and soil texture. NOAH also predicts soil ice and fractional snow cover effects, and considers surface emissivity properties. The Eta PBL scheme is a 1.5-order local closure scheme that computes vertical eddy diffusivities based on turbulence kinetic energy (TKE) predicted by a prognostic equation as a function of local vertical wind shear, stability, and turbulence length scale (Mellor and Yamada 1982; Janjic 1990, 1994, 1996, 2002). An upper limit is imposed on the master length scale, depending on the TKE as well as the buoyancy and wind shear. In the unstable range, the functional form of the upper limit is derived from the requirement that the TKE production be nonsingular in the case of growing turbulence. In the stable range, the upper limit is derived from the requirement that

42

the ratio of the variance of the vertical velocity deviation and TKE cannot be smaller than that corresponding to the regime of vanishing turbulence. The ACM2 PBL scheme combines local and non-local closure techniques to account for both small-scale shear driven turbulence and larger-scale convective turbulence (Pleim and Chang 1992; Pleim 2007a,b). As a result, ACM2 produces smoother near-surface profiles compared to non-local closure schemes and more well-mixed profiles compared to local closure (pure eddy diffusion) schemes. Well-characterized turbulence is extremely important for accurately modeling the sea breeze circulation. Thus, in convective conditions the ACM2 can simulate rapid upward transport in buoyant plumes and local shear-induced turbulent diffusion. The partitioning between the local and non-local transport components is derived from the fraction of non-local heat flux according to the model of Holtslag and Boville (1993). The algorithm transitions smoothly from eddy diffusion in stable conditions to the combined local and nonlocal transport in unstable conditions. 3.5. Optimization of Model Configuration: Nudging and Assimilation Strategies Once the various types of observations have been collected and an error analysis performed on each, necessary for their assimilation into the model via 3DVAR, and the model physics have been tuned and upgraded to include necessary schemes important for conditions in the Beaufort Sea region (e.g., addressing the significance of sea ice), the next challenge that must be confronted is how, specifically, the final product, a 30-year Beaufort Sea mesoscale meteorology simulation, will ultimately be produced. The Phase I study (Section 2.2.3) demonstrated that forecast simulations of as little as a few days generate significant errors due to errors in the forcing data and limitations of the model itself. The longer the model runs, the more the errors accumulate. Thus, it is necessary to constrain the model solution in some form, so that it is not allowed to deviate too far from reality. There are generally two ways that this can be accomplished: (1) Nudging the model solution to a known, “correct” reanalysis, such as that provided to the model at initialization. (2) Assimilating additional observations into the model, thereby generating a new reanalysis as was done for the initialization of the model run, and repeating this action whenever additional observations are available until the end of the desired simulation period is reached. In the Phase I study (Section 2.2.3), we conducted experiments comparing nudged model data to that for which no nudging was performed, and found that nudging the model generated significantly superior results relative to allowing the model to develop completely on its own, as is to be expected. It was also found that solely nudging the free atmosphere above the boundary layer resulted in more accurate surface winds than when the entire model domain was nudged, which is also to be expected since one of the primary benefits of using a mesoscale model is in its enhanced resolution. The use of a high-resolution grid allow the model to better represent complex surface properties, such as land use and topography, and consequently more accurately simulate near-surface atmospheric conditions. When the entire domain is nudged to a coarserresolution reanalysis, this advantage is lost and the accuracy of the simulation necessarily suffers.

