Computer simulation models are becoming an ever more popular means for management of an agricultural production system. This study was carried out to determine the accuracy of the cotton growth simulation model GOSSYM COMAX when utilized for cotton mangement under southwestern desert conditions, specifically Arizona. This field validation was conducted with the idea of comparing the effects of planting date and irrigation termination dates on yield and other crop growth parameters (i.e. plant height, number of mainstem nodes, Height(inches):Node ratios, Heat Unit to Node ratios). Data was taken from two studies conducted at the Maricopa Ag Center in Maricopa, AZ over a period of three years: Planting date by irrigation termination (1990, 1991, 1992, 2 dates of planting, 2 irrigation terminations) and variety by irrigation termination (1991, 1992, 3 varieties and 3 irrigation terminations). These experiments were used as the source of actual data for comparison. Plant growth parameters were expressed as functions of Heat Units (HU 86/55°F Thresholds). Planting dates were based upon HU accumulated since Jan1. Planting date 1 was generally planted anywhere from 250 HU/Jan1 to 500 HU/Jan1. Planting date 2 was planted from 550 HU/Jan1 to 900 HU/Jan1. Irrigation terminations were based upon stage of crop growth. The first irrigation termination was after cut-out. The goal was to mature all fruit that was set during the first fruiting cycle. Irrigation termination treatments two and three were one and two additional irrigations respectively. These were spaced approximately two weeks apart. GOSSYM runs were made for each treatment combination using cultural inputs (i.e. irrigation, fertilization), soil characteristic data and weather data corresponding to that specific treatment. GOSSYM data was then compared to actual data collected from these experiements. Results showed that GOSSYM is very capable of tracking the effects of PD (planting dates) on yield where PD2 was consistently lower than PD1 for every simulation. The actual yields also demonstrated this. Irrigation termination (IT) treatment effects were not well simulated by GOSSYM. Yields were identical for all three IT treatments. Yield predictions overall varied greatly by season. In 1990, 1991 GOSSYM under predicted yield while in 1992 GOSSYM over predicted. Plant height was tracked very closely by GOSSYM overall and by treatment. However, large discrepancies were found between the simulated and actual data in number of mainstem nodes. GOSSYM consistently simulated fewer node numbers and appeared to hit a plateau towards the end of the season where node production completely stopped. The termination of node production was, in every case, highly correlated to a simulated nitrogen stress. The HU:node ratio comparison showed that towards the end of the season it would take approximately 500 HU for GOSSYM to produce a new node while the actual data never never exceeded 120 HU/node. With the discrepancy in node number the height to node ratios (HNR's) calculated from GOSSYM data were much higher than the actual HNR's. The fewer node numbers might account for the under prediction of yield because there would be fewer fruiting sites. However, the over prediction of yield also occured with fewer mainstem nodes. This would point to an error in simulated fruit retention (higher than normal). The simulated nitrogen stress which was highly correlated to the termination of node production did not actually occur in the field. The large amounts of irrigation water applied (¼6 acre in/irrigation) might have caused GOSSYM to simulate undue leaching of nitrates through the soil profile.