Integrated modelling software platform development ...

Ecological Modelling 288 (2014) 195–202

Contents lists available at ScienceDirect

Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel

Integrated modelling software platform development for effective use of ecosystem models Guy R. Larocque a,∗ , Jagtar Bhatti b , André Arsenault c a b c

Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, Quebec, QC, Canada Natural Resources Canada, Canadian Forest Service, Northern Forestry Centre, Edmonton, AB, Canada Natural Resources Canada, Canadian Forest Service, Atlantic Forestry Centre, Corner Brook, NL, Canada

a r t i c l e

i n f o

Article history: Received 14 April 2014 Accepted 12 June 2014 Keywords: Modelling platform Numerical analysis Simulation results interpretation Software integration components

a b s t r a c t Ecological modelling is increasing in importance to facilitate the development of sustainable management planning of terrestrial ecosystems and integrate social and economic objectives. As models have become more complex and include many state variables, modelling software platforms have been developed to support development projects. However, few modelling software platforms integrate software components or applications to facilitate the interpretation of simulation results. Recent advances in computing technology and the increasing availability of free or open-source software allow the efficient integration of different software applications to support modelling development efforts and analysis of simulation results. The application AMSIMOD (“Application for the Management of SImulation MODels”), which was developed to manage the ZELIG-CFS gap model, is introduced as an example of an integrated modelling software platform to facilitate and analyze simulation results with the Stand Visualization System and the Quantum GIS geographic information system. Future advancements in the development of this type of modelling platform are discussed. Typical frameworks for model management and analytical applications should have the capacity to manage several models and simulation results at different spatial scales and the ability to generate management scenarios. Also, end users should have access to various types of reporting tools and analytical and numerical processing utilities. Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved.

1. Introduction The development of ecosystem simulation models to predict the dynamics of terrestrial ecosystems is at the heart of ecological modelling. Ecosystem simulation models, through the application of the basic principles of systems analysis, have contributed to improving the understanding of the processes that govern ecosystem dynamics (see Doren et al., 2009; Sierra et al., 2009). They have also been used to predict the impacts of different types of disturbances (Klenner et al., 2000). Due to environmental concerns expressed by the public, ecosystem models are increasingly being used to ensure the sustainable management of terrestrial ecosystems and the integration of social and economic objectives (Kelly et al., 2013). Indeed, it is increasingly recognized that management planning requires predictions of ecosystem dynamics and simulation models are among the most robust tools to achieve this goal (Seidl et al., 2012).

∗ Corresponding author. Tel.: +1 418 648 5791; fax: +1 418 648 5849. E-mail address: [email protected] (G.R. Larocque). http://dx.doi.org/10.1016/j.ecolmodel.2014.06.011 0304-3800/Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved.

As knowledge on ecosystem processes and dynamics has improved, many ecosystem models have become inherently more complex in the representation of nonlinearity in the processes and feedback mechanisms and heterogeneous spatio-temporal variability. However, the causal relationship between model complexity and predictive capacity still remains controversial (Larocque, 2012). Even if model complexity in terms of representation of processes and feedback mechanisms might not generally increase as much as might be anticipated in the forthcoming decades, it is likely that ecosystem models will simulate more processes with different degrees of complexity in the algorithms, contain more state variables representing ecosystem components or processes and will be increasingly integrated. Jakeman and Letcher (2003) identified five types of integration: (1) combined treatments of widely different subjects, such as environmental, economic and social issues; (2) involvement of stakeholders in the modelling process; (3) use of a complex systems perspective merging knowledge from different disciplines; (4) combination of biological, chemical or physical processes and models in a unified system; and (5) linkages of different temporal and spatial scales. In particular, one of the reasons that will justify the intensification

