Can big data explain yield variability and water productivity in intensive cropping systems?

Silva, João Vasco; Tenreiro, Tomás R.; Spätjens, Léon; Anten, Niels P.R.; Ittersum, Martin K. van; Reidsma, Pytrik


Yield gaps and water productivity are key indicators to monitor the progress towards more sustainable and productive cropping systems. Individual farmers are collecting increasing amounts of data (‘big data’), which can help monitor the process of sustainable intensification at local level. In this study, we build upon such data to quantify the magnitude and identify the biophysical and management determinants of on-farm yield gaps and water productivity for the main arable crops cultivated in the Netherlands. The analysis focused on ware, seed and starch potatoes, sugar beet, spring onion, winter wheat and spring barley and covered the period 2015–2017. A crop modelling approach based on crop coefficients (kc) and daily weather data was used to estimate the potential yield (Yp), radiation intercepted and potential evapotranspiration (ETP) for each crop. Yield gaps were estimated to be ca. 10% of Yp for sugar beet, 25–30% of Yp for ware, seed and starch potato and spring barley, and 35–40% of Yp for spring onion and winter wheat. Variation in actual yields was associated with water availability in key periods of the growing season as well as with sowing and harvest dates. However, the R2 of the fitted regressions was rather low (20–49%). Current levels of crop water productivity ranged between 13 kg DM ha−1 mm−1 for spring barley, ca. 15 kg DM ha−1 mm−1 for seed potato, spring onion and winter wheat, 23 kg DM ha−1 mm−1 for ware potato and ca. 25 kg DM ha−1 mm−1 for starch potato and sugar beet. These values are about half of their potential, but increasing actual water productivity further is restricted by rainfall amount and distribution. However, doing so should not be prioritized over reducing environmental impacts of these intensive cropping systems in the short-term and may require large investments from farm to regional levels in the long-term. Although these findings are most relevant to similar cropping systems in NW Europe, the underlying methods are generic and can be used to benchmark crop performance in other cropping systems. Based on this work, we argue that ‘big data’ are currently most useful to describe cropping systems at regional scale and derive benchmarks of farm performance but not as much to predict and explain crop yield variability in time and space.