Developing rapid and non-destructive methods for chlorophyll estimation over large spatial areas is a topic of much interest, as it would provide an indirect measure of plant photosynthetic response, be useful in monitoring soil nitrogen content, and offer the capacity to assess vegetation structural and functional dynamics. Traditional methods of direct tissue analysis or the use of handheld meters, are not able to capture chlorophyll variability at anything beyond point scales, so are not particularly useful for informing decisions on plant health and status at the field scale. Examining the spectral response of plants via remote sensing has shown much promise as a means to capture variations in vegetation properties, while offering a non-destructive and scalable approach to monitoring. However, determining the optimum combination of spectra or spectral indices to inform plant response remains an active area of investigation. Here, we explore the use of a machine learning approach to enhance the estimation of leaf chlorophyll (Chl_t), defined as the sum of chlorophyll a and b, from spectral reflectance data. Using an ASD FieldSpec 4 Hi-Res spectroradiometer, 2700 individual leaf hyperspectral reflectance measurements were acquired from wheat plants grown across a gradient of soil salinity and nutrient levels in a greenhouse experiment. The extractable Chl_t was determined from laboratory analysis of 270 collocated samples, each composed of three leaf discs. A random forest regression algorithm was trained against these data, with input predictors based upon (1) reflectance values from 2102 bands across the 400–2500 nm spectral range; and (2) 45 established vegetation indices. As a benchmark, a standard univariate regression analysis was performed to model the relationship between measured Chl_t and the selected vegetation indices. Results show that the root mean square error (RMSE) was significantly reduced when using the machine learning approach compared to standard linear regression. When exploiting the entire spectral range of individual bands as input variables, the random forest estimated Chl_t with an RMSE of 5.49 µg·cm⁻² and an R² of 0.89. Model accuracy was improved when using vegetation indices as input variables, producing an RMSE ranging from 3.62 to 3.91 µg·cm⁻², depending on the particular combination of indices selected. In further analysis, input predictors were ranked according to their importance level, and a step-wise reduction in the number of input features (from 45 down to 7) was performed. Implementing this resulted in no significant effect on the RMSE, and showed that much the same prediction accuracy could be obtained by a smaller subset of indices. Importantly, the random forest regression approach identified many important variables that were not good predictors according to their linear regression statistics. Overall, the research illustrates the promise in using established vegetation indices as input variables in a machine learning approach for the enhanced estimation of Chl_t from hyperspectral data.

With salt stress presenting a major threat to global food production, attention has turned to the identification and breeding of crop cultivars with improved salt tolerance. For instance, some accessions of wild species with higher salt tolerance than commercial varieties are being investigated for their potential to expand food production into marginal areas or to use brackish waters for irrigation. However, assessment of individual plant responses to salt stress in field trials is time-consuming, limiting, for example, longitudinal assessment of large numbers of plants. Developments in Unmanned Aerial Vehicle (UAV) sensing technologies provide a means for extensive, repeated and consistent phenotyping and have significant advantages over standard approaches. In this study, 199 accessions of the wild tomato species, Solanum pimpinellifolium, were evaluated through a field assessment of 600 control and 600 salt-treated plants. UAV imagery was used to: (1) delineate tomato plants from a time-series of eight RGB and two multi-spectral datasets, using an automated object-based image analysis approach; (2) assess four traits, i.e., plant area, growth rates, condition and Plant Projective Cover (PPC) over the growing season; and (3) use the mapped traits to identify the best-performing accessions in terms of yield and salt tolerance. For the first five campaigns, >99% of all tomato plants were automatically detected. The omission rate increased to 2–5% for the last three campaigns because of the presence of dead and senescent plants. Salt-treated plants exhibited a significantly smaller plant area (average control and salt-treated plant areas of 0.55 and 0.29 m², respectively), maximum growth rate (daily maximum growth rate of control and salt-treated plant of 0.034 and 0.013 m², respectively) and PPC (5–16% difference) relative to control plants. Using mapped plant condition, area, growth rate and PPC, we show that it was possible to identify eight out of the top 10 highest yielding accessions and that only five accessions produced high yield under both treatments. Apart from showcasing multi-temporal UAV-based phenotyping capabilities for the assessment of plant performance, this research has implications for agronomic studies of plant salt tolerance and for optimizing agricultural production under saline conditions.

