Yield gap estimation by combining remote sensing and crop growth model

Conventional analytical crop growth models cannot handle actual land use systems because of massive data needs, algorithm complexity and prohibitive error propagation. So far, actual production could only be determined through field measurements. A methodology was presented for estimating regional levels of actual crop production. The difference between remotely sensed canopy temperature and ambient air temperature was used to estimate the degree of stomata closure of the crop. Introducing this remote sensing based degree of stomata closure in calculations of assimilatory activity permits to calculate the actual rate of crop growth over regions. An integrated model (the PS-X model) introduces NOAA-14 AVHRR remote sensing data into crop growth modeling. The PS-X model simulates crop growth and production at several levels of abstraction: it calculates the bio-physical production potential (PS-1), the water-limited production potential (PS-2) and the actual production (PS-X). It was run using data on summer maize collected in Handan, North China Plain (NCP) in 1998. Regionally differentiated yield gaps were analyzed by combining production calculations at PS-1, PS-2 and PS-X levels with farmer's household survey data. The yield gap between the PS-1 and PS-2 levels was largely explained by differences in soil type and rainfall. However the larger part of the yield gap between the bio-physical production potential and actual production (81.4%) exists between the PS-2 and PS-X levels and is largely caused by management factors. The estimation accuracy of summer maize production exceeded 90% for counties in the plain. Canopy temperature and crop water stress factor estimates were higher on sandy soil than those on loam or clay soils. Hence the modeled yield was lower for plots on sandy soil than that on other soils. This result was consistent with results obtained with analyses at the PS-2 level and with farmers interviews. This study confirms that it is feasible to estimate regional crop production by coupling instantaneous remotely sensed temperature information and crop growth simulation.