Abstract:
|
The availability of large-scale open-source remote sensing data, paired with machine learning methods, allows for real-time prediction of poverty and design of index-based livestock insurance (IBLI). A pilot IBLI in 2010 was designed using NDVI (Normalized Difference Vegetation Index), a satellite indicator of “greenness”, for pastoralists in Northern Kenya, where uninsured livestock losses are a major cause of poverty. However, the NDVI index suffers from quality issues such as pixel contaminations from cloud covers. Recently, advances in satellite Solar-Induced chlorophyll Fluorescence (SIF), which detects emissions from photosynthetic activity, holds great promise for real-time monitoring of vegetation and are unaffected by contamination. We propose using the novel SIF data and regime-switching threshold models to predict livestock mortality in Northern Kenya, based on ground-truth economic surveys of household livestock losses. Initial results show substantial improvement compared to traditional models using NDVI vegetation indices.
|