It is known that the zenith path delay (ZPD) is a valuable meteorological parameter. In this paper, ZPD values from. Australian and Malaysian Global Positioning ...
ATMOSPHERIC REMOTE SENSING USING GNSS IN THE AUSTRALASIAN REGION: FROM TEMPERATE CLIMATES TO THE TROPICS Chris Rizos1, Samsung Lim1, Tajul A. Musa2, Sahrum Ses2, Amir Sharifuddin2, Kefei Zhang3 1
School of Surveying & Spatial Information Systems, UNSW, Sydney, NSW 2049, Australia 2 Faculty of Geoinformation Science & Engineering, UTM, 81310 Skudai, Johor, Malaysia 3 School of Mathematical & GeoSpatial Sciences, RMIT, Melbourne, VIC 3001, Australia ABSTRACT
It is known that the zenith path delay (ZPD) is a valuable meteorological parameter. In this paper, ZPD values from Australian and Malaysian Global Positioning System (GPS) reference stations are derived and compared in order to show the climatic contrast between the two regions. Based on the analysis of the two data sets, a higher and shorterterm variability of the estimated ZPD is observed in Malaysia. The quality of the estimated ZPD is examined by benchmarking against International GNSS Service (IGS) troposphere products. The differences range from 4-12mm RMS. Index Terms— meteorology
GPS,
ZPD,
IGS,
troposphere,
1. INTRODUCTION Water vapour – part of the Earth’s atmosphere – is crucial for meteorological processes. Improved knowledge of the water vapour field is vital for understanding atmospheric processes, hydrological cycle and climate system. Moreover, many global meteorological phenomena, such as tropical storms, El Niño and La Niña, require information on the regional and global spatial (in 3-dimensions) and temporal variability of the atmospheric water vapour. The Global Positioning System (GPS) is a versatile and cost-effective tool for the remote sensing of atmospheric water vapour. The GPS signals propagate through the Earth’s atmosphere, and due to neutral molecules in the troposphere layer, the signals are delayed and refracted. There are essentially two types of GPS-based atmospheric remote sensing techniques: ground-based techniques and satellite-based radio occultation techniques. The trend in the establishment of GPS Continuously Operating Reference Stations (CORS) creates a unique opportunity not only for precise positioning applications, but also for ground-based GPS meteorology [1]. The total delay, or Zenith Path Delay (ZPD), can be extracted from the GPS signals after careful data processing. Together with surface pressure and temperature data, the Integrated Water Vapour (IWV) can be inferred from the estimated ZPD – a useful quantity for meteorological applications. Ground-based GPS
meteorology projects such as the Water Vapour Experiment For Regional Operation Network (WAVEFRONT) in Europe, Meteorological Applications of GPS Integrated Column Water Vapour Measurement (MAGIC) in the Western Mediterranean, the GPS Atmospheric Sounding Project (GASP) in Germany, and GPS Meteorology in Japan, have consistently demonstrated that atmospheric water vapour can be estimated to an accuracy commensurate with existing measurement techniques such as water vapour radiometers and radiosondes [1]. In addition, the CORS infrastructure is relatively inexpensive to setup and operate, and provides higher spatial and temporal resolution. GPS radio occultation (RO) is also a very promising atmospheric remote sensing technology. Approximately 2,500 global daily GPS RO events (ROEs) can be obtained as a result of the launch of the six Constellation Observing Systems for Meteorology, Ionosphere, and Climate (COSMIC) Low Earth Orbit (LEO) satellites in April 2006. COSMIC satellites are equipped with GPS receivers, and the analysis of ROEs provides information on both tropospheric and ionospheric conditions. Over 100 of these daily ROEs from COSMIC can be retrieved over Australasia, which represents ten times more observations in comparison with the (single satellite) Challenging Minisatellite Payload for Geophysical Research (CHAMP) mission. The planned new satellite launches in the next few years by countries in South America, Europe and Asia will offer opportunities for geodesists to address critical issues such as climate changes, weather monitoring, cyclones, and drought and environmental problems in general, through satellite-based atmospheric remote sensing techniques, complemented by CORS-based approaches. Furthermore, the opportunity to study the active atmosphere in tropical areas (both the ionosphere and troposphere) over the coming years is extremely exciting. 2. AUSTRALIAN STUDIES Research groups in Australia and Malaysia are cooperating in a number of GPS meteorology projects. This paper discusses the challenges of atmospheric remote sensing in the Australasian region – which stretches from the temperate zones (at latitudes as far south as 45 degrees) to the equator
– and presents some preliminary outcomes. ZPD estimates from two CORS networks – the Australian Regional GPS Network (ARGN) including the Sydney CORS network (SydNET) in the temperate zone, and the Malaysian RealTime Kinematic network (MyRTKnet) in the tropics – have been obtained. Moreover, the International GNSS Service (IGS) stations were included in the ZPD estimation process so as to provide an opportunity to assess the estimated ZPD against the IGS-derived troposphere products. From these ZPD values, animations of ZPD time series were generated within the network coverage through an inverse distance weighting interpolation method. GPS RO results from CHAMP and COSMIC over a two year period (from July 2006 to March 2008) have been obtained, and were evaluated using both numerical weather prediction models and radiosonde data. Studies show that GPS RO-derived Earth’s atmospheric profiles agree well with both modelled data and radiosonde observations in Australasia, across different geographic areas (e.g. low, mid and high latitude regions), and provide useful information on the spatial and temporal error characteristics of the GPS RO data for a number of different scenarios (e.g. seasonal performance).
