Forecasting wheat yields using a water budgeting model

Abstract

A model has been developed to forecast soil water variations and wheat crop growth in dry land situations. The forecast of the yield to be expected if sowing occurred today is obtained by running the calibrated model for all years for which meteorological data are available. The soil water content on today's date in each year is fixed at today's observed soil water value. From each year of observed meteorological data, an estimate is made of the yield. These yield data allow construction of a frequency distribution of yield which can be used to make a probabilistic forecast. The model involves two sub-models, a water balance model and acrop development model. The two sub-models interact to provide 5-day estimates of soil water content, actual evaporation and transpiration, runoff and increments to biomass and grain yield. The water balance model takes inputs of daily rainfall and estimated potential evapotranspiration. Available energy is partitioned between evaporation and transpiration depending on leaf area index. There are two soil layers plus a surface interception and depression store. Water removal from the soil layers is dependent on root development and the location of available water. Biomass production is driven by actual transpiration and transpiration efficiency and so biomass and grain development are dependent on the timing and amount of water and energy utilization by the crop. The model was first calibrated in northern New South Wales with 13 years of research station data. With minor recalibration, it provided good estimates of observed district wheat harvests for a continuous period of 75 years. Further recalibration with 30 years of shire data from Queensland, 29 years of single farm data in southern New South Wales and with 31 years of county data from northwestern USA., indicated the model is able to accurately reproduce observed yields and has the potential to provide reliable forecasts of yield, in a range of different climates.