A forecast says where the storm will hit, not what crops will suffer. WxCrop links weather indices, causal analysis and insurance data to estimate losses, to within 2% over years of testing.
Extreme weather events are announced several days in advance, but their impact on agriculture remains difficult to anticipate: the forecast indicates the trajectory of a hurricane, not the crops that will suffer from it or the amount involved. Researchers present WxCrop, an agent based on physically interpretable meteorological indices, scrutinized by causal analysis, and then crossed with agricultural data to predict harvest losses, both for past and anticipated dates.
The system has been demonstrated in Florida on three hazards, hurricanes, frosts and dry heat, each with a calibrated loss model on insurance compensation. The verifications are encouraging: the hurricane index returns to cyclone days recorded with a AUC of 0.98, losses in years away from training are replicated to within 2 per cent, and independent production statistics confirm county ranking (rank correlation of 0.63).
In concrete terms, the user questions the agent and obtains, for each crop group, the loss expressed as a share of the expected value of the season, each figure being sourcing. The authors target public authorities, operators and scientists who need to react quickly after a disaster, and consider the method to be applicable to other hazards and regions.
Read the full article on the original source
Go to the source →