Simple regression models to estimate light interception in wheat crops with Sentinel‐2 and a handheld sensor

Abstract

Capture of radiation by crop canopies drives growth rate, grain set, and yield. Since the fraction of photosynthetically active radiation absorbed by green area (fAPARg) correlates with normalized difference vegetation index (NDVI), remote sensors have been used to monitor vegetation. With a 10-m spatial resolution and 5-d revisiting time, the recently launched Sentinel-2 satellite is a promising tool for fAPARg monitoring. However, the available algorithm to estimate fAPARg is based on simulations of canopy interception of several vegetation types and was never tested in field crops. Handheld sensors, such as GreenSeeker, are another alternative to estimate fAPARg. Our objectives were (a) to test the ability of indices derived from Sentinel-2 and GreenSeeker NDVI to capture fAPARg of wheat (Triticum aestivum L.) crops, (b) to compare these sensors’ performance against the moderate resolution imaging spectroradiometer (MODIS), and (c) to compare our Sentinel-2 model estimations with the available algorithm. In wheat fields in the southwest Argentinean Pampas, on several sampling dates, we measured fAPARg with a quantum light sensor and NDVI with a GreenSeeker. We regressed fAPARg measurements with vegetation indices from the different sources and selected the best models. Sentinel-2 and GreenSeeker NDVI precisely estimated fAPARg, with a performance similar to MODIS (p <.05; RMSD = 0.09, 0.11, and 0.08; R2 =.89,.88, and.95, respectively). The available algorithm to estimate fAPARg with Sentinel-2 yielded biased estimations, mainly in the lower range of fAPARg. These results suggest that simple models may provide fAPARg estimations with Sentinel-2 and GreenSeeker in wheat crops with an accuracy suitable for agricultural applications.