Figure 1. ZPD interpolation from IGS troposphere products
3. EXPERIMENTAL RESULT Preliminary result of the ZPD interpolation (Figure 1) shows higher magnitude ZPD is found in the tropics, especially in South-East Asia, Brazil and northern Australia. Study by [2] also confirmed that South-East Asia always have higher magnitude ZPD during the monsoon periods. This confirms that research should be focused in this area. Participation of CORS networks from this area plays a very important role in support of GPS meteorology. In addition, more CORS will provide better ZPD estimation, and therefore ultimately benefit both global and local scientific and meteorological communities in this region of the world.
Figure 2. IGS, MyRTKnet and ARGN stations
The MyRTKnet, ARGN and IGS CORS provides an opportunity to utilise the combined CORS network for meteorological applications. Figure 2 shows the locations of the MyRTKnet, ARGN and IGS stations. In this research, five MyRTKnet stations in Peninsula Malaysia operated by the Department of Survey and Mapping Malaysia (DSMM) were selected. These stations were: • GETI in Geting, Kelantan • PEKN in Pekan, Pahang • USMP in Kampus Universiti Sains Malaysia, Pulau Pinang • BANT in Banting, Selangor • JHJY in Johor Jaya, Johor Bahru Selected GPS stations from the ARGN consist of 9 stations maintained by Geoscience Australia: • COCO in Coco Island • KARR in Karratha • DARW in Darwin • TOW2 in Townsville • SYDN in Sydney • MOBS in Melbourne • CEDU in Ceduna • HIL1 in Hillarys • ALIC in Alice Springs In addition four IGS stations were used: • BAKO in Indonesia • NTUS in Singapore • IISC in India • PIMO in the Philippines Data from all 18 stations were processed as a single network first, and then as individual networks. Absolute ZPD can be estimated only for each individual network because the same satellite is seen from different elevation angles at each station. In other words, long baselines contribute to a better separation of correlated parameters of ZPD. The observation data were selected from 1 June to 31 August 2008. Dual-
frequency GPS data (L1 and L2) were used with sampling rate at 30 seconds for 24-hour periods. The GPS data are processed as a single network in postprocessing mode using the Bernese GPS software (version 5.0). In order to obtain high quality ZPD estimates, the data processing parameters and models are appropriately configured, as listed in Table 1. Rapid variations in ZPD are evident as explained by [3]. Table 1. Parameters and models for GPS data processing Parameters Input data Network design Elevation cut-off angle Weighting of the observables Sampling rate Orbits / EOP Station coordinates
Strategy Daily OBS-MAX 3° With cos2 (z), z = zenith angle 30 - 180 s IGS final orbits (SP3) and EOP (Earth Orientation Parameters). Tightly constrained to the ITRF2005 reference frame.
Absolute antenna phase PHAS COD.I05, SATELLIT.I05 center corrections Ocean loading model FES2004 Double-differenced ionosphere-free Ionosphere (IF) linear combination (L3). Ambiguities solution Fixed, resolved using QIF strategy Ionosphere model for Global ionosphere model from CODE ambiguity fixing Horizontal gradient parameters: tilting Gradient Estimation (24 h interval). A-priori Saastamoinen model A-priori model (hydrostatic part) with dry Niell mapping function. Wet-Niell mapping function (1h Mapping Function interval)
Figure 3 shows the time series of estimated ZPD for five MyRTKnet stations (BANT, GETI, JHJY, PEKN and USMP). Table 2 gives the statistics of the time series in mm. According to Figure 3 and Table 2, all MyRTKnet stations produced high mean ZPD values (>2600mm). These stations are located in the equatorial region which is experiencing a high variability and large amount of water vapour. The highest mean ZPD value is from station GETI, followed by BANT, PEKN, USMP and JHJY. There are sudden drops in the estimated ZPD values (see Figure 3) between Day of Year (DoY) 209-215 for all MyRTKnet stations. This situation can be explained by reference to the amount of rainfall at five nearby weather stations operated by the Malaysian Meteorological Department (MMD) in Peninsula Malaysia. From Figure 3, for DoY 209-214, no precipitation was observed at the weather stations. Hence, the lower ZPD values confirm the absence of precipitation.
