Overview of Proximal Sensing and Field Phenomics.

Meeting the projected demands for agricultural production requires a transformation in the accuracy and speed with which crop growth and development are monitored in field research. Proximal sensing based on reflectance spectroscopy, imaging, thermometry and other approaches can provide required throughput under field conditions, and georeferencing via GPS allows such measurements to be located at sub-plot scales. While great scope remains for improving instrumentation, the critical challenges in field phenomics currently are in integrating the existing instruments onto robust platforms, whether tractors, cranes, aircraft or other vehicles, and in efficiently estimating phenotypic values from multiple data streams. Ecophysiological models offer unique potential for synthesizing raw sensor data due to their ability to quantify the complex effects of weather, soil conditions and crop management. Through inverse modeling, biologically fundamental phenotypes can be estimated as genotype-specific model parameters. Treating the raw or partially processed data from phenotyping as observed data, parameter values for traits such as photoperiod sensitivity or potential leaf expansion rate can be estimated through model optimization procedures. An as-yet untested proposition is that these estimated phenotypes will show higher heritability than phenotypes derived directly from high-throughput or conventional, manual phenotyping. While widely used ecophysiological models simulate crop cycles from emergence to harvest, proximal sensing at sub-daily time scales suggests a potential demand for simulating detailed physiology and energy balances for single days at sub-hourly time scales. Ultimately, as genetic control of traits is better understood, models should be revised to incorporate this knowledge, leading to cycles of improved phenotyping, greater understanding of genetic and physiological controls, and further model improvement.