The land surface memory - a drought prediction tool

Food security in Africa is a major problem. Approximately 886 million people in Africa rely on agricultureas their main means of survival. They are therefore susceptible to changes in seasonal rains from year to year that can result in agricultural drought. Agricultural drought is determined by low soil moisture content. Soil moisture responds to rainfall, but also depends on many other factors including the soil characteristics and, crucially, on the past soil moisture. Seasonal rainfall forecasts are thus of limited use when attempting to predict agricultural drought because they do not predict the soil moisture, nor take account of the current and past values. Here we demonstrate that predictive skill can be gained from knowledge of the current state of the land surface – how wet or dry the soil is – as the growing season evolves. This skill arises from the land surface memory - the soil moisture content at a particular time depends to a large extent on the historical soil moisture. Reversing this, the state of the soil moisture at a point in the future must depend somewhat on the state of the soil moisture now. By forcing a land surface model with observed data up to a present day and then repeatedly with historical data (to represent the range of possible future conditions) we show that it is possible to be confident of an ensuing agricultural drought 50 days ahead of the end of the growing season. This can be achieved without any rainfall forecast information, but requires the modelled soil moisture to be accurate. The basic premise of the system is that after a certain date we can be confident of an ensuing drought because, no matter how much it rains, the soil moisture will not recover quickly enough to avoid one before the end of the growing season. We also present results from a current operational trial of a forecast system using this method for an undisclosed location in Northern Ghana.