Satellite Soil Moisture Fields as Simple Predictors of Flood Event Onsets

Floods are the most impactful disaster by annual cost and also the most frequent. Large and disastrous events are still very difficult to accurately predict with appropriate lead times (beyond 1-2 days). In simple words, the dominant drivers for flood events are rainfall and consequently antecedent soil moisture conditions. Satellites measure these variables at global scale with credible accuracies, and over a long enough record period now to begin some trend analysis. In this paper, we investigated the use of satellite soil moisture fields at daily time steps over 7 years from ESA’s SMOS mission and correlated that with the same 7 years of recorded large flood events. The Dartmouth Flood Observatory is collecting data on large flood events since 1985, such as area impacted, duration, damage and type of flood, and since 2010 reported around 800 events.

First results look very promising and indicate that indeed antecedent soil moisture conditions measured by satellites such as SMOS or SMAP, at low spatial but high temporal resolution, can be used as simple predictors of the onset of large flood events. Prediction skill depends on type and magnitude of flood, but generally speaking miss rates are low. More detailed analysis and probably longer soil moisture records are needed to substantiate those first findings but such a simple predictor for large flood events would have a very positive impact on flood forecasting and disaster preparedness.