Assessing water availability over peninsular Malaysia using public domain satellite data products

Water availability monitoring is an essential task for water resource sustainability and security. In this paper, the assessment of satellite remote sensing technique for determining water availability is reported. The water-balance analysis is used to compute the spatio-temporal water availability with main inputs; the precipitation and actual evapotranspiration rate (AET), both fully derived from public-domain satellite products of Tropical Rainfall Measurement Mission (TRMM) and MODIS, respectively. Both these satellite products were first subjected to calibration to suit corresponding selected local precipitation and AET samples. Multi-temporal data sets acquired 2000-2010 were used in this study. The results of study, indicated strong agreement of monthly water availability with the basin flow rate (r2 = 0.5, p < 0.001). Similar agreements were also noted between the estimated annual average water availability with the in-situ measurement. It is therefore concluded that the method devised in this study provide a new alternative for water availability mapping over large area, hence offers the only timely and cost-effective method apart from providing comprehensive spatio-temporal patterns, crucial in water resource planning to ensure water security.


Introduction
Remote sensing satellite missions over the past decade has seen great success in generating and sharing high level products, particularly for applications related to environmental management and natural resources. Among them are the TRMM (Tropical Rainfall Measurement Mission) and MODIS (Moderate Resolution Imaging Spectro-radiometer) satellite data products. Numerous studies related water resources using both TRMM and MODIS data data sets as the main input for precipitation and actual-evapotranspiration, respectively [1,2,3,4]. TRMM data were used to estimate recharge potential of fresh water shallow aquifers for their sustainable management of water resources in arid ecosystem. Water balance analyses for planning for water resources have also successfully in studies using TRMM and MODIS along with Gravity Recovery And Climate Experiment (GRACE) satellite data sets [2].
Data processing involved in retrieving precipitation rate from radar data of the TRMM is quite elaborate involving data pre-processing of level 0 data, up to the validation and calibration of the outputs to the corresponding in-situ measurement. Such studies [5,6,7,8] are now considered 4 To whom any correspondence should be addressed matured. Similarly for deriving AET from MODIS data having same data processing trends [9]. In in fact in Malaysia such studies have also noted in [10,11]. Higher level satellite data products are now created using specific algorithms for specific biophysical parameters such rainfall-related information from TRMM and AET from MODIS data products. Usability of such higher level products still require one to address the local validation and calibration for operational applications but avoid the most bulk of data pre-processing where bulk of the uncertainties due to systematic and random errors persist. However, there are less studies addressing the local validation of these satellite biophysical products, apart from devising alocal calibration method. This paper highlights the assessment of TRMM and MODIS higher level products for determining water availability. The focus is emphasized on absolute accuracy of water yeild over large watersehed from multi-temporal data sets, and finally this paper concluded how this study could contribute the digital earth environment.

Test site
The study area comprised all watersheds in peninsular Malaysia. The climate is governed by the yearly alternation of the northeast, southwest monsoons and inter monsoon. The Northeast Monsoon occurs from November till March, and the Southwest Monsoon between May and September. The Northeast Monsoon brings heavy rains and extensive flooding to the east coast of Malaysia Peninsular. The rainfall is dominantly govern by wet and dry climate, and also renewable of water resource that area. The amount of rainfall is much influenced by monsoon system at meso-scale level and orographic convective system and land-sea interactions at local level.

Satellite precipitation data -TRMM products
The TRMM 3B42.v6 three-hourly data were used to generate precipitation information; downloaded from http://disc.sci.gsfc.nasa.gov/site for study period June-2000 to July 2010. The TRMM rain rate data are available at a temporal resolution of three-hourly with spatial resolution of 0.25° x 0.25° latitude-longitude grid [12]. The corresponding rainfall data for representative rain gauge stations within the research area were obtained from the Malaysia Meteorological Service (

Satellite AET -MODIS products
The AET for entire study area for June 2000 to July 2010, were derived using derived using MODIS satellite product NDVI (normalised difference vegetation index). The AET derivation method is based on [13], monthly composite NDVI data used, where the data sources are obtainable from http://neo.sci.gsfc.nasa.gov/.

Data processing for water availability
The downloaded TRMM product sets were transformed into ASCII format files containing both the geographical location coordinates and the rainfall intensity for each location. Subsequently, exported to the *.tiff data format ready for further processing within a GIS platform. We used ArcGIS 9.2 to define coordinate system for TRMM data sets to WGS84 and sub-set to study area (1°-7°N and 99.5°-104.5°E). Then, monthly and yearly rainfall also derived from TRMM data sets over the study area for the period Jul 2000 -Jun 2010.  used to determine the accuracy by root mean square error (RMSE). The calibrated TRMM data sets and the derived AET were then input into the water-balance analysis, where this computation is carried in ArcGIS system. Figure 1 illustrates the regression output of the in-situ (set 1) against the corresponding TRMM data set, best shown with y= 0.9x + c, (R 2 =0.71, P<0.0001, n=1337), where y is calibrated TRMM data set, x= bias coefficient, and c is an offset value =7.9094. An independent test to further validated the calibrated TRMM output yield RMSE+83mm (n=2308).   Figure 2).

Calibration and validation
Water yeild in mm

Conclusion
The devised methodology has provided an undeniably best alternative method, that relies fully on satellite higher-level products. It was also a cost-effective method as these data are from publicdomain satellite sources. In term of the reliability of outputs, it was noted the good agreement between satellite-based and the gauge-based water yield. This allows best analysis of spatio-temporal water availability for water resource planning, in tandem with sustainable development within the watershed. In the larger context of this research, it provides a platform the devised method to the