43

WRF contains three options when nudging to an existing analysis: one can nudge the entire domain, only the area above the model-defined boundary layer, or above a specified model level. While we have demonstrated that the second is superior to the first, it is unknown whether or not using the third option would be better still. Theoretically the boundary layer is completely uncoupled from the free atmosphere above; nudging to the free atmosphere alone should therefore not impact the boundary layer circulations, which is what we are primarily interested in. However, it is unclear if this is in fact the case. In our experiments, we have noted that the model tends to define the boundary layer as being very shallow over the Beaufort Sea, presumably due to the lack of any convective activity there, while over land it is much deeper. When a sea breeze circulation develops, outflow from the rising air on land is transported northward over the ocean, where it subsequently sinks before being drawn onshore near the surface, completing the cycle. A problem thus develops if the air over the ocean that is a part of circulation is being nudged to a coarser analysis in which this circulation is not present, or insufficiently resolved. Nudging with this method could therefore dampen the sea breeze, leading to its inaccurate representation in the model. Thus, it is necessary to investigate the third nudging option, whereby everything above a specified model level is nudged, disregarding how the model chooses to define the boundary layer. If this level is sufficiently high, it should allow all boundary-influenced circulations to develop unimpeded, while at the same time nudging enough of the free atmosphere to constrain the model and maintain an accurate overall synoptic environment. The alternate method to constrain the model solution and reduce the model errors is the usage of a data assimilation technique, i.e., initialize the model with a reanalysis and run it for a short period of time, then perform a 3DVAR assimilation of additional observations, resulting in a new reanalysis, which is then used as the initial condition of the next cycle, and the process is repeated indefinitely. This is the technique that is used in many reanalyses, such as ERA-40 (Uppala et al. 2005) and NARR (Mesinger et al. 2006). While this sounds good in theory, and obviously works in those cases, there is the potential that this method would not be as successful for our particular application. Within the interior of our modeling domain, there are only three major radiosonde stations (Barrow, Kotzebue, and Inuvik, Canada),which is likely an insufficient amount of upper-air data with which to generate a new reanalysis. Due to this lack of observational data, it is possible that the model could in time begin to veer away from the “correct” analysis at locations away from these stations, leading to poor results. One way to counteract this would be to assimilate satellite sounding data, such as from the COSMIC, MODIS, or TOVS instruments. WRF-Var is already capable of ingesting some of these data, but as was found in the Arctic Reanalysis Project (Fan et al. 2008), the usage of MODIS data often leads to unsatisfactory results, probably due to the fact that the data is often very error-prone. To assimilate any new type of observational data into the model, there is a need to make the necessary modifications to the WRF-Var package. This primarily involves modifying the program to suit the new data, a procedure which has been done in the past to assimilate MODIS and TOVS atmospheric profile data via the 3DVAR package for the MM5 model. We will perform a similar procedure on the WRF-Var system. Once the necessary changes are made to the software package and it is able to successfully assimilate the new observation type, we will proceed to perform sensitivity tests to determine whether or not the observations have a positive impact on the simulation results. As discussed above, assimilating extra data (such as MODIS

44

profiles) into an analysis does not necessarily improve the quality of the subsequent simulation for all parameters. It is therefore necessary to test the effect of assimilating each observation type independently of the others, in order to determine whether or not to include it in the final production simulation. Once the final set of observations to be used in the assimilation has been determined, additional sensitivity tests using the entire collection will be performed in order to ensure that the combination of all types generates satisfactory results. Which of the two methods (nudging or assimilation) of producing mesoscale model output is better for the study region cannot be answered at this time. Perhaps a combination of the two, whereby new analyses are constantly generated via 3DVAR, with the model constantly being nudged to the next NARR reanalysis at upper levels, will prove to be best. Ultimately, it will require additional testing of various techniques for longer-term (~1 month) simulation periods, verifying the results against both surface and upper-air observation data, in order to determine which method is superior, and which assimilated data improves the analyses and which does not. 3.6. Validation with Experimental Simulation and Production Simulation Due to a sparse observational network, imperfect model physics, and limited numerical precision, uncertainty and systematic errors cannot be completely avoided in any model or assimilation system. However, we can optimize the data-modeling system by choosing appropriate model physics and a suitable assimilation approach in order to reduce the uncertainties and minimize the errors. Towards this end, we will therefore first conduct a 5-year experimental simulation, which will be a prototype of the final 30-year production simulation. The experimental simulation will be comprehensively validated against the collected observational data. In addition, because of the sensitivity of oil spills to the details of surface wind forcing, small errors in numerical simulations, which may be of minor concern for a weather simulation, may nevertheless lead to erroneous oil spill predictions. Therefore, a test simulation of the oil spill model (or wave model), using the experimental simulation results, is also an important procedure for the model’s validation. The feedback received from the test simulation, as well as the existence of any problems encountered in the validation processes with the observational data will all be taken into account and remedial actions will be taken. Finally, a best-possible production simulation will be produced, including the calculation of assessed uncertainties. This will be made available to users for different applications, e.g., ocean circulation, wave/surge, and oil spill simulations. A flowchart (Fig. 3.2) below shows the step-by-step procedures we will follow for producing a high quality and high resolution Beaufort Sea mesoscale meteorology database. The selection of the period for the 5-year experimental simulation will be determined by which period has the most available observational data for validation. As guided by the World Meteorological Organization (WMO) standard verification system (WMO 2004), the following metrics are defined for the validation of the simulation outputs: the physical parameters to be validated; a set of observations against which the model is evaluated; the space and time domain over which statistics are computed; and the statistical measures used. We will seek to validate the model quantities that are well-represented in our collected observational database as described above. 45

As the surface wind is the most significant parameter in this study, it will naturally be the primary parameter selected for validation. However, since the model is a physically constrained interactive system, bias in the surface wind may also result from model discrepancies related to other parameters in the model. It is imperative to evaluate and understand the overall performance of the model simulations and to monitor their dynamic and thermodynamic consistency. Accordingly, we will utilize a larger set of fundamental parameters for validation, which will include: the surface wind field, surface pressure, surface temperature and specific humidity, as well as the coverage of sea ice and snow.