196

G.R. Larocque et al. / Ecological Modelling 288 (2014) 195–202

of model integration is that decision-makers on ecosystem management issues will require knowledge information from different disciplines (Kelly et al., 2013). For effective implementation and use of more complex or large integrated models, there is a need for efficient integrated modelling software platforms. Advances in computer technology in recent decades, including the exponential increase in the performance of computers and progress in software engineering such as modularity or objectoriented designs (e.g., Voinov and Shugart, 2013; David et al., 2013), have led to the development of large simulation systems with efficient execution times and of modelling software platforms to create and manage ecological models (Voinov and Fishwick, 2008; Scheller et al., 2010; Yang et al., 2011; Kelly et al., 2013). Many modelling software platforms have been developed. They were classified in three groups by Lorek and Sonnenschein (1999): software frameworks, modelling tools and simulators. Software frameworks are applications with utilities that facilitate the programming of simulation models using languages, such as C++ or logo, within an integrated development environment. Good examples are ECOSIM (Steenbeek et al., 2013), NetLogo (Wilensky, 1999) and Source Integrated Modelling System (Welsh et al., 2013). Modelling tools consist of applications with user-friendly graphical interfaces to develop models, automate their execution and display results. Good examples include Stella (Richmond, 2004) and Probˇ et al., 2012). Thus, modellers focus more on MoT (Cerepnalkoski modelling concepts than programming. Simulators include applications that are based on the use of a specific model. They do not allow users to change the code, but the values of parameters can be modified for adaptation to specific ecosystem types or conditions. Simulators are best suited to meet the needs of ecologists who do not develop models, but conduct experiments using existing models. A good example is the Universal Simulator (Holst, 2013). In addition to the examples mentioned above, many other efforts resulted in the development of rich and useful modelling platforms that have facilitated modeller’s work by significantly reducing the time required for programming and by facilitating the visualization of simulation results. In particular, the advent of graphical user interfaces has greatly improved the examination of simulation results on high-quality line or scatter graphs or histograms. These user-friendly graphical interfaces are essential for visualizing the patterns of change in the variables of interest. Examining simulation results only at the end of simulation periods is not always sufficient. For instance, examining how predicted species abundance in grassland or forest ecosystems changes over time may allow users to draw inferences on potential successional pathways, as the abundance of some species may grow and decline, oscillate or simply decline. Also, the analysis of ecosystem dynamics over time may allow model users to modify simulation conditions or parameters based on previous results. However, the majority of modelling platforms either facilitated the model development process or provided user-friendly visualization graphic interfaces. As far as can be evaluated from the literature in ecological modelling, there are relatively few integrated modelling software platforms, with different applications integrated within the same application framework to efficiently display simulation results in different scales or formats or conduct complex numerical analysis. For instance, the dynamics of forest ecosystems can be simulated at the individual-tree, sample plot, landscape or regional levels. The possibility of displaying and analyzing these different types of simulation results at different scales using graphics, Geographic Information Systems (GIS) or numerical analysis tools within an integrated environment can greatly simplify the work of modellers and end users (Steenbeek et al., 2013). There are many GIS or numerical analytical applications that can be used independently of each other, but considerable time can be wasted in transferring simulation results between

applications. First, there are compatibility issues between applications, such as data format. Second, the management of different types of simulation results in different output files can quickly become cumbersome, particularly when they are exported to other applications. Third, license issues may prevent software integration, especially for commercial software. However, the increasing availability of sophisticated free or open-source software allows the development of integrated tools. The integrated approach used by Steenbeek et al. (2013), one of the few examples developed specifically for the modelling of terrestrial ecosystems, combines modelling tools and GIS. Due to the variety of ecosystem models available and requirements of model developers and end users, it is unlikely that a single software platform can meet all the needs with respect to model development facilities, visualization of simulation results and numerical analysis. For this reason, there must be a compromise between the level of complexity in the modelling platforms and analytical tools that can be used to manage as many models as possible. In this paper, we introduce an example of an integrated modelling software platform used to facilitate the management of the simulations of a forest succession model and display simulation results. Future directions in the development of integrated modelling software platforms are discussed. 2. Example of an integrated modelling software platform: AMSIMOD AMSIMOD (“Application for the Management of SImulation MODels”) is an integrated modelling software platform that was developed to manage the simulations of the ZELIG-CFS gap model (Larocque et al., 2011a). Even though the software platform was originally designed for the management of ZELIG-CFS, the following functional requirements were considered important to facilitate the use of AMSIMOD for other models in the future: • The platform must be user-friendly, with meaningful menus and submenus strategically located in the main window of the integrated modelling software platform. Therefore, users should not have to read an extensive user’s manual to understand the basic functionalities. • The executable version of the model must be independent of AMSIMOD. Thus, a particular model can be compiled independently using programming languages such as Fortran, C/C++ or C# and be made available to AMSIMOD upon request. • Three groups of users were identified: (i) researchers developing ecosystem models, (ii) decision-makers who are responsible for the management of terrestrial ecosystems and (iii) undergraduate and graduate students who use ecological models to learn about ecosystem functioning or are being educated in modelling. It is assumed that these users need a software platform that will enable them to conduct simulations with ease and visualize results in different formats. • Functional linkages with other applications must be automated as much as possible. 2.1. The ZELIG-CFS model ZELIG-CFS originates from ZELIG, a gap model with the basic structure of JABOWA and FORET, that was originally developed by Urban (1990, 2000) and Urban et al. (1991). It simulates competition between individual trees, the occurrence of single-tree mortality and regeneration. Tree growth and seedling establishment depend on the intensity of light interception, a site fertility factor and monthly precipitation and mean temperature. Basic ecological characteristics, such as minimum and maximum


