Mauritius suffered from chronic water shortage problems that had a severe impact on its economy and the well-being of its population. For instance, in 1998-1999, the island faced a drought that resulted in a 40% decrease in sugar production, and a drop of about MUR 2 billion in GDP. On the social level, the water deficit the country faces is a major problem for residents. Water availability in reservoirs and major aquifers are influenced by precipitation regimes, which are in turn affected by large-scale circulation patterns such as the El Niño Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD). In this study we (i) investigated the relationship between both ENSO and IOD with precipitation, (ii) developped an Artificial Neural Network (ANN) for precipitation prediction based on ENSO and IOD, (iii) develop statistical and time-series models for precipitation forecasting, and (iv) conduct a drought analysis based on multiple precipitation deficit variables (duration, severity and inter-arrival time). Monthly precipitation data for the period between 1961 to 2012 for the Vacoas station are used in this study. We found some correlation between average winter ENSO indices and precipitation, while the correlation for summer was negligible. Statistically significant correlation was found between average winter precipitation and IOD index. We also found that the correlations fluctuate over time. With ANNs, we obtained an average winter precipitation prediction accuracy of 86%. Prediction of summer precipitation was less accurate than winter precipitation. Results obtained from ANN were more accurate than those from other statistical techniques, such as linear regression and autoregressive integrated moving average (ARIMA). This may be attributed to the ability of ANNs to capture non-linearity in the system. Both ANNs and regression based models predict winter precipitation with remarkably higher precision than summer precipitation. Standardized Precipitation Index (SPI) is proposed as a simple and effective index that can be used for drought definition. It is a spatially invariant quantity that can be computed to give precipitation deficit at multiple timescales. Based on SPI for six months for Vacoas, we found that drought durations vary between 1 and 9 months with a mean of 2.6 months. The mean interarrival time is 15 months. We also identified the severity of all drought events between 1961 and 2012 and the computed the conditional probability of drought of a certain duration given severity and vice versa. The findings from this study can help in more efficient planning and management of scarce water resources on the island. This study can be extended in several ways: (i) cover longer time periods to investigate possible impact of climate change, (ii) cover more stations across the island, (iii) incorporate other meteorological parameters, and (iv) develop models to predict other key variables such as water levels in reservoirs.
Keywords
Water shortage,Drought,Multivariate,Precipitation,ION index