Spatial Analysis and GIS Applications for Estimating Monthly Rainfall Totals on Mauritius

Abstract:

Reliable gridded rainfall data are critical for GIS-based climate change impact assessments, water resources planning and management, design of hydraulic works and urban development. Small Island Developing States (SIDS) are highly dependent on rainfall, more sensitive, and have a lower adaptive capacity to climate change than mainland countries yet are poorly studied. Extensive hydrometeorological records exist in Mauritius, offering a unique opportunity to model rainfall distribution and produce high resolution gridded datasets for GIS-based models. Multiple regression is used to model mean annual and monthly rainfall on the island for the period 1997 - 2011 and derive a physical basis for understanding spatial rainfall patterns. The models incorporate latitude, longitude, slope, distance to coast, elevation and their interactions accounting for 68% of the variance in mean annual rainfall and 55-72% of variance in mean monthly rainfall across the island. Spatial trends are removed from observed monthly rainfall totals and ordinary kriging is applied to the residuals. The regression and kriging results are combined to produce a high resolution, physically consistent gridded time-series dataset. Estimate and variance values from each month are then used to calculate 95% confidence interval surfaces. Cross-validation reveals close correspondence between predicted and observed values. This regression kriging approach captures what is currently understood about the spatial and temporal variability of precipitation in this mountainous sub-tropical location, giving us greater confidence in the reliability of the new rainfall estimates.