Timely crop water stress detection can help precision irrigation management and minimize yield loss. A two-year study was conducted on non-invasive winter wheat water stress monitoring using state-of-the-art computer vision and thermal-RGB imagery inputs. Field treatment plots were irrigated using two irrigation systems (flood and sprinkler) at four rates (100, 75, 50, and 25% of crop evapotranspiration [ET_c]). A total of 3200 images under different treatments were captured at critical growth stages, that is, 20, 35, 70, 95, and 108 days after sowing using a custom-developed thermal-RGB imaging system. Crop and soil response measurements of canopy temperature (T_c), relative water content (RWC), soil moisture content (SMC), and relative humidity (RH) were significantly affected by the irrigation treatments showing the lowest T_c (22.5 ± 2 °C), and highest RWC (90%) and SMC (25.7 ± 2.2%) for 100% ET_c, and highest T_c (28 ± 3 °C), and lowest RWC (74%) and SMC (20.5 ± 3.1%) for 25% ET_c. The RGB and thermal imagery were then used as inputs to feature-extraction-based deep learning models (AlexNet, GoogLeNet, Inception V3, MobileNet V2, ResNet50) while, RWC, SMC, T_c, and RH were the inputs to function-approximation models (Artificial Neural Network (ANN), Kernel Nearest Neighbor (KNN), Logistic Regression (LR), Support Vector Machine (SVM) and Long Short-Term Memory (DL-LSTM)) to classify stressed/non-stressed crops. Among the feature extraction-based models, ResNet50 outperformed other models showing a discriminant accuracy of 96.9% with RGB and 98.4% with thermal imagery inputs. Overall, classification accuracy was higher for thermal imagery compared to RGB imagery inputs. The DL-LSTM had the highest discriminant accuracy of 96.7% and less error among the function approximation-based models for classifying stress/non-stress. The study suggests that computer vision coupled with thermal-RGB imagery can be instrumental in high-throughput mitigation and management of crop water stress.

Geospatial crop water use mapping is critical for field-scale site-specific irrigation management. Landsat 7/8 satellite imagery with a widely adopted METRIC (Mapping Evapotranspiration at high Resolution with Internalized Calibration) energy balance model (LM approach) estimates accurate evapotranspiration (ET) but limits field-scale spatiotemporal (30 m pixel⁻¹, ~16 days) mapping. A study was therefore conducted to map actual ET of commercially grown irrigated-field crops (spearmint, potato, and alfalfa) at very high-resolution (7 cm pixel⁻¹). Six small unmanned aerial system (UAS)-based multispectral and thermal infrared imagery campaigns were conducted (two for each crop) at the same time as the Landsat 7/8 overpass. Three variants of METRIC model were used to process the UAS imagery; UAS-METRIC-1, -2, and -3 (UASM-1, -2, and -3) and outputs were compared with the standard LM approach. ET root mean square differences (RMSD) between LM-UASM-1, LM-UASM-2, and LM-UASM-3 were in the ranges of 0.2–2.9, 0.5–0.9, and 0.5–2.7 mm day⁻¹, respectively. Internal calibrations and sensible heat fluxes majorly resulted in such differences. UASM-2 had the highest similarity with the LM approach (RMSD: 0.5–0.9, ET_dep,abs (daily ET departures): 2–14%, r (Pearson correlation coefficient) = 0.91). Strong ET correlations between UASM and LM approaches (0.7–0.8, 0.7–0.8, and 0.8–0.9 for spearmint, potato, and alfalfa crops) suggest equal suitability of UASM approaches as LM to map ET for a range of similar crops. UASM approaches (Coefficient of variation, CV: 6.7–24.3%) however outperformed the LM approach (CV: 2.1–11.2%) in mapping spatial ET variations due to large number of pixels. On-demand UAS imagery may thus help in deriving high resolution site-specific ET maps, for growers to aid in timely crop water management.

Site-specific irrigation management for perennial crops such as grape requires water use assessments at high spatiotemporal resolution. In this study, small unmanned-aerial-system (UAS)-based imaging was used with a modified mapping evapotranspiration at high resolution with internalized calibration (METRIC) energy balance model to map water use (UASM-ET approach) of a commercial, surface, and direct-root-zone (DRZ) drip-irrigated vineyard. Four irrigation treatments, 100%, 80%, 60%, and 40%, of commercial rate (CR) were also applied, with the CR estimated using soil moisture data and a non-stressed average crop coefficient of 0.5. Fourteen campaigns were conducted in the 2018 and 2019 seasons to collect multispectral (ground sampling distance (GSD): 7 cm/pixel) and thermal imaging (GSD: 13 cm/pixel) data. Six of those campaigns were near Landsat 7/8 satellite overpass of the field site. Weather inputs were obtained from a nearby WSU-AgWeatherNet station (1 km). First, UASM-ET estimates were compared to those derived from soil water balance (SWB) and conventional Landsat-METRIC (LM) approaches. Overall, UASM-ET (2.70 ± 1.03 mm day⁻¹ [mean ± std. dev.]) was higher than SWB-ET (1.80 ± 0.98 mm day⁻¹). However, both estimates had a significant linear correlation (r = 0.64–0.81, p < 0.01). For the days of satellite overpass, UASM-ET was statistically similar to LM-ET, with mean absolute normalized ET departures (ET_d,MAN) of 4.30% and a mean r of 0.83 (p < 0.01). The study also extracted spatial canopy transpiration (UASM-T) maps by segmenting the soil background from the UASM-ET, which had strong correlation with the estimates derived by the standard basal crop coefficient approach (T_d,MAN = 14%, r = 0.95, p < 0.01). The UASM-T maps were then used to quantify water use differences in the DRZ-irrigated grapevines. Canopy transpiration (T) was statistically significant among the irrigation treatments and was highest for grapevines irrigated at 100% or 80% of the CR, followed by 60% and 40% of the CR (p < 0.01). Reference T fraction (T_rF) curves established from the UASM-T maps showed a notable effect of irrigation treatment rates. The total water use of grapevines estimated using interpolated T_rF curves was highest for treatments of 100% (425 and 320 mm for the 2018 and 2019 seasons, respectively), followed by 80% (420 and 317 mm), 60% (391 and 318 mm), and 40% (370 and 304 mm) of the CR. Such estimates were within 5% to 11% of the SWB-based water use calculations. The UASM-T-estimated water use was not the same as the actual amount of water applied in the two seasons, probably because DRZ-irrigated vines might have developed deeper or lateral roots to fulfill water requirements outside the irrigated soil volume. Overall, results highlight the usefulness of high-resolution imagery toward site-specific water use management of grapevines.