Improved nitrogen retrievals with airborne-derived fluorescence and plant traits quantified from VNIR-SWIR hyperspectral imagery in the context of precision agriculture

• DOI: 10.1016/j.jag.2018.04.013
• Publication Date: 2018-05-03T23:27:54Z
• Authors: Carlos Camino; Victoria González...; Pilar Hernández; J.C. Sillero; Pablo J. Zarco‐Te...
• Keywords: 2. Zero hunger; Hyperspectral; Nitrogen concentration; Airborne; 0401 agriculture, forestry, and fisheries; NIR indices; 04 agricultural and veterinary sciences; 15. Life on land; Chlorophyll fluorescence; Chlorophyll content

Peer reviewed<br/>The authors gratefully acknowledge the financial support of the Spanish Ministry of Science and Education (MEC) for projects AGL2012-40053-C03-01, and AGL2012-35196 and the Junta de Andalucia for projects P12-AGR-2521 and P12-AGR-0482.<br/>In semi-arid conditions, nitrogen (N) is the main limiting factor of crop yield after water, and its accurate quantification remains essential. Recent studies have demonstrated that solar-induced chlorophyll fluorescence (SIF) quantified from hyperspectral imagery is a reliable indicator of photosynthetic activity in the context of precision agriculture and for early stress detection purposes. The role of fluorescence might be critical to our understanding of N levels due to its link with photosynthesis and the maximum rate of carboxylation (Vcmax) under stress. The research presented here aimed to assess the contribution played by airborne-retrieved solar-induced chlorophyll fluorescence (SIF) to the retrieval of N under irrigated and rainfed Mediterranean conditions. The study was carried out at three field sites used for wheat phenotyping purposes in Southern Spain during the 2015 and 2016 growing seasons. Airborne campaigns acquired imagery with two hyperspectral cameras covering the 400–850 nm (20 cm resolution) and 950–1750 nm (50 cm resolution) spectral regions. The performance of multiple regression models built for N quantification with and without including the airborne-retrieved SIF was compared with the performance of models built with plant traits estimated by model inversion, and also with standard approaches based on single spectral indices. Results showed that the accuracy of the models for N retrieval increased when chlorophyll fluorescence was included (r2LOOCV ≥ 0.92; p < 0.0005) as compared to models only built with chlorophyll a + b (Cab), dry matter (Cm) and equivalent water thickness (Cw) plant traits (r2LOOCV ranged from 0.68 to 0.77; p < 0.005). Moreover, nitrogen indices (NIs) centered at 1510 nm yielded more reliable agreements with N concentration (r2 = 0.69) than traditional chlorophyll indices (TCARI/OSAVI r2 = 0.45) and structural indices (NDVI r2 = 0.57) calculated in the VNIR region. This work demonstrates that under irrigated and non-irrigated conditions, indicators directly linked with photosynthesis such as chlorophyll fluorescence improves predictions of N concentration.<br/>