AMSIMOD (Application for the Management of SImulation MODels)

ZELIG-CFS

Partial cut scenarios

Graphic outputs

Stand Visualization System

Quantum GIS

Fig. 1. A simplified view of AMSIMOD, an integrated modelling software platform that manages simulation runs of the ZELIG-CFS gap model. It includes different modules to modify input files for the simulation of partial cut scenarios, generate graphic outputs to visualize simulation results on line graphs or histograms, and prepare the required information and data for the Stand Visualization System and Quantum GIS.

degree-days or potential maximum dbh growth rate within species-specific area distributions, are defined within the ZELIG structure. ZELIG-CFS was modified considerably by testing and implementing new algorithms on crown interaction effects, mortality and regeneration, which contributed to improving the predictive capacity of the dynamics of North American mixed forest types with complex structures (Larocque et al., 2011a). The mechanistic nature of ZELIG-CFS is characterized by a sufficient degree of complexity to simulate the dynamics of mixed hardwood or conifer forest ecosystems with complex structures. In the forest types examined by Larocque et al. (2011a), there were ecosystems with as many as eight species differing in size, age and social status (e.g., dominant, co-dominant, etc.). ZELIG-CFS is very flexible for evaluating the potential long-term effects of disturbances on forest succession. Forest management is moving towards the development of forests with complex structures and sustainability issues of forest ecosystems are increasing in importance. As a consequence, succession pathways should be considered in forest management planning for periods as long as 200 years (Taylor et al., 2009). 2.2. Description of AMSIMOD The basic structure of AMSIMOD is composed of the main application with five different modules that manage ZELIG-CFS runs, set up the conditions for partial cut treatments, display simulation results on line graphs or histograms and prepare the necessary input files for the Stand Visualization System (SVS) (McGaughey, 1997) and Quantum GIS (www.qgis.org) (Fig. 1). For each sample plot included in a simulation, users can request line graphs in the module graphic outputs to show temporal predictions of state variables, such as basal area, for a maximum of five species in each line graph (Fig. 2). For both SVS and Quantum GIS, AMSIMOD manages the functional linkage. SVS is an application that displays virtual images of stands using a list of individual stand components and tree characteristics (McGaughey, 1997). For each sample plot requested by a user, AMSIMOD generates the necessary input files for SVS. The simulation results are shown in SVS for each simulation year that users may request (Fig. 3). The same principle applies for the linkage with Quantum GIS. After a user identifies a digital map of a forest landscape or region, AMSIMOD creates the layer files to show the location of individual sample plots and the plugin “TimeViewer” in Quantum GIS is activated to show changes in species proportions over time using pie charts (Fig. 4). 3. Future perspectives Different types of software components or supporting applications can be developed or attached to integrated modelling software platforms, such as AMSIMOD, to facilitate the tasks associated with the management of simulation runs, reporting tools and

197

analytical and numerical post-processing utilities (Fig. 5). The term “integrated modelling software platform” is used here to describe a software system that consists of a main application with software components to perform different analytical tasks and manage the execution of independent supporting applications. For instance, AMSIMOD may be viewed as the main application and Quantum GIS as a supporting application that enables users to visualize simulation results in different formats and spatial scales. Both applications can be run independently of each other, but the fact that AMSIMOD creates the input files for Quantum GIS and manages its execution enables users to be very efficient in the management of model outputs. In particular, users do not have to convert simulation results into a format acceptable for Quantum GIS and only need to know the basic functionalities of this software. Thus, integrated modelling software platforms must be designed to be as user-friendly as possible to facilitate the use of complex analytical tools and to automate the management of simulation results and linkages between applications. In other words, end users must spend more time on the development and use of their models than dealing with complex programming or data crunching duties (see Steenbeek et al., 2013). 3.1. Combined used of models Efficient and powerful integrated modelling software platforms should have the capacity to manage the simulation results from several models executed either in sequential or parallel mode. There are more and more integrated modelling projects based on the combined use of different models to meet specific objectives. For instance, Holguin-Gonzalez et al. (2013) integrated hydraulic and physicochemical water quality with aquatic ecological models in their modelling framework for decision support in river management. Lindim et al. (2011) predicted spatial and temporal changes in the Alqueva reservoir, Portugal, using water quality and hydrodynamic models. Johnston et al. (2011) simulated changes in water quality and quantity, biomass of fish populations and habitat suitability in the Albemarle/Pamlico watershed located in Virginia and North Carolina, USA, using five independent models. For these examples, the main objective was to predict the dynamics of ecosystem components that required the use of different independent models, with each one focusing on specific ecosystem components and processes modelled. Integrated modelling efforts can also include models that simulate the same type of state variables or ecosystem components, but with different approaches. For instance, forest ecosystem models can be classified into three major types: empirical, gap and process-based (Larocque, 2008). The three model types can simulate the same state variables based on forest productivity. Each type has strengths and weaknesses, but can be used together in a global management decision-making exercise. The main strength of empirical models is to simulate the effect of silvicultural treatments. Gap models simulate long-term successional pathways and process-based models may predict the impact of disturbances, such as climate change, on forest ecosystem processes, including carbon and nutrient cycles, tree physiology or competition within stands. If used interactively, the three model types can feed each other. For instance, a process-based model that predicts the impacts of climate change on productivity can provide adjustment factors (e.g., negative or positive percentage effect at the species level) on potential long-term impacts of climate change to an empirical model. Within the same model type, there are models that make predictions at the tree, ecosystem, landscape or regional levels. For instance, gap models, such as ZELIG or SORTIE (Canham et al., 1999), predict forest dynamics at the tree and ecosystem levels, while small-scale models, such as LANDIS (Scheller and Mladenoff, 2004), simulate forest succession over large landscapes. Different spatial

