Modelling crop productivity, water fluxes and yield in winter wheat from remotely-sensed drone data under differential sulphur, nitrogen and or sugar application.

Understanding how crops contribute to carbon, water and nitrogen cycling under different fertiliser regimes will be crucial for improving ecosystem models and predicting future yields. Synthetic fertilisers hugely boost crop yields, but excessive application often leads to negative environmental impacts including increased nitrous oxide emissions (about c. 300x more potent than CO₂). To maximise crop yields and optimise fertiliser and water application, rapid retrieval of plant traits and fluxes will be critical. Here, we explore the effectiveness of optical (trait-based) and thermal (flux-based) remotely-sensed data collected from ground-based and drone platforms for quantifying differences in plant physiological performance and overall yield in field-grown wheat under different nitrogen, sulphur and or sugar treatments.

Research was undertaken at a winter wheat variable nutrient field trial in North Yorkshire, UK during June, 2021. Across 24 treatment plots (3 plot replicates per treatment), leaf level hyperspectral reflectance data was obtained using a Spectral Evolution PSR+ 3500 Spectroradiometer which was paired with stomatal conductance (g_sw) measurements (collected using a LI-COR LI-600 porometer) and photosynthetic capacity (V_cmax) measurements (collected using a Li-6800 portable infra-red gas analyser). Plant thermal images were captured using a handheld FLIR T650-C thermal imaging camera (640x480). Field-assessed leaves were destructively harvested for leaf chlorophyll and nitrogen content analysis. Drone flights were conducted using a DJI Matrice M200 with a MicaSense RedEdge-Mx multispectral imaging sensor (1456 x 1088) and a Parrot Analfi thermal drone (160 x 120) at a 10 m altitude above ground.

Results show that plants fertilised with sulphur and nitrogen had the highest or equal-highest leaf chlorophyll values (c. 60-70 µg/cm²), followed by plants that only received nitrogen (c. 40-55 µg/cm²), with unfertilised controls having the lowest chlorophyll values (c. 15-20 µg/cm²). Sugar did not significantly affect leaf chlorophyll values but an interaction was detectable between sugar and fertiliser at the plot level (Two-way ANOVA, p < 0.05). Strong relationships were found between the MERIS terrestrial chlorophyll index (MTCI) spectral vegetation index, calculated from drone-acquired optical reflectance, and both leaf chlorophyll content (R² = 0.76; p < 0.0001) and crop yield (R²= 0.90; p < 0.0001). V_cmax assessment also revealed a strong relationship with chlorophyll across fertiliser treatments (R² = 0.77, p < 0.0001), with sulphur and nitrogen application again producing the highest trait values. Plants receiving nitrogen or nitrogen and sulphur had c. 50% higher g_sw, with leaf temperatures that were c. 1-3 °C cooler than unfertilised controls. Sugar did not significantly affect leaf g_sw or temperature. Using ground-based and drone-mounted thermal cameras, strong correlations were shown between leaf temperature and g_sw (R² = 0.64; p < 0.0001  and R² = 0.6; p < 0.001 respectively).  Data captured during the drone flights enabled the production of spatial maps of V_cmax (~3 cm spatial resolution) across the field trails to reveal clear differences in photosynthetic capacity both across and within nutrient treatments. Overall, remote sensing data accurately captured subtle differences in plant traits and water fluxes paving the way for field-scale mapping of crop physiological, biochemical and structural traits.