Evaluation of interpolated rainfall data for weather index insurance spatial basis risk management in the Muda irrigation area, Kedah, Malaysia

Weather index insurance allows small-scale farmers to secure income fluctuations caused by adverse weather conditions. However, the problem of basis risk hinders the demand for weather index insurance by farmers, while the systemic weather risk problem impedes the supply of weather index insurance by insurers. This work aimed to construct a hypothetical weather index insurance that protects small-scale rice farmers against ex-treme weather-induced risks for the Muda Agricultural Development Authority (MADA) in north-western Peninsular Malaysia. We first developed a monthly rainfall dataset over the Muda area for 16 years (2001-2016) using different spatial interpolation methods such as Nearest Neighbour (NN), Inverse Distance Weighting (IDW), Ordinary Kriging (OK), and Kriging with External Drift (KED). The performance of the methods was evaluated and compared in a cross-validation framework using the Root Mean Square Error (RMSE) and Relative Root Mean Square (RRMSE). We then derived var-ious climate indices that include cumulative rainfall (CR) based on interpolated rainfall data and rainfall data from a rainfall station located in the centre of the insured area, cumulative Growing Degree Days (GDD), Standardised Precipitation Index (SPI) at dif-ferent time scales, and average relative humidity. The CR, GDD, and humidity indices were derived at monthly, mid-season and seasonal time scales. The correlation between these indices and seasonal rice yield (e.g, off-season, and main season) at different levels of spatial aggregation was then analysed. A separate hypothetical index insurance con-tract for each season and area was developed, and their risk reduction potential was eval-uated by comparing granary level contracts with sub-granary level contracts and inter-polated-based rainfall contracts with station-based rainfall contracts. We found that the IDW2 method was the best approach to interpolate rainfall data, as evidenced by the RMSE and RRMSE, followed by IDW3, OK, and KED, while the NN method performed the worst. For the results of the correlation analysis, the GDD indices in June were mainly associated with rice yield in the dry off-season across most of the areas. Moreo-ver, cumulative rainfall (i.e, CR_station and CR_interpolated) and SPI1 indices in De-cember were mainly associated with rice yield in the wet main season across most areas. In GDD-based contracts of the dry off-season, the standard deviation reduction (average of all sub-granaries) at the sub-granary level was less than the standard deviation reduc-tion at the granary level by approximately 28%, revealing an increased risk reduction potential at higher levels of spatial aggregation compared to lower levels. In CR-based contracts of the wet main season, a marginal risk reduction potential was achieved at higher levels of aggregation using rainfall-based contracts, compared to a higher risk reduction potential achieved with temperature-based contracts in off-season. The use of the CR_interpolated index did not reduce spatial basis risk at both granary and sub-gran-ary levels. The developed methodology, which analyses spatial basis risk using various interpolation methods, could serve as a foundation for future research on rainfall index insurance. This work could help farmers and policy makers with their risk management plans for rice production to respond to crop losses related to extreme weather.

Text 115752.pdf Download (1MB) Official URL or Download Paper: http://ethesis.upm.edu.my/id/eprint/18248

Actions (login required)

View Item