198


Fig. 2. Example of a line graph that AMSIMOD users can request to visualize simulation results over time. This example reports simulation results of change in basal area (m2 ha−1 ) over time in a mixed uneven-aged forest in southeastern Canada.

resolution models can interact across a scaling-up exercise. Tree and ecosystem level models can contribute to calibrating or validating regional-level models. Using the same model type, such as process-based models, there are modelling projects that compared the predictions of different models differing in concepts. For the simulation of the dynamics of the same carbon pools in black spruce (Picea mariana [Mill] B.S.P.) forest ecosystems in northern Ontario, Canada, Luckai and Larocque (2002) compared the simulation results of two conceptually different carbon cycle models, CENTURY and Forest-BGC. Both models focus on the dynamics of the carbon cycle, but differ in the representation of processes. This type of exercise allows end users or modellers to determine the degree to which different models compare with reality, assuming that independent datasets are available, or if predictions from different models converge or diverge. In both cases, this kind of exercise allows modellers to identify the strengths, weaknesses or promising approaches in existing models, which may contribute to guiding future modelling research efforts. The examination of model convergence or divergence may be useful in the decisionmaking process during management planning (Larocque et al., 2011b). 3.2. Integrated applications As previously mentioned, a desirable feature of integrated software modelling platforms is to integrate supporting applications that interface with models or allow users to visualize simulation results in different forms or formats. One of the best examples is the combined use of models and GIS, which is appropriate for spatial analysis. A common type of integration consists in using GIS utilities, such as macros or plugins, that include model code and compute spatial or temporal changes in stocks or flows

that are represented in a spatial GIS layer. Wang et al. (2012) used this approach to simulate land use allocation in the TaipeiTaoyuan region of Taiwan. Their model framework was composed of an environmental component for the simulation of energy and material flows and a land component for the simulation of changes in biomass, agricultural lands and urban populations. Steenbeek et al. (2013) used a similar approach to model the food web in the North–Central Adriatic, but their model was first developed using the ecosystem modelling software platform EwE (Ecopath with Ecosim). Another type of integration is to have GIS support within a software modelling platform, such as Swarm (Minar et al., 1996), NetLogo (Wilensky, 1999) or GAMA (Amouroux et al., 2009). GAMA is particularly well advanced with respect to interfacing with a GIS. A close integration between software modelling platforms and GIS presents several possibilities. As indicated above, users of AMSIMOD and ZELIG-CFS can visualize simulation results at the stand and landscape levels over time. A second example of an integrated supporting application is the use of virtual imagery software that creates two- or threedimensional images showing changes in ecosystem components. Several applications were developed for forest ecosystems, such as SVS (discussed above) or The Virtual Forest (Buckley et al., 1998). The software package ADAGE was developed for the analysis of sample plot data with visualization utilities linked to SVS and Pov-Ray (www.povray.org) (Turbis et al., 2002). When closely interconnected with dynamic models, virtual imagery applications may become powerful analytical or educational tools. For instance, AMSIMOD creates the input files for SVS using temporal simulation results. As a consequence, users can launch SVS from AMSIMOD and visualize on three-dimensional images the changes in species composition.


199

Fig. 3. Screenshots showing pictures of simulation results of a mixed uneven-aged forest in southeastern Canada using the Stand Visualization System at years 0 (a), 25 (b), 50 (c) and 100 (d). AMSIMOD creates the input files from simulation results for the different ages of the simulation. Users can visualize the changes over time simply by clicking on “Next”.