46

Collect observations (station + satellite)

Observational database

5-year experimental model run

assimilation

Validation of model results

Oil spill or wave test simulation and feedback

Corrective action

30-year production run

Beaufort Sea mesoscale meteorology database Figure 3.2. Flowchart of the step-by-step procedures for producing Beaufort Sea mesoscale meteorology database.

47

A statistical analysis will be employed for validation, which will comprise the temporal/spatial mean, variance, mean square error (MSE), and mean squared skill score (MSSS) for each parameter. In particular, assuming xon and xmn are the observed and modeled time series of the parameter x at either one grid point or all grid points over a domain, and n ( n =1, 2, 3, …, N) are either time steps over the period of validation or grid points over the modeling domain, or both, the temporal and/or spatial means for both are defined as: xo =

1 N

and

∑ xon

xm =

n

1 N

(1)

∑ xmn n

Their variances are given by:

So2 =

1 ∑ xon − x o N n

(

)2

and

Sm2 =

1 ∑ x mn − x m N n

(

)

2

(2)

and MSE is defined as: MSE =

1 N



∑ (xmn − xon )2

(3)

n

2  N  2 MSEc =   So  N −1

(4)

where MSEC is the MSE for observations, in which the observations at the initial time are withheld, and MSSS is defined as: MSSS = 1 −

MSE MSEc

(5)

The statistical significance of MSSS can be calculated by using a standard procedure, such as a Student’s t-test. For the first step, we will examine to what extent the model realistically captures the regional/large scale atmospheric features characterized by the selected parameters described above. The mean annual cycle will be computed by separately averaging the selected parameters from both the model output and the observations located within the modeling domain over each month. Consequently, a pair of 12 modeled and analyzed monthly climatological data sets will be created for each of the selected parameters. Comparisons between the two sets will then be quantified by using the statistical analyses described above, with the overall errors of the simulated 5-year climatological annual cycle to be estimated by MSSS. The primary advantage of producing high-resolution model output in this project is to capture mesoscale atmospheric features. Thus, a majority of the validation work will focus on the

48

capability of the model to simulate surface mesoscale structures. The collected high-resolution satellite data and in situ field observational data will be used for validation. The satellite data and in situ observations exhibit different spatial and temporal characteristics. The former usually provides 2-D coverage over a portion of the entire modeling domain when available. Thus, the MSSS will be computed spatially for those hours when the satellite data is available with coverage of greater than 50% of the modeling domain. An average of the MSSS over the entire simulation time period will also be computed to represent an overall evaluation of the model results. In contrast, in situ observational data are available in time series at irregularly distributed locations. The validation with these data will therefore be conducted by interpolating model data to selected observational locations. The MSSS will then be calculated temporally at each observation site in order to evaluate the model simulation. The validated and remedied model physics, model configuration, and data assimilation approach as described above will thereafter be used for the final 30-year production simulation. 3.7. Climatology, Interannual Variability, and Long-Term Change of Beaufort Sea Surface Winds The new high-resolution (both spatial and temporal) surface wind data from the 30-year production simulation will provide an unprecedented opportunity to enable us to examine the detailed structures of the distribution and evolution of the surface wind across the Beaufort Sea region. We will use this high spatial- (10 km) and temporal- (hourly) resolution surface wind data to examine its climatological features, interannual variability and long-term change. 3.7.1. High Resolution Climatological Features of Beaufort Sea Surface Winds Diurnal and seasonal cycles It has been demonstrated that the use of high-resolution hourly forcing data is critical for ocean models (including wave and oil spill models) because coarse-resolution data filters out high variability and damps wind strength, which can weaken the ocean currents and associated mixing. Using the newly-produced 30-year high-resolution simulation, we will construct the climatological monthly-mean diurnal cycle of the Beaufort Sea surface wind field using the hourly model output at each grid point for each month from January through December, in order to examine the detailed diurnal structures of the Beaufort Sea surface wind distribution and evolution. This construction will present how the diurnal cycle of the averaged surface wind varies in magnitude and direction from the coast to the deep basin over the Beaufort Sea. The climatological diurnal cycle and regional maximum, minimum, and amplitude of the diurnal cycle will be identified. The sea breeze is an integral component of the boundary layer wind circulation over the Beaufort Sea, and it primarily occurs on a daily basis. Using the experience we obtained from modeling the sea breeze circulation in the Phase I study, we will separate the sea breeze from the total wind vector and follow the approach described above to construct a climatological 3-dimensional spatial structure of the sea breeze along selected representative vertical cross-sections, portraying the diurnal cycle of its strength and direction, as well as its contribution to the total wind vector along the Beaufort Sea coastal region. 49

