Multi-sensors data fusion approach for site-specific seeding of consumption and seed potato production

This study evaluated the agronomic and economic prospects of Site-Specific Seeding (SSS) for consumption and seed potato production based on Management Zone (MZ) maps delineated with the fusion of multiple soil and crop attributes at four experimental sites in Belgium. Soil pH, organic carbon, P, K, Mg, Ca, Na, moisture content, cation exchange capacity, apparent electrical conductivity and crop normalized difference vegetation index were measured with an on-line visible and near-infrared reflectance spectroscopy sensor, electromagnetic induction sensor, and Sentinel-2 constellation, respectively. Spatial alignment of the different data layers generated a co-georeferenced data matrix for data fusion by k-means clustering. Per field MZ classes were ranked according to their fertility status and the prescription rule of sowing more seeds to the more fertile zones and vice versa was adopted and compared against a Uniform Rate Seeding (URS) treatment in a strip plot experiment. Cost–benefit analysis revealed that the SSS improved tuber yields, hence, increased gross margin (137.81 to 457.83 €/ha) of production compared to the URS, although SSS consumed relatively higher amount of seeds. The percentage of gross margin increase varied between 2.34 and 27.21%, with the highest profitability in fields with low productivity. Larger seed-to-seed spacing than the control increased the proportion of the most demanded and profitable tuber category, suggesting the seeding interval is a key determinant of tuber size distribution. It is suggested to adopt SSS for potato production using the proposed multi-sensor data-fusion approach to manage in-field soil and crop variabilities, and improve productivity and profitability.