3.3. Analytical and numerical processing Simulation projects based on the use of several models and the examination of different management scenarios can produce an enormous amount of data. For end users, utilities that facilitate the analytical and numerical analysis of model results should be available and easy to use. Relatively few modelling software platforms contain analytical and numerical processing utilities. A recent good example is the FRAMES integrated modelling framework (Johnston et al., 2011). The importance of sensitivity and uncertainty analysis for ecological models has been a subject of discussion for a long time. Sensitivity analysis allows modellers or end users to evaluate the performance of models by identifying model limitations, reduce parameter uncertainty or draw conclusions on future improvements (Jakeman et al., 2006; Cariboni et al., 2007; Makler-Pick et al., 2011; Pappas et al., 2013). For models that contain many parameters for which the rate of change has greater influence on model outputs than the other parameters, sensitivity analysis is a mandatory exercise (see Makler-Pick et al., 2011; Pappas et al., 2013). Uncertainty analysis allows a more precise comparison of different ecosystem models because the error estimates in model predictions enable modellers or end users to better evaluate the likelihood of the significance of differences in the predictions

(Hollinger and Richardson, 2005). This statement is supported by the fact that some believe that model predictions without uncertainty estimates are less useful for decision-making than models with uncertainty analysis (Radtke et al., 2001; Allen et al., 2004). Modellers or end users can benefit from the availability of utilities that can facilitate the undertaking of both sensitivity and uncertainty analysis integrated within ecosystem modelling platforms, as both require extensive data manipulations and analysis for end users (Makler-Pick et al., 2011; Larocque et al., 2008). The importance of well designed software to conduct these tasks efficiently must be emphasized, as the lack of funds or difficulty to interpret the results may become limiting factors (Jakeman et al., 2006). The prototype architecture in Fig. 5 suggests that uncertainty analysis is conducted after simulation runs. However, routines that perform uncertainty analysis can be integrated within the source code of models. For instance, when Monte Carlo procedures are used, it is preferable to have them programmed in routines integrated within the source code of models. However, uncertainty analysis can also be conducted after simulation runs using applications, such as GEM-SA (Kennedy et al., 2006) or SIMLAB (Saltelli et al., 2004). Analytical and numerical processing can include utilities to perform statistical or spatial analysis (Fig. 5). If modelling exercises aim at comparing different management scenarios, simulation results can be compared using statistical analytical methods to

200


Fig. 4. Screenshots showing pictures of the location of sample plots and pie charts of species proportions for mixed uneven-aged forests in southeastern Canada using Quantum GIS at years 0 (a), 25 (b), and 100 (c). AMSIMOD creates the layer files for sample plot locations and species proportions from simulation results. A plugin allows users to visualize the changes in species proportions on pie charts over time.


201

Framework for model management and analytical software

Model 1

Model 2

Results

Analytical & numerical processing

Ecosystem level

Sensitivity analysis Uncertainty analysis Statistical analysis Spatial analysis Optimization (linear or nonlinear programming) Artificial intelligence

Landscape level Model n

Regional level

Reporting tools Graphs Geographic Information Systems Virtual imagery applications

Generation of management scenarios

Fig. 5. Prototype architecture of an integrated modelling software platform to manage the simulation runs of several models, coordinate the computation of simulation results at different levels, conduct analytical and numerical processing, report simulation results in different formats and generate simulation scenarios.

evaluate the statistical significance of differences between scenarios. For instance, Wang et al. (2013) performed simulation experiments that consisted in comparing three climate change and seven forest management scenarios using the TRIPLEX model. The simulation outputs generated a lot of time series predictions. Twoway ANOVA was used to evaluate the extent to which differences among the treatments (climate change and management scenarios) were statistically significant. Spatial analysis includes many numerical analysis procedures to study topological, geometric, ecological or geographic properties in small or large areas, including watersheds, landscapes, regions or continents. The majority of the different categories of spatial analyses are facilitated using GIS. In fact, it is likely that GIS advancements have contributed quite significantly to the development of spatial analysis. Even though there is an abundance of literature on the use of spatial analysis and GIS, there are still few examples of applications within integrated modelling software platforms. In the case of ZELIG-CFS and AMSIMOD, the examples mentioned above shows how the linkage with Quantum GIS allows users to visualize simulation results from many sample plot data in different formats over time, including species proportions, over a landscape. Other variables could be illustrated on additional GIS layers, such as species-specific mean dbh, stem height, volume, or biomass. Ecological indicators can also be computed and analyzed using spatial analysis procedures. Additional analytical components can be developed to facilitate spatial analysis, such as impact of changes in species composition over time at the landscape level, considering that partial cut treatments can be simulated at the plot level, and make linkages with topographic, ecological or soil information to show how spatial variation may change over time following simulation results in forest succession. In forest management, mathematical optimization methods, including linear or nonlinear programming, have been used for a long time to optimize partial cut treatments or harvest operations. The integration of these mathematical tools into an integrated modelling software platform would allow end users to be more efficient in the analysis and interpretation of simulation results. For instance, many management scenarios can be considered for large forest landscapes. Optimization utilities can ultimately contribute to determining the best management long-term scenarios in the decision-making process. As previously mentioned, the results