Finally, the climatological seasonal cycle of the Beaufort Sea wind fields will be analyzed, which will also provide a reference for the follow-up examination of interannual variability and longterm change. The climatological monthly-mean surface wind will be calculated at each grid point by averaging over the corresponding time frame. The monthly climatology of the surface wind and sea breeze will be used to examine both their seasonal cycle and the regional distribution. Extreme wind events The analysis of the climatological diurnal and seasonal cycles discussed above provides a fundamental description of the wind field structure in time and space. However, extreme wind events may result in abrupt changes to ocean currents, waves, and surges, causing significant consequences to the transport of oil spills. Therefore, we will augment the analyses to include the climatology of extreme wind events over the Beaufort Sea. Extreme wind events are usually caused by intense synoptic storm systems. We will conduct our analysis by identifying the daily maximum wind speed Wmax at each grid point and then calculating W90, representing the mean of the 90th percentile Wmax for each individual month. Extreme wind events will be defined as those for which Wmax exceeds W90. According to this definition, we will use the 30-year simulation to calculate the frequency and intensity (mean wind speed) of extreme wind events at each grid point for each month, and then construct the monthly climatology, which will provide a view of how often and where extreme wind events occur over the Beaufort Sea during the study period. Meanwhile, we will perform further statistical analyses by using the probability density function (PDF) for the entire Beaufort Sea or selected sub-domains. Wind stress curl The wind stress curl has significant implications for ocean circulations and is regarded as the driving force for the ocean gyre. It generates ocean surface divergence and convergence, and forces upwelling and downwelling ocean water movements, which is especially important for understanding coastal ocean currents. The vertical velocity w associated with Ekman pumping, induced by the wind stress curl, can be expressed as:

  τ w = k •∇× ρf



(1) 

where k is a unit vector in the local vertical direction, τ is the surface wind stress, ρ is the sea water density, and f is the Coriolis parameter. The lack of high-resolution wind data with good spatial and temporal coverage has hampered the evaluation of the wind stress curl and its impacts on ocean circulations over the Beaufort Sea. The newly-modeled dataset will thus make quantifying that process possible for the first time. By using the new data, we will construct the 2-D monthly-mean climatology of the wind stress curl as derived by first-order differencing the wind stress field to provide background information for the oil spill model. In addition, the climatology of the wind curl corresponding to the extreme wind event database will also be constructed.

50

3.7.2. Interannual Variability and Long-Term Change of Beaufort Sea Surface Winds Substantial environmental changes have occurred in the Beaufort Sea region over the past several decades, e.g., an increased storm invasion (e.g., Zhang et al. 2004), a large fluctuation in the surface pressure pattern represented by the Arctic/North Atlantic Oscillation (e.g., Thompson and Wallace 1998), a conspicuous warming of surface air and ocean water temperatures (e.g., Comiso 2006; Shimada et al. 2006), and a drastic retreat of sea ice (e.g., Comiso et al. 2008; Zhang et al. 2008). Changes in the large-scale wind pattern over the Arctic as well as the entire northern hemisphere have been noticed in concert with these environmental changes. However, changes in the high-resolution structure of the surface wind field over the Beaufort Sea have not been well-investigated. We will employ the newly-produced high-resolution data to fill this gap. Diurnal cycle To examine the interannual variability and long-term change of the Beaufort Sea surface wind diurnal cycle, we will employ a composite approach. Specifically, we will first follow the same method described above to sequentially construct three 10-year monthly-mean diurnal cycle climatologies of the Beaufort Sea surface wind, including daily maximum, minimum, and amplitude, at each grid point for each month from January to December. We will then compare changes among those climatologies by calculating their differences. We will also employ a statistical significance test (e.g., the Student’s t-test) to identify where significant changes occur. Through these analyses, we can detect how the daily wind strength and prevailing wind direction may change and how the spatial pattern of the wind may shift in association with synoptic- and large-scale environmental variability and change. Similarly, we will also apply the composite approach to examine changes in the sea breeze diurnal cycle along the selected representative vertical cross-sections. Seasonal winds Based on the climatology constructed above, we will calculate monthly anomalies of the surface wind vector and wind stress curl. The Beaufort Sea region demonstrates distinct seasonality in many aspects of the atmosphere, sea ice, and ocean fields. We will thus group the monthly data into four seasons representing winter (January–March), spring (April–June), summer (July– September), and fall (October–December). We will then make a 2-D vector Empirical Orthogonal Function (EOF) analysis for the surface wind vector and a normal scalar EOF analysis for the wind stress curl in each season. Through these analyses, we will identify the predominant spatial patterns of the surface wind vector and wind stress curl which explain the maximum amount of variance of the variability and change seen during the study time period. The spatial pattern will demonstrate where the wind vector or wind stress curl varies, and which areas predominantly experience large variation. We will also perform a spectral analysis of the amplitude temporal evolution (principal component) of those spatial patterns to detect their major oscillatory and trend components that may play a leading or modulating role in those patterns’ variability and change. Extreme wind events

