|Jones, D - UNIV OF NEBRASKA|
|Pinter Jr, Paul|
Submitted to: Biological Systems Simulation Group Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: January 31, 2002
Publication Date: N/A
Technical Abstract: Remotely sensed (RS) estimates of state variables such as leaf area index, above-ground-biomass, and plant water status have all been used as inputs to or for calibration of crop simulation models. In most cases, the RS estimate is considered the "correct" estimate of the state variable and the model in error. There can be many situations where this assumption cab be violated (e.g., errors in atmospheric correction of the RS data, or 0inaccurate empirical relationships between the state variable and RS data). Therefore, it is advisable to have an error limit for the RS estimate and some strategy for determining if the model's prediction is truly in error. The first step is to assign error limits to the RS estimate through standard statistical approaches or using uncertainty estimates derived from Monte Carlo Simulation or fuzzy number theory. Once error limits have been assigned to the RS estimate, one must decide if the model's prediction is better or worse than the RS estimate. One approach could be to adjust the model to the RS data only if it is outside of the error limits of the RS estimate. However, this approach loses any information contained in the model's prediction at that time. The uncertainty for any replacement should be considered when the next RS observation becomes available. If the next RS estimate still disagrees with the model, but the previous RS estimate had higher uncertainty than the present one, the simulation could be repeated without the previous adjustment and see if the model then agrees with the current RS estimate. If so, there is reason to question if the model really was wrong that previous time.