Leveraging browse and grazing forage estimates to optimize index-based livestock insurance

African pastoralists suffer recurrent droughts that cause high livestock mortality and vulnerability to climate change. The index-based livestock insurance (IBLI) program offers protection against drought impacts. However, the current IBLI design relying on the normalized difference vegetation index (NDVI) may pose limitation because it does not consider the mixed composition of rangelands (including herbaceous and woody plants) and the diverse feeding habits of grazers and browsers. To enhance IBLI, we assessed the efficacy of utilizing distinct browse and grazing forage estimates from woody LAI (LAI_W) and herbaceous LAI (LAI_H), respectively, derived from aggregate leaf area index (LAI_A), as an alternative to NDVI for refined IBLI design. Using historical livestock mortality data from northern Kenya as reference ground dataset, our analysis compared two competing models for (1) aggregate forage estimates including sub-models for NDVI, LAI (LAI_A); and (2) partitioned biomass model (LAI_P) comprising LAI_H and LAI_W. By integrating forage estimates with ancillary environmental variables, we found that LAI_P, with separate forage estimates, outperformed the aggregate models. For total livestock mortality, LAI_P yielded the lowest RMSE (5.9 TLUs) and higher R² (0.83), surpassing NDVI and LAI_A models RMSE (9.3 TLUs) and R² (0.6). A similar pattern was observed for species-specific livestock mortality. The influence of environmental variables across the models varied, depending on level of mortality aggregation or separation. Overall, forage availability was consistently the most influential variable, with species-specific models showing the different forage preferences in various animal types. These results suggest that deriving distinct browse and grazing forage estimates from LAI_P has the potential to reduce basis risk by enhancing IBLI index accuracy.