Abstract:
Information gatheringfor early warning and crop assessment in Tanzania is based on physical inspection
of standing crop in sample jields. This process is subject to human error, inadequate and is also time consuming.
Recent developments in computer simulation have paved the way for more efficient methods of
analysing datafor purposes of early warning and crop assessment. Two such sch~mes based on soil water
balance simulation, viz. IRSIS and CRPSM models were used in this study fo see how closeZv they could
predict grain yieldsfor selected stations in Tanzania. Inputfor the models comprised of weather, crop and
soil data collected from jive selected stations. Simulation results show that IRSIS model tends to
over predict grain yields of maize, sorghum and wheat, a fact that could be attributed to the inadequacy of
the model to accurately account for rainfall excess. On the other hand, the CRPSA1 model simulated results
were not significantZv different (P>O. 05) from the actual grain yields ojmaize, sorghum,. wheat and
beans. Although the agreement between actual and simulated yield data was good, it was observed that
mean valuesfor predicted grain yields were consistently lower thanfor actual grain yields. This could be
attributed to the use of approximate rather than location specific input parameters required by the
CRPSM model. Locally calibrated input parameters in the CRPSM model could filrther improve the accuracy
of the model and hence its ability to predict grain yields.
