Yield Prediction

Crop yield is largely defined by interactions between genotype and environmental factors. Data on a broad range of environmental factors throughout the growing season for specific local environments, including that from satellite crop remote sensing and climate data, can guide computer deep learning systems for accurate yield estimations.

Canopy spectral reflectance data from satellites, used in conjunction with weather data is used to gather crop status data for computer driven yield prediction. Crop biophysical and biochemical status affects canopy photosynthesis which can be indirectly measured as the crop fractional absorbed photosynthetically active radiation (fAPAR). Crop biomass accumulation is proportional to the cumulative absorbed photosynthetically active radiation (APAR); therefore, fAPAR can be used to model crop yields with algorithms programmed from radiation use efficiency (RUE) models. Temporary absence of satellite data due to climatic conditions and low temporal resolution are overcome by optimisation algorithms, allowing researchers to predict wheat yield with ca. 75% accuracy 2 months before the end of the growing season.

Yield prediction helps make informed precision crop management choices and guide financial decisions. This also includes crop loss assessment and mitigation in difficult seasons. Some systems are now. Yield prediction during the season when decision making is still ongoing is of far more value than at the point of just preharvest.  There are two types of systems a) giving 50-60 days pre harvest with 90 % accuracy, and b) an continual indication throughout the growing season which is of more use.

https://www.frontiersin.org/articles/10.3389/fpls.2019.00621/full

https://www.researchgate.net/publication/331738561_Toward_Precision_in_Crop_Yield_Estimation_Using_Remote_Sensing_and_Optimization_Techniques

https://www.futurefarming.com/Tools-data/Articles/2019/5/Machine-learning-accurately-predicts-wheat-yield-427203E/