51

Vector and normal scalar EOF analysis will also be used to investigate the spatial structure, variability, and long-term changes in extreme wind events. As described above for the monthlymean wind fields, we will compute monthly seasonal anomalies of the extreme wind vectors and identify the leading spatial patterns and principal components for each season. Simultaneously, we will conduct a scalar EOF analysis of the anomalies of the frequency and strength of extreme wind events to examine their different perspectives. The EOF analyses will reveal how the leading component of the extreme wind events is distributed spatially in the study area and how it varies with time. In addition, in order to illustrate systematic changes in extreme wind events, we will conduct an analysis by using the PDF, which has been used in previous studies to examine changes in storm frequency and intensity in future warming climate scenarios compared with the present-day climate (Figure 3.3). We will make comparisons among the three 10-year periods by plotting PDFs. The PDF in this planned study will display a distribution of the frequency of the extreme wind events as a function of their intensity for each time window. The shift of the PDF pattern among the compared time periods will provide an insight into how the frequency of the extreme wind events with different intensities changes with time, e.g., whether stronger or weaker wind events occur more frequently later in the 30-year period.

Figure 3.3. The PDF showing the changes in the frequency distribution of maximum storm intensity over the northern hemisphere in winter (December–February) between the 20th and 21st centuries, suggesting that there will be fewer weak storms in a future warming climate. 3.7.3. Physics for Shaping the Wind Field Climatology, Variability, and Long-Term Change Along with the diagnostic and statistical analyses of various aspects of the wind field climatology, variability, and long-term change, we will investigate the physical processes and mechanisms behind them. Synoptic-scale weather patterns, which play a primary steering role in the surface wind distribution and temporal evolution, will be examined first. For example, the Beaufort high is a semi-permanent weather system that generally occupies the Beaufort Sea area, resulting in an anticyclonic surface wind pattern. However, the Beaufort high also demonstrates obvious variability in its spatial positions and intensity, which accordingly causes changes to the surface wind field. Figure 3.4 displays the year-by-year positions of the Beaufort high in January 52

(Zhang, 2008). It generally weakens and contracts toward the North Pacific side in the years when the AO is in its positive phase, while it intensifies and shifts towards the central Arctic Ocean in the years when the AO is in its negative phase. By using the AO index and the 1-sigma criterion for defining the polarized AO positive and negative phases, we will conduct a composite analysis through which we will document how the changes in the Beaufort high, e.g., its weakening and contraction, or its strengthening and expansion, alter the surface wind structure, including its intensity, direction, and the nature of extreme events.

(a) (b) Figure 3.4. The year-by-year positions of the Beaufort high in January when the AO is in its positive (a), and negative phase (b). Cyclonic storms are the other predominant weather systems that invade the Beaufort Sea region. Zhang et al. (2004) indicates that climatologically there are more cyclones in summer than winter, and that there has also been an intensifying trend in cyclone activity over the Arctic. Increased winter invasion of storms into the Beaufort Sea region unquestionably erode the Beaufort high and its associated surface winds. Cyclonic storms occurring over the Beaufort Sea can either originate from the North Pacific Ocean, eastern Siberia, or North America, or are generated locally over the Arctic Ocean. By using the information from the above vector EOF/PC analysis of the seasonal wind field, we will identify peak positive and negative amplitudes of the leading wind field pattern. Composition of storm center locations and intensities in the selected seasons with peak amplitude will be examined to elucidate how the intensity and pathway of cyclonic storms control or modulate the spatial structure and temporal evolution of the Beaufort Sea surface wind. As mentioned above, the physical environment in the Beaufort Sea region has substantially changed in recent decades in the context of global warming, e.g., the dramatic summer sea ice retreat, and an increase in air and water temperatures. These changes also inevitably influence the spatial structure and temporal evolution of the surface wind. To better understand this, we will specifically investigate how this environmental change impacts the primary controlling factors of surface wind, including synoptic-scale weather systems and local mesoscale meteorological processes. We hypothesize that the changed environment weakens the Beaufort high and the local sea breeze circulation, but strengthens cyclonic storm activity. To test this

53

hypothesis, we will specifically examine changes in land-sea and sea-ice thermal contrasts and the associated atmospheric baroclinity, and uncover how these changes influence the persistence, intensity, and position of the Beaufort high, storms, and sea breeze. The land-sea and sea-ice thermal contrasts will be measured by examining the meridional surface air temperature gradients, which will be calculated from the model output. As described above, we will identify the two extreme states of the Beaufort high variation based on the AO index, and the two extreme states of cyclonic storm activity based on the leading wind field pattern. Through the grouping of the computed meridional surface air temperature gradients, sea ice anomalies, and open water sea surface temperature anomalies into the two corresponding extreme states of the Beaufort high and storm activity, we will detect their correlative relationship and reveal the impacts of recent environmental changes. The composite sea breeze along the selected vertical cross-sections as described in Section 3.7.2 will be separated into two groups containing those present with more and less sea ice coverage over the Beaufort Sea, respectively. A comparison will be conducted to examine how the distribution of sea ice and open water impacts the intensity and duration of the sea breeze.

