UNCERTAINTIES IN CROP, SOIL AND WEATHER INPUTS USED IN GROWTH-MODELS - IMPLICATIONS FOR SIMULATED OUTPUTS AND THEIR APPLICATIONS

Deterministic crop growth models require several inputs relating to cr op/variety, soil physical properties, weather and crop management. The input values used could be significantly uncertain due to random and systematic measurement errors and spatial and temporal variation obser ved in many of these inputs. Often soil and weather data are approxima ted using GIS and/or weather generators. In this paper total uncertain ty in simulated yield, evapotranspiration and crop N uptake has been q uantified considering uncertainties in crop, soil and weather inputs. WTGROWS, a crop model that simulates the effect of genotypic, climatic , edaphic and management factors on productivity of spring wheat was u sed. The uncertainty in each input was represented by a statistical di stribution of values based on literature review, actual measurement an d subjective expert judgement. The Monte Carlo simulation technique wa s used to analyze total uncertainty. The results showed that uncertain ties in crop, soil and weather inputs resulted in uncertainty in simul ated grain yield, ET and N uptake, which varied depending upon the pro duction environment. Uncertainties in outputs increased as the product ion system changed from a potential production level to a level where crop growth was constrained by limited availability of water rand nitr ogen. There was an 80% probability that the bias in the deterministic model outputs was always less than 10% in potential and irrigated prod uction systems. In rainfed environments this bias was larger. The bias in simulated outputs was less than or equal to model error. Most of t he uncertainty in outputs caused by variable soil, crop and weather in puts could be represented if the outputs were determined using fixed s oil and crop data, and a large series of weather data. In potential an d irrigated production systems, inputs relating to crop photosynthesis and leaf area estimation had a large 'uncertainty importance'. Uncert ainties in soil N inputs and vapor pressure were also of great importa nce in irrigated environments. In rainfed environments, uncertainties in soil and weather inputs were dominant and crop parameters had only limited 'uncertainty importance'. The implications of these results in estimates of potential and rainfed productivity, database development and guiding refinement of models are discussed.