High-throughput phenotyping of a large tomato collection under water deficit: Combining UAVs’ remote sensing with conventional leaf-level physiologic and agronomic measurements

Abstract

Field high-throughput phenotyping (HTPP) studies are highly needed to study water use efficiency (WUE), stress tolerance capacities, yield and quality in tomato to improve crop breeding strategies and adapt them to the climatic change scenario. In this study, UAV remote sensing is tested by comparison with leaf-level physiologic and agronomic measurements in a collection including 91 tomato genotypes. These genotypes include long shelf-life (LSL) and non-LSL (CON) Mediterranean landraces, cultivated under well-watered (WW, covering 100% crop evapotranspiration demands) and water deficit (WD, irrigation stopped one month after plantlet transplantation to field) conditions. Aerial remote sensing (including multispectral imaging), leaf gas-exchange, leaf carbon isotope composition (δ13C), fruit production and quality measurements, including total soluble solids and acidity, were performed. Differences between CON and LSL genotypes were observed in leaf-level physiologic and remote sensing measurements under both WW and WD conditions, while for agronomic measurements differences were only found for quality traits under WW conditions. Significant relationships were detected between remote sensing and leaf-level physiologic and agronomic measurements when considering all genotypes and treatments. However, different regressions were described for CON and LSL genotypes, mainly due their different physiologic behavior and response to WD. For instance, for the same NDVI value LSL genotypes showed near 30% lower AN and half gs than CON, and therefore higher intrinsic water use efficiency (WUEi). Also, tomato fruit quality was approached through remote sensing measurements, being correlated with multispectral indices. In conclusion, this study shows how remote sensing can help to optimize tomato physiologic and agronomic phenotyping processes. However, it also points out that the inclusion of genotypes with a different water use efficiency behavior and response to WD lead to a large scattering in the relationships between remote sensing and physiologic and agronomic traits and prevents to obtention of reliable models.