54

4. References Belousov S.L., L.S. Gandin, and S.A. Mashkovich, 1968: Computer Processing of Current Meteorological Data. Ed. V. Bugaev. Meteorological Translation No. 18, 1972, Atmospheric Environment Service, Downsview, Ontario, Canada, 227 pp. Bromwich, D. H., R. L. Fogt, K. E. Hodges, and J. E. Walsh, 2007: A tropospheric assessment of the ERA-40, NCEP, and JRA-25 global reanalyses in the polar regions. J. Geophys. Res., 112, doi:10.1029/2006JD007859. Carmack, E. and D. C. Chapman, 2003: Wind-driven shelf-basin exchange on an Arctic shelf: The joint roles of ice cover extent and shelf-break bathymetry. Geophys. Res. Lett., 30, 1778, doi:10.1029/2003GL017526. Cavalieri D., P. Gloerson, and J. Zwally, 1990, updated 2005: DMSP SSM/I daily polar gridded sea ice concentrations. Edited by J. Maslanik and J. Stroeve. Boulder, CO: National Snow and Ice Data Center. Digital media. Chen, S.-H., 2007: The impact of assimilating SSM/I and QuikSCAT satellite winds on hurricane Isidore simulations. Mon. Wea. Rev., 135, 549–566. Chen, F. and J. Dudhia, 2001: Coupling an advanced land-surface hydrology model with the PSU/NCAR MM5 modeling system. Part I: Model description and implementation. Mon. Wea. Rev., 129, 569–585. Chen, S.-H. and W.-Y. Sun, 2002: A one-dimensional time dependent cloud model. J. Meteor. Soc. Japan, 80, 99–118. Chou M.-D. and M. J. Suarez, 1994: An efficient thermal infrared radiation parameterization for use in general circulation models. NASA Tech. Memo. 104606, 3, 85 pp. Collins, W.D. et al., 2004: Description of the NCAR Community Atmosphere Model (CAM3.0), NCAR Technical Note, NCAR/TN-464+STR, 226 pp. Comiso, J. C., 2006: Abrupt decline in the Arctic winter sea ice cover, Geophys. Res. Lett., 33, L18504, doi:10.1029/2006GL027341. Comiso, J. C., C. L. Parkinson, R. Gersten, and L. Stock, 2008: Accelerated decline in the Arctic sea ice cover, Geophys. Res. Lett., 35, L01703, doi:10.1029/2007GL031972. DeGaetano A. T., 1997: A quality control routine for hourly wind observations. Journal of Atmospheric and Oceanic Technology, 14: 308–317. DeGaetano A. T., 1998: Identification and implications of biases in U.S. surface wind observation, archival, and summarization methods. Theoretical and Applied Climatology, 60: 151–162. Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model, J. Atmos. Sci., 46, 3077–3107. Dudhia, J., S.-Y. Hong, and K.-S. Lim, 2008: A new method for representing mixed-phase particle fall speeds in bulk microphysics parameterizations. J. Met. Soc. Japan, in press. Ebuchi, N., H. C. Graber, and M. J. Caruso, 2002: Evaluation of Wind Vectors Observed by QuikSCAT/SeaWinds Using Ocean Buoy Data. J. Atmos. Oceanic Technol., 19, 2049– 2062. Ek, M. B., M. K. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land model advances in the NationalCenters for Environmental Prediction operational Eta model. J. Geophys. Res., 108(D22), 8851–8867. Fan, X., J. E. Walsh, and J. R. Krieger, 2008: A one year experimental Arctic reanalysis and comparisons with ERA-40 and NCEP/NCAR reanalyses. Geophys. Res. Lett., in press.

