TRAC is a new optical instrument for measuring the Leaf Area Index (LAI) and the Fraction of Photosynthetically Active Radiation absorbed by plant canopies (FPAR). TRAC measures canopy 'gap size' distribution in addition to canopy 'gap fraction'. Gap fraction is the percentage of gaps in the canopy at a given solar zenith angle. It is usually obtained from radiation transmittance. Gap size is the physical dimension of a gap in the canopy. For the same gap fraction, gap size distributions can be quite different.

A multiple sensor payload for a multi-rotor based UAV platform was developed and tested for measuring land surface albedo and spectral measurements at user-defined spatial, temporal, and spectral resolutions. The system includes a Matrice 600 UAV with an RGB camera and a set of four downward pointing radiation sensors including a pyranometer, quantum sensor, and VIS and NIR spectrometers, measuring surface reflected radiation. A companion ground unit consisting of a second set of identical sensors simultaneously measure downwelling radiation. The reflected and downwelling radiation measured by the four sensors are used for calculating albedo for the total shortwave broadband, visible band and any narrowband at a 1.5 nm spectral resolution within the range of 350–1100 nm. The UAV-derived albedo was compared with those derived from Landsat 8 and Sentinel-2 satellite observations. Results show the agreement between total shortwave albedo from UAV pyranometer and Landsat 8 (R² = 0.73) and Sentinel-2 (R² = 0.68). Further, total shortwave albedo was estimated from spectral measurements and compared with the satellite-derived albedo. This UAV-based sensor system promises to provide high-resolution multi-sensors data acquisition. It also provides maximal flexibility for data collection at low cost with minimal atmosphere influence, minimal site disturbance, flexibility in measurement planning, and ease of access to study sites (e.g., wetlands) in contrast with traditional data collection methods.

Digital hemispherical photography (DHP) is widely used to derive forest biophysical variables including leaf, plant, and green area index (LAI, PAI and GAI), the fraction of intercepted photosynthetically active radiation (FIPