Development of a cotton crop model that uses remote sensing data

A new cotton crop model able to use remote sensing data as input was developed and tested using field data sets based on GRAMI. Based on analysis of a 2002 field data set obtained at Halfway, TX, the primary parameters that affect the cotton simulation model were estimated. These included radiation use efficiency (RUE: 2.3 g MJ-1), light extinction coefficient (K: 0.9), specific leaf area (SLA: 0.01 m^2 g^-1), and a coefficient of boll production ( ã : 0.57 GDD^-1). The model also included boll growth and lint yield estimation functions. The model was first verified using cotton field data obtained at Halfway, TX, in 2002, and then validated using data sets obtained in Lamesa County, TX, in 1999 and 2001. Model verification was performed using both leaf area index (LAI) and remotely sensed ground cover (GC) data. Simulated results using LAI as input are as follows: simulated values matched with measured values with R^2 0.91-0.97 and RMSE 0.16-0.35 for LAI; with R^2 0.92-0.96 and RMSE 95.7-142.7 for AGDM; with R' 0.83-0.9 and RMSE 13.9-36.0 for boll number; with R^ 0.37 and RMSE 15.9 for harvestable boll number; and with R^2 0.5 and RMSE 153.3 for lint yield in three fields. Simulated results using remotely sensed GC as input were as follows: simulated values agreed with measured values with R^2 0.95-0.98 and RMSE 0.06-0.07 for GC; with R^ 0.91-0.95 and RMSE 94.7-131.6 for AGDM; with R^ 0.82-0.91 and RMSE 17.4-21.9 for boll number; with R^2 0.3 and RMSE 14.5 for simulated harvestable bolls; and with R^ 0.3 and RMSE 121.8 for lint yield. The validated results were as follows. Simulated values matched with measured values with R^ 0.94 and RMSE 0.12 for LAI, and with R^ 0.97 and RMSE 48.5 for AGDM in 1999 data; with R' 0.93 and RMSE 0.39 for LAI, and with R^ 0.98 and RMSE 131.1 for AGDM in 2001 data. In both years, simulated boll number, harvestable boll number, and lint yield matched with measurements with R"^ 0.88 and RMSE 19.3 for boll number; with R^2 0.73 and RMSE 12.6 for harvestable boll number; and with R^ 0.75 and RMSE 181.8 for lint yield. In general, simulated values obtained with the new cotton model showed reasonable agreements with their corresponding measurements. The new model not only has simple input requirements but is also easy to use, and applicable to regional cotton growth monitoring and lint yield mapping projects.