Current seasonal crop forecast techniques either stem from a rainfall forecast or develop a forecast from historical yields. Making a forecast from simply rainfall predictions is limiting, for many other climatic factors influence crop development. A forecast that derives from a historical relationship between national yields and other climatic indices is also limited. The historical records are both unreliable and corrupted by socio-economic variability.
Explored within is a new approach to seasonal crop forecasting, one derived from crop water-stress for southern Africa. Potential evapotranspiration (PET) simulations from historical temperature, precipitation, and other climatic factors over the period 1961-1994 are compared with calculated available water from precipitation and soil water holding capacity to develop a water-stress index. As water-stress is the primary determinant of yield in water-limited environments such as southern Africa, this forecast methodology attempts to clearly link climate and maize forecasts.
The forecasts utilize global sea surface temperature (SST) and sea level pressure (SLP) readings to anticipate water-stress six months prior to harvest, with a hindcast correlation over 16 seasons of 0.92 for South Africa and 0.62 for Zimbabwe. Over 17 seasons of independent validation, they achieve similarly remarkable success: a correlation of 0.85 between forecast and actual maize water-stress in South Africa, and a correlation of 0.79 for the same relationship in Zimbabwe. Respective RMS errors against validation data are 18% and 10%.
Other discoveries are regional alternatives to both the Southern Oscillation Index (SOI) and the NINO3 index, alternatives that provide a better regional indication of ENSO's influence upon southern African climate. They account for 58% of seasonal maize water-stress for South Africa, and 49% of that variance for Zimbabwe over a 33 year period.
The full text of this paper is available as a [pdf].