obtained from the simulation of the dynamics of forest ecosystems, including the impact of partial cut treatments, can generate enormous amounts of data, which may confuse end users to the extent that it might be difficult for them to design additional relevant and efficient simulation scenarios. For this reason, artificial intelligence techniques can be developed to generate management scenarios and automate the analysis of all the results. Users could outline general conditions for management scenarios, along with optimization if necessary, that could be applied to different ecosystems within a landscape. 4. Conclusion Many ecological models have been developed in the last few decades and it is likely that many more will be developed in the near future. Advances in computer technology and software have created favourable conditions for the development of integrated modelling software platforms that can manage ecological models and provide end users and modellers with sophisticated visualization tools and numerical and analytical applications to facilitate the analysis of simulation results. The developers of these platforms must keep in mind that end users must spend more time on model development or analysis of simulation results than programming. This may require more complex programming efforts, but may facilitate the work of end users in the development of models or analysis of different potential management scenarios. Acknowledgement Sincere thanks are extended to the International Society for Ecological Modelling (ISEM) for financial support to G.R. Larocque, which allowed him to present an oral communication at ISEM 2013 in Toulouse, France. References Allen, M.R., Booth, B.B.B., Frame, D.J., Gregory, J.M., Kettleborough, J.A., Smith, L.A., Stainforth, D.A., Stott, P.A., 2004. Observational constraints on future climate: distinguishing robust from model-dependent statements of uncertainty in climate forecasting. In: Martin, M., Petit, M., Easterling, D., Murphy, J., Patwardham, A., Hans-Holger, R., Swart, R., Yohe, G. (Eds.), Describing Scientific Uncertainties

