Search

Crop yield prediction typically involves the utilization of either theory-driven process-based crop growth models, which have proven to be difficult to calibrate for local conditions, or data-driven machine learning methods, which are known to require large datasets. In this work we investigate potato yield prediction using a hybrid meta-modeling approach. A crop growth model is employed to generate synthetic data for (pre)training a convolutional neural net, which is then fine-tuned with observational data. When applied in silico, our meta-modeling approach yields better predictions than a baseline comprising a purely data-driven approach. When tested on real-world data from field trials (n=303) and commercial fields (n=77), the meta-modeling approach yields competitive results with respect to the crop growth model. In the latter set, however, both models perform worse than a simple linear regression with a hand-picked feature set and dedicated preprocessing designed by domain experts. Our findings indicate the potential of meta-modeling for accurate crop yield prediction; however, further advancements and validation using extensive real-world datasets is recommended to solidify its practical effectiveness.

Crop yields are determined by the biophysical environment and by farm management decisions, which in turn depend on socio-economic conditions of the farm(er). The interaction of these factors results in spatial and temporal yield variability. We assessed ware potato yield variability in the Netherlands across four agronomically relevant scales (among provinces, farms and fields and within fields) using five datasets with data on potato yield across space and time. Furthermore, we disseminated an online questionnaire among farmers to identify the perceived yield gap and the key yield gap explaining factors at farm level. Spatial yield variability was largest among fields, with a standard deviation of 8.5–11.1 t ha−1, and within fields, with a standard deviation of 7.7–8.7 t ha−1. Spatial yield variability decreased at higher aggregation levels, i.e., the standard deviation of among-farm yield variability was 4.0–6.1 t ha−1 and that of among-provinces 1.6–3.5 t ha−1. Mean yields of the datasets ranged from 46 to 52 t ha−1. Temporal yield variability explained 10–55 % of the total observed variation in crop yield and its magnitude was equal or larger than the spatial yield variability for almost all datasets. Farmers estimated the ware potato yield gap at 13–18 t ha−1, corresponding to 20–24 % of estimated yield potential, depending on the soil type and variety. Water deficit and water excess were considered the most important yield gap explaining biophysical factors. In addition, soil structure was an important biophysical factor on clay soils and diseases on sandy soils. Irrigation and fertilization were identified as the most important yield gap explaining management factors, whereas legislation and potato prices were identified as the key socio-economic factors influencing potato yields. However, the perceived yield gap explaining factors varied with soil type, variety and year. We conclude that reducing potato yield variability in the Netherlands can be achieved b