Figure 3. Time series of estimated ZPD from MyRTKnet Table 2. Statistics of estimated ZPD from MyRTKnet Station Latitude Climate Min Max GETI 6.226 Tropical 2507.1 2709.5 USMP 5.358 Tropical 2492.3 2694.1 PEKN 3.493 Tropical 2507.4 2690.7 BANT 2.826 Tropical 2484.6 2721.9 JHJY 1.537 Tropical 2497.6 2694.8
Mean 2626.0 2617.9 2622.1 2623.8 2616.1
RMS 26.0 30.3 26.5 33.9 30.5
Figure 4. Time series of estimated ZPD from ARGN Table 3. Statistics of estimated ZPD from ARGN Station COCO DARW TOW2
Latitude -12.188 -12.844 -19.269
Climate Tropical Tropical Tropical
Min Max 2423.2 2688.0 2319.3 2581.1 2336.7 2614.9
Mean 2544.4 2439.7 2459.1
RMS 48.0 53.9 57.1
KARR ALIC HIL1 CEDU SYDN
-20.981 Tropical 2305.4 2570.7 -23.67 Arid 2177.5 2362.6 -31.826 Subtropical 2314.6 2550.2 -31.867 Temperate 2299.4 2458.2 -33.781 Temperate 2310.4 2539.6
2378.0 2237.3 2414.6 2367.6 2402.7
51.7 32.7 29.1 25.8 41.8
MOBS
-37.829
2401.6
24.7
Temperate 2321.4 2497.3
In Figure 4, the time series of estimated ZPD for nine ARGN stations (ALIC, CEDU, COCO, DARW, HIL1, KARR, MOBS, SYDN and TOW2) are shown. Table 3 gives the statistics of the time series in mm. The statistics of estimated ZPD in Table 3 shows that the mean value of ZPD estimates for the COCO station is the highest (2544mm) and ALIC station is the lowest (2237mm). This is of course related to the geographical location of the COCO station, located in the tropics, which is experiencing high variability and large amounts of water vapour. In the case of the ALIC station, the location is in an arid region, resulting in the lowest mean ZPD value.
determined. For example, station PIMO has 42.9mm ZPD difference at DoY 170. 4. CONCLUSION In this paper, ZPDs from Australian and Malaysian GPS reference stations were derived and compared in order to show the climatic contrast between the two regions. Based on the analysis of the two data sets, a higher and shorterterm variability of the estimated ZPD is observed in Malaysia, as is expected. The quality of the estimated ZPD is indicated by benchmarking against IGS troposphere products. Challenges for Australia and Malaysia identified during the course of this study include: • To collaborate in producing near real-time IWV and other troposphere products using as many real-time data streaming CORS as possible. • To establish an online GPS meteorology data processing centre that services the region, from the tropics to the temperate climes of Australia. • To encourage other countries within the focus area to participate in GPS meteorology studies by providing access to real-time CORS data to researchers. ACKNOWLEDGEMENTS
Figure 5. The difference of estimated ZPD from IGS stations Table 4. Statistics of the differences in Figure 5 Station
Statistics (mm) Min
Max
Mean
RMS
BAKO
-31.6
27.6
-1.3
7.2
CEDU
-15.9
14.5
-2.0
4.2
COCO
-25.2
38.1
-3.0
7.6
KARR
-21.2
19.2
-2.0
6.0
PIMO
-43.5
42.9
0.1
11.6
SYDN
-30.4
21.4
-3.1
8.0
TOW2
-23.3
19.7
-1.5
6.1
NTUS
-25.1
16.6
-0.2
4.8
The quality of estimated ZPD is examined by comparing ZPD at IGS stations (BAKO, CEDU, COCO, KARR, PIMO, SYDN, TOW2 and NTUS). The difference between the expected and observed values was in the range 43.5 to 42.9mm (see Figure 5) and the RMS was from 4.2 to 11.6mm (see statistics in Table 4). A possible explanation for the main ZPD difference between the expected and observed is due to the different network configuration (in term of network size and design) and processing strategies. A technique to assess the ZPD value needs to be established by exploiting information from the expected IGS ZPDs. From Figure 5, a “worse” ZPD from IGS stations can be
The authors wish to thank the Korea Astronomy and Space Science Institute, the Department of Survey and Mapping Malaysia, Geoscience Australia and the International GNSS service for providing GPS data or troposphere products. The authors also would like to thank Jasmin Kim for contributing Figure 1. REFERENCES [1] Bevis, M., Businger, S., Herring, T.A., C. Rocken., Anthes, R.A. and R. Ware (1992) “GPS Meteorology: Remote Sensing of Atmospheric Water Vapor Using the Global Positioning System”, Journal of Geophysical Research, 97(14):787-801. [2] Musa. T.A., Lim, S. and Rizos, C. (2005) “Low latitude Troposphere: A Preliminary Study Using GPS CORS Data in South East Asia”, U.S. Institute of Navigation National Tech. Meeting, San Diego, California, 24-26 January, 685-693. [3] Lim, S., Musa, T.A. and Rizos, C. (2008) “Application of Running Average Function to Non-dispersive Errors of NetworkBased Real-Time Kinematic Positioning”, Journal of GPS, 7(2): 58-65.