Monitoring leaf Chlorophyll (Chl) in-situ is labor-intensive, limiting representative sampling for detailed mapping of Chl variability at field scales across time. Unmanned aeria-l vehicles (UAV) and hyperspectral cameras provide flexible platforms for observing agricultural systems, overcoming this spatio-temporal sampling constraint. Here, we evaluate a customized machine learning (ML) workflow to retrieve multi-temporal leaf-Chl levels, combining sub-centimeter resolution UAV-hyperspectral imagery (400–1,000 nm) with leaf-level reflectance spectra and SPAD measurements, capturing temporal correlations, selecting relevant predictors, and retrieving accurate results under different conditions. The study is performed within a phenotyping experiment to monitor wild tomato plants’ development. Several analyses were conducted to evaluate multiple ML strategies, including: (1) exploring sequential versus retraining learning; (2) comparing insights gained from using 272 spectral bands versus 60 pigment-based vegetation indices (VIs); and (3) assessing six regression methods (linear, partial-least-square regression; PLSR, decision trees, support vector, ensemble trees, and Gaussian process; GPR). Goodness-of-fit (R²) and accuracy metrics (MAE, RMSE) were determined using training/testing and validation data subsets to assess the models’ performance. Overall, while equally good performance was obtained using either PLSR, GPR, or random forest, results show: (1) the retraining strategy improved the ability of most of the approaches to model SPAD-based Chl dynamics; (2) comparative analysis between retrievals and validation data distributions informed the models’ ability to capture Chl dynamics through SPAD levels; (3) VI predictors slightly improved R² (e.g., from 0.59 to 0.74 units for GPR) and accuracy (e.g., MAE and RMSE differences of up to 2 SPAD units) in specific algorithms; (4) feature importance examined through these methods, revealed strong overlaps between relevant bands and VI predictors, highlighting a few decisive spectral ranges and indices useful for retrieving leaf-Chl levels. The proposed ML framework allows the retrieval of high-quality spatially distributed and multi-temporal SPAD-based chlorophyll maps at an ultra-high pixel resolution (e.g., 7 mm).

Hyperspectral systems integrated on unmanned aerial vehicles (UAV) provide unique opportunities to conduct high-resolution multitemporal spectral analysis for diverse applications. However, additional time-consuming rectification efforts in postprocessing are routinely required, since geometric distortions can be introduced due to UAV movements during flight, even if navigation/motion sensors are used to track the position of each scan. Part of the challenge in obtaining high-quality imagery relates to the lack of a fast processing workflow that can retrieve geometrically accurate mosaics while optimizing the ground data collection efforts. To address this problem, we explored a computationally robust automated georectification and mosaicking methodology. It operates effectively in a parallel computing environment and evaluates results against a number of high-spatial-resolution datasets (mm to cm resolution) collected using a push-broom sensor and an associated RGB frame-based camera. The methodology estimates the luminance of the hyperspectral swaths and coregisters these against a luminance RGB-based orthophoto. The procedure includes an improved coregistration strategy by integrating the Speeded-Up Robust Features (SURF) algorithm, with the Maximum Likelihood Estimator Sample Consensus (MLESAC) approach. SURF identifies common features between each swath and the RGB-orthomosaic, while MLESAC fits the best geometric transformation model to the retrieved matches. Individual scanlines are then geometrically transformed and merged into a single spatially continuous mosaic reaching high positional accuracies only with a few number of ground control points (GCPs). The capacity of the workflow to achieve high spatial accuracy was demonstrated by examining statistical metrics such as RMSE, MAE, and the relative positional accuracy at 95% confidence level. Comparison against a user-generated georectification demonstrates that the automated approach speeds up the coregistration process by 85%.

The use of unmanned aerial vehicles (UAVs) for Earth and environmental sensing has increased significantly in recent years. This is particularly true for multi- and hyperspectral sensing, with a variety of both push-broom and snap-shot systems becoming available. However, information on their radiometric performance and stability over time is often lacking. The authors propose the use of a general protocol for sensor evaluation to characterize the data retrieval and radiometric performance of push-broom hyperspectral cameras, and illustrate the workflow with the Nano-Hyperspec (Headwall Photonics, Boston USA) sensor. The objectives of this analysis were to: (1) assess dark current and white reference consistency, both temporally and spatially; (2) evaluate spectral fidelity; and (3) determine the relationship between sensor-recorded radiance and spectroradiometer-derived reflectance. Both the laboratory-based dark current and white reference evaluations showed an insignificant increase over time (99.5% of pixel–waveband combinations. Using a mercury/argon (Hg/Ar) lamp, the hyperspectral wavelength bands exhibited a slight shift of 1-3 nm against 29 Hg/Ar wavelength emission lines. The relationship between the Nano-Hyperspec radiance values and spectroradiometer-derived reflectance was found to be highly linear for all spectral bands. The developed protocol for assessing UAV-based radiometric performance of hyperspectral push-broom sensors showed that the Nano-Hyperspec data were both time-stable and spectrally sound.

Developments in computer vision, such as structure from motion and multiview stereo reconstruction, have enabled a range of photogrammetric applications using unmanned aerial vehicles (UAV)-based imagery. However, some specific cases still present reconstruction challenges, including survey areas composed of steep, overhanging, or vertical rock formations. Here, the suitability and geometric accuracy of four UAV-based image acquisition and data processing scenarios for topographic surveying applications in complex terrain are assessed and compared. The specific cases include the use of: 1) nadir imagery; 2) nadir and oblique imagery; 3) nadir and façade imagery; and 4) nadir, oblique, and façade imagery to reconstruct a topographically complex natural surface. Results illustrate that including oblique and façade imagery to supplement the more traditional nadir collections significantly improves the geometric accuracy of point cloud data reconstruction by approximately 35% when assessed against terrestrial laser scanning data of near-vertical rock walls. Most points (99.41%) had distance errors of less than 50 cm between the point clouds derived from the nadir imagery and nadir–oblique–façade imagery. Apart from delivering enhanced spatial resolution in façade details, the geometric accuracy improvements achieved from integrating nadir, oblique, and façade imagery provide value for a range of applications, including geotechnical and geohazard investigations. Such gains are particularly relevant for studies assessing rock integrity and stability, and engineering design, planning, and construction, where information on the position of rock cracks, joints, faults, shears, and bedding planes may be required.