Evaluation of a Simple Model to Estimate Wheat Yield Through Data Assimilation of Remote Sensing Images.

Early crop yield estimation is extremely relevant to assist on management decisions and to plan harvest logistics and commercialization. The objective of this work was to evaluate the performance of a simple model of crop growth that assimilates remote sensing leaf area index (LAI) data to estimate wheat yield at the field level. The crop growth model used was a derivative of SAFY. The model was designed with the constraints of the data assimilation problem in mind, it is extremely simple, lacks soil components and only describes above ground growth. It is a model based on light use efficiency for dry mater production, allometry for partitioning, and thermal time for phenology. Inputs are divided in site-cultivar specific (date of emergence, date of anthesis) and species constants (rate of senescence). Through the assimilation of LAI data using the levenburg-marquardt algorithm, two parameters were adjusted (initial above ground biomass per unit area, and light use efficiency). Model calibration was performed with data from field trials of 2012, where time series of LAI, evolution of dry matter and nitrogen, and grain yield were collected. The model was validated in a 114 ha field in 2012 where harvester yield map and six high resolution multispectral images acquired from UAV were available. LAI was estimated from CI_green vegetation index and latter fed into the model to estimate growth and yield. The use of a simple model with few parameters to adjust proved useful in achieving sufficiently reliable and robust estimates of final grain yield.