202


in Climate Change to Support Analysis of Risk and of Options. National University of Ireland, Maynooth, Ireland, pp. 53–55. Amouroux, E., Chu, T.-Q., Boucher, A., Drogoul, A., 2009. GAMA: An environment for implementing and running spatially explicit multi-agent simulations. In: Ghose, A., Governatori, G., Sadananda, R. (Eds.), Agent Computing and MultiAgent Systems. Lecture Notes in Computer Science 5044. , pp. 359–371. Buckley, D.J., Ulbricht, C., Berry, J., 1998. The Virtual Forest: Advanced 3-D Visualization Techniques for Forest Management and Research. http://proceedings. esri.com/library/userconf/proc98/proceed/to350/pap337/p337.htm Canham, C.D., Coates, K.D., Bartemucci, P., Quaglia, S., 1999. Measurement and modeling of spatially explicit variation in light transmission through interior cedar-hemlock forests of British Columbia. Can. J. For. Res. 29, 1775–1783. Cariboni, J., Gatelli, D., Liska, R., Saltelli, A., 2007. The role of sensitivity analysis in ecological modelling. Ecol. Modell. 203, 167–182. ˇ Cerepnalkoski, D., Taˇskova, K., Todorovski, L., Atanasova, N., Dˇzeroski, S., 2012. The influence of parameter fitting methods on model structure selection in automated modeling of aquatic ecosystems. Ecol. Modell. 245, 135–165. David, O., Ascough II, J.C., Lloyd, W., Green, T.R., Rojas, K.W., Leavesley, G.H., Ahuja, L.R., 2013. A software engineering perspective on environmental modeling framework design: the Object Modeling System. Environ. Modell. Softw. 39, 201–213. Doren, R.F., Richards, J.H., Volin, J.C., 2009. A conceptual ecological model to facilitate understanding the role of invasive species in large-scale ecosystem restoration. Ecol. Indic. 98, S150–S160. Holguin-Gonzales, J.E., Boets, P., Alvarado, A., Cisneros, F., Carrasco, M.C., Wyseure, G., Nopens, I., Goethals, P.L.M., 2013. Integrating hydraulic, physicochemical and ecological models to assess the effectiveness of water quality management strategies for the River Cuenca in Ecuator. Ecol. Modell. 254, 1–14. Hollinger, D.Y., Richardson, A.D., 2005. Uncertainty in eddy covariance measurements and its applications to physiological models. Tree Physiol. 25, 873–885. Holst, N., 2013. A universal simulator for ecological models. Ecol. Inf. 13, 70–76. Jakeman, A.J., Letcher, R.A., 2003. Integrated assessment and modelling: features, principles and examples for catchment management. Environ. Modell. Softw. 18, 491–501. Jakeman, A.J., Letcher, R.A., Norton, J.P., 2006. Ten iterative steps in development and evaluation of environmental models. Environ. Modell. Softw. 21, 602–614. Johnston, J.M., McGarvey, D.J., Barber, M.C., Laniak, G., Babendreier, J., Parmar, R., Wolfe, K., Kraemer, S.R., Cyterski, M., Knightes, C., Rashleigh, B., Suarez, L., Ambrose, R., 2011. An integrated modeling framework for performing environmental assessments: application to ecosystem services in the Albemarle-Pamlico basins (NC and VA, USA). Ecol. Modell. 222, 2471–2484. Kelly, R.A., Jakerman, A.J., Barreteau, O., Borsuk, M.E., ElSawah, S., Hamilton, S.H., Henriksen, H.J., Kuikka, S., Maier, H.R., Rizzoli, A.E., van Delden, H., Voinov, A.A., 2013. Selecting among five common modelling approaches for integrated environmental assessment and management. Environ. Modell. Softw. 47, 159–181. Kennedy, M., Anderson, C., O’Hagan, A., Lomas, M., Woodward, I., Heinemeyer, A., 2006. Quantifying uncertainty in the biospheric carbon flux for England and Wales. In: Research Report No. 564/06. Department of Probability and Statistics, University of Sheffield. Klenner, W., Walton, R., Kurz, W., 2000. Habitats for tomorrow: understanding the consequences of today’s decisions and natural disturbances on future habitat condition. In: Darling, L.M. (Ed.), Proceedings of a Conference on the Biology and Management of Species and Habitats at Risk. 15–19 February 1999, Kamloops, BC Ministry of Environment, Lands and Parks, Victoria, BC and University College of the Caribou, Kamloops, BC, pp. 199–206, Vol. 1. Larocque, G.R., 2008. Forest models. In: Jøgensen, S.E., Fath, B.D. (Eds.), Ecological Models. Vol. [2] of Encyclopedia of Ecology. Elsevier, Oxford, pp. 1663–2167 (5. vols). Larocque, G.R., 2012. Ecological modelling in the 21st century: examining potential research directions and challenges. Proc. Environ. Sci. 13, 331–339. Larocque, G.R., Archambault, L., Delisle, C., 2011a. Development of the gap model ZELIG-CFS to predict the dynamics of North American mixed forest types with complex structures. Ecol. Modell. 222, 2570–2583. Larocque, G.R., Bhatti, J.S., Boutin, R., Chertov, O., 2008. Uncertainty analysis in carbon cycle models of forest ecosystems: research needs and development of a theoretical framework to estimate error propagation. Ecol. Modell. 219, 400–412. Larocque, G.R., Bhatti, J.S., Ascough II, J.C., Liu, J., Luckai, N., Mailly, D., Archambault, L., Gordon, A.M., 2011b. An analytical framework to assist decision makers in the use of forest ecosystem model predictions. Environ. Modell. Softw. 26, 280–288. Lindim, C., Pinho, J.L., Vieira, J.P.M., 2011. Analysis of spatial and temporal patterns in a large reservoir using water quality and hydrodynamic modeling. Ecol. Modell. 222, 2485–2494.