55

Graybeal D. Y., 2006: Relationships among daily mean and maximum wind speeds, with application to data quality assurance, International Journal of Climatology, 26: 29–43. Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of the fifth-generation Penn State/NCAR mesoscale model (MM5). NCAR Tech. Note TN-398+STR, 122 pp. [Available from UCAR Communications, P.O. Box 3000, Boulder, CO 80307.] Hibler, W. D. III, 1979: A dynamic thermodynamic sea ice model, J. Phys. Oceanogr., 9, 815– 846. Hong, S.-Y. and H.-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model, Mon. Wea. Rev., 124, 2322–2339. Hong, S.-Y., H.-M. H. Juang, and Q. Zhao, 1998: Implementation of prognostic cloud scheme for a regional spectral model, Mon. Wea. Rev., 126, 2621–2639. Hong, S.-Y., J. Dudhia, and S.-H. Chen, 2004: A Revised Approach to Ice Microphysical Processes for the Bulk Parameterization of Clouds and Precipitation, Mon. Wea. Rev., 132, 103–120. Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with anexplicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341. Janjic, Z. I., 1990: The step–mountain coordinates: physical package. Mon. Wea. Rev, 118, 1429–1443. Janjic, Z. I., 1994: The step-mountain Eta coordinate model: further developments of the convection, viscous sublayer and turbulence closure schemes. Monthly Weather Review, 122, 927–945. Janjic, Z. I., 1996: The Mellor-Yamada level 2.5 scheme in the NCEP Eta Model. 11th Conference on Numerical Weather Prediction, Norfolk, VA, 19–23 August 1996; American Meteorological Society, Boston, MA, 333–334. Janjic, Z. I., 2002: Nonsingular Implementation of the Mellor–Yamada Level 2.5 Scheme in the NCEP Meso model, NCEP Office Note, No. 437, 61 pp. Kain J. S. and J. M. Fritsch, 1990: A One-Dimensional Entraining/Detraining Plume Model and Its Application in Convective Parameterization. J. Atmos. Sci., 47, No. 23, pp. 2784–2802. Kantha, L. H. and C. A. Clayson, 1994: An improved mixed layer model for geophysical applications. J. Geophys. Res., 99, 25235–25266. Key, J. R., D. Santek, C. S. Velden, N. Bormann, J.-N. Thépaut, L. P. Riishojgaard, Y. Zhu, and W. P. Menzel, 2003: Cloud-Drift and Water Vapor Winds in the Polar Regions from MODIS. IEEE Trans. Geosci. Remote Sens., 41, 482–492. Kozo, T., 1979: Evidence for sea breezes on the Alaskan Beaufort Sea coast, Geophysical Research Letters, 6, 849–852. Kozo. T., 1980: Mountain Barrier Baroclinity effects on surface winds along the Alaskan coast, Geophysical Research Letters, 7, 377–380. Kozo, T., 1982: An observational study of sea breezes along the Alaskan Beaufort Sea coast: Part I. J. Appl. Meteor., 21(7), 891–905. Krieger, J. R. and J. Zhang, 2005: Numerical simulation of different complex terrain flows in south-central Alaska: Implication for air pollution transport. 8th Conference on Polar Meteorology and Oceanography, January 9–13, 2005, San Diego, CA. (Available at http://ams.confex.com/ams/pdfpapers/87699.pdf) Lin, Y.-L., R. D. Farley, and H. D. Orville, 1983: Bulk parameterization of the snow field in a cloud model. J. Climate Appl. Meteor., 22, 1065–1092.

56

Lott J. N., The quality control of the integrated surface hourly database, Fourteenth Conference on Applied Climatology, American Meteorological Society, 7.8, 7 pp. Mass, F. C., D. Ovens, K. Westrick, and B. A. Colle, 2002: Does increasing horizontal resolution produce more skillful forecast? Bull. Am. Meteorol. Soc., 83, 407–430. McCumber, M., W.-K. Tao, J. Simpson, R. Penc, and S.-T. Soong, 1991: Comparison of icephase microphysical parameterization schemes using numerical simulations of tropical convection, J. Appl. Meteor., 30, 985–1004. Mellor, G. L. and T. Yamada, 1982: Development of a turbulence closure model for geophysical fluid problems, Reviews of Geophysics and Space Physics, 20(4), 851–875. Mesinger, F. et al., 2006: North American regional reanalysis. Bull. Am. Meteorol. Soc., 87, 343–360. Mitchell, K., M. Ek, D. Lohmann, V. Koren, J. Schaake, Q. Duan, P. Grunmann, G. Gayno, Y, Lin, E. Rogers, D. Tarpley, C. Peters-Lidard, 2002: Reducing near-surface cool/moist biases over snowpack and early spring wet soils in NCEP ETA model forecasts via land surface model upgrades. 16th Conference on Hydrology, Orlando. Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102 (D14), 16663–16682. Pan, H.-L. and W.-S. Wu, 1995: Implementing a Mass Flux Convection Parameterization Package for the NMC Medium-Range Forecast Model. NMC Office Note, No. 409, 40pp. [Available from NCEP/EMC, W/NP2 Room 207, WWB, 5200 Auth Road, Washington, DC 20746-4304] Parkinson, C. L. and W. M. Washington, 1979: A large scale numerical model of sea ice, J. Geophys. Res., 84, 311–337. Pichel, W., F. Monaldo, and J. Nicoll, 2005: SAR-derived coastal winds. Alaska Satellite Facility News and Notes, winter 2005, vol. 2:4. (available at http://www.asf.alaska.edu/publications/newsletter/ASFNNV.2No.4.pdf) Pickett, M. H., W. Tang, L. K. Rosenfeld, and C. H. Wash, 2003: QuikSCAT Satellite Comparisons with Nearshore Buoy Wind Data off the U.S. West Coast. J. Atmos. Oceanic Technol., 20, 1869–1879. Pleim, J. E. and J. S. Chang, 1992: A non-local closure model for vertical mixing in the convective boundary layer. Atm. Env., 26A, 965–981. Pleim, J. E. and A. Xiu, 1995: Development and testing of a surface flux and planetary boundary layer model for application in mesoscale models. J. Applied Meteorology, 34, 16–32. Pleim, J. E., 2007a: Combined Local and Non-local Closure Model for the Atmospheric Boundary Layer. Part 1: Model Description and Testing. J. Appl. Meteor. Clim., 46, 1383– 1395, Pleim, J. E., 2007b: Combined Local and Non-local Closure Model for the Atmospheric Boundary Layer. Part 2: Application and Evaluation in a Mesoscale Model. J. Appl. Meteor. Clim., 46, 1396–1409. Rutledge, S. A. and P. V. Hobbs, 1984: The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. XII: A diagnostic modeling study of precipitation development in narrow cloud-frontal rainbands. J. Atmos. Sci., 20, 2949– 2972. Satheesan, K., A. Sarkar, A. Parekh, M. R. Kumar, and Y. Kuroda, 2007: Comparison of wind data from QuikSCAT and buoys in the Indian Ocean, Int. J. Rem. Sens., 28, 2375–2382.

57

Shimada, K., T. Kamoshida, M. Itoh, S. Nishino, E. Carmack, F. A. McLaughlin, S. Zimmermann, and A. Proshutinsky, 2006: Pacific Ocean inflow: Influence on catastrophic reduction of sea ice cover in the Arctic Ocean, Geophys. Res. Lett., 33, L08605, doi:10.1029/2005GL025624. Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang and J. G. Powers, 2005: A Description of the Advanced Research WRF Version 2, NCAR Tech Note, NCAR/TN–468+STR, 88 pp. [Available from UCAR Communications, P.O. Box 3000, Boulder, CO, 80307]. Available on-line at: http://box.mmm.ucar.edu/wrf/users/docs/arw_v2.pdf) Smirnova, T. G., J. M. Brown, and S. G. Benjamin, 1997: Performance of different soil model configurations in simulating ground surface temperature and surface fluxes. Mon. Wea. Rev., 125, 1870–1884. Smirnova, T. G., J. M. Brown, S. G. Benjamin, and D. Kim, 2000: Parameterization of cold season processes in the MAPS land-surface scheme. J. Geophys. Res., 105 (D3), 4077– 4086. Tao, W.-K., J. Simpson, and M. McCumber 1989: An ice-water saturation adjustment, Mon. Wea. Rev., 117, 231–235. Troen, I. and L. Mahrt, 1986: A simple model of the atmospheric boundary layer: Sensitivity to surface evaporation. Boundary Layer Meteor., 37, 129–148. Thompson, D. W. J., and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime geopotential height and temperature fields, Geophys. Res. Lett., 25, 1297–1300. Thompson, G., R. M. Rasmussen, and K. Manning, 2004: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon. Wea. Rev., 132, 519–542. Uppala, S. M. et al., 2005: The ERA-40 re-analysis. Q. J. Roy. Meteorol. Soc., 131, 2961–3012. Xiu, A. and J. E. Pleim, 2001: Development of a land surface model part I: Application in a mesoscale meteorology model. J. Appl. Meteor., 40, 192–209. Zhang, J. and X. Zhang, 2004: Modeling study of the Arctic storm with the coupled MM5-sea ice-ocean model. The 5th WRF and 14th MM5 Users’ Workshop, Boulder, CO. (Available at http://www.mmm.ucar.edu/mm5/workshop/ws04/PosterSession/Zhang.Jing.pdf) Zhang, J., X. Fan, J. Krieger, M. Shulski, D. Morton, and A. Klene, 2007: Mesoscale Meteorology Model Development and Associated Data Assimilation Efforts, Project Report for MMS Contract 0106CT39787 “Beaufort Sea Mesoscale Meteorology Model Study”, 42 pp. Zhang, J., J. Krieger, and D. Morton, 2008: Beaufort Sea Mesoscale Meteorology Model Evaluation: Initial Sensitivity Analysis, Project Report for MMS Contract 0106CT39787 “Beaufort Sea Mesoscale Meteorology Model Study”, 46 pp. Zhang, X. and J. Zhang, 2001: Heat and freshwater budgets and pathways in the Arctic Mediterranean in a coupled ocean/sea-ice model, J. Oceanography, 57, 207–237. Zhang, X., J. E. Walsh, J. Zhang, U. Bhatt, and M. Ikeda, 2004: Climatology and interannual variability of Arctic cyclone activity: 1948–2002, J. Clim., 17, 2300–2317. Zhang, X., A. Sorteberg, J. Zhang, R. Gerdes, and J. C. Comiso, 2008: Recent radical shifts in atmospheric circulations and rapid changes in Arctic climate system. Science, submitted. Zhang, X., 2008: The Beaufort high and the Beaufort Sea freshwater anomaly in the framework of Arctic Oscillation, to be submitted.

58