Lorek, H., Sonnenschein, M., 1999. Modelling and simulation software to support individual-based ecological modelling. Ecol. Modell. 115, 199–216. Luckai, N., Larocque, G.R., 2002. Challenges in the application of existing processbased models to predict the effect of climate change on C pools in forest ecosystems. Clim. Change 55, 39–60. Makler-Pick, V., Gal, G., Gorfine, M., Hipsey, M.R., Carmel, Y., 2011. Sensitivity analysis for complex ecological models—a new approach. Environ. Modell. Softw. 26, 124–134. McGaughey, R.J., 1997. Visualizing forest stand dynamics using the stand visualization system. In: Proceedings of the 1997 ACSM/ASPRS Annual Convention and Exposition, April 7–10, 1997, Seattle, WA., Bethesda, MD: American Society of Photogrammetry and Remote Sensing 4, pp. 248–257. Minar, N., Burkhart, R., Langton, C., Askenazi, M., 1996. The Swarm Simulation System: A Toolkit for Building Multi-agent Simulations. Santa Fe Institute, Santa Fe, NM. Pappas, C., Fatichi, S., Leuzinger, S., Wolf, A., Burlando, P., 2013. Sensitivity analysis of a process-based ecosystem model: pinpointing parameterization and structural issues. J. Geophys. Res. Biogeosci. 118, 505–528. Radtke, P.J., Burk, T.E., Bolstad, P.V., 2001. Estimates of the distributions of forest ecosystem model inputs for deciduous forests of eastern North America. Tree Physiol. 21, 505–512. Richmond, B., 2004. An Introduction to Systems Thinking with STELLA. ISEE Systems, Lebanon, NH. Saltelli, A., Tarantola, S., Campolongo, F., Ratto, M., 2004. Sensitivity Analysis in Practice. A Guide to Assessing Scientific Models. John Wiley and Sons Publishers, New York, NY. Scheller, R.M., Mladenoff, D.J., 2004. A forest growth and biomass module for a landscape simulation model, LANDIS: design, validation, and application. Ecol. Modell. 180, 211–229. Scheller, R.M., Sturtevant, B.R., Gustafson, E.J., Ward, B.C., Mladenoff, D.J., 2010. Increasing the reliability of ecological models using modern software engineering techniques. Front. Ecol. Environ. 8, 253–260. Seidl, R., Rammer, W., Scheller, R.M., Spies, T.A., 2012. An individual-based process model to simulate landscape-scale forest ecosystem dynamics. Ecol. Modell. 231, 87–100. Sierra, C.A., Loescher, H.W., Harmon, M.E., Richardson, A.D., Hollinger, D.Y., Perakis, S.S., 2009. Interannual variation of carbon fluxes from three contrasting evergreen forests: the role of forest dynamics and climate. Ecology 90, 2711–2723. Steenbeek, J., Coll, M., Gurney, L., Mélin, F., Hoepffner, N., Buszowski, J., Christensen, V., 2013. Bridging the gap between ecosystem modeling tools and geographic information systems: driving a food web model with external spatial–temporal data. Ecol. Modell. 263, 139–151. Taylor, A.R., Chen, H.Y.H., VanDamme, L., 2009. A review of forest succession models and their suitability for forest management planning. For. Sci. 55, 23–36. Turbis, S., Mailly, D., Pothier, D., 2002. ADAGE: un logiciel d’analyse et de description de placettes d’inventaire forestier. Gouvernement du Québec, Ressources naturelles Québec. Urban, D.L., 1990. A Versatile Model to Simulate Forest Pattern: A User’s Guide to ZELIG Version 1. 0. University of Virginia, Charlottesville, VA. Urban, D.L., 2000. Using model analysis to design monitoring programs for landscape management and impact assessment. Ecol. Appl. 10, 1820–1832. Urban, D.L., Bonan, G.B., Smith, T.M., Shugart, H.H., 1991. Spatial applications of gap models. Forest Ecol. Manage. 42, 95–110. Voinov, A.A., Fishwick, P.A., 2008. Modules in modeling. In: Jøgensen, S.E., Fath, B.D. (Eds.), Ecological Models. Vol. [2] of Encyclopedia of Ecology. Elsevier, Oxford, pp. 2419–2425 (5. vols). Voinov, A., Shugart, H.H., 2013. ‘Integronsters’, integral and integrated modeling. Environ. Modell. Softw. 39, 149–158. Wang, S.-H., Huang, S.-L., Budd, W.W., 2012. Integrated ecosystem model for simulating land use allocation. Ecol. Modell. 227, 46–55. Wang, W., Peng, C., Kneeshaw, D.D., Larocque, G.R., Lei, X., Zhu, Q., Song, X., Tong, Q., 2013. Modeling the effects of varied forest management regimes on carbon dynamics in jack pine stands under climate change. Can. J. For. Res. 43, 469–479. Welsh, W.D., Vaze, J., Dutta, D., Rassam, D., Rahman, J.M., Jolly, I.D., Wallbrink, P., Podger, G.M., Bethune, M., Hardy, M.J., Teng, J., Lerat, J., 2013. An integrated modelling framework for regulated river systems. Environ. Modell. Softw. 39, 81–102. Wilensky, U., 1999. NetLogo. Center for Connected Learning and ComputerBased Modeling, Northwestern University, Evanston, IL http://ccl.northwestern. edu/netlogo/ Yang, J., He, H.S., Shifley, S.R., Thompson, F.R., Zhang, Y., 2011. An innovative computer design for modeling forest landscape change in very large spatial extents with fine resolutions. Ecol. Modell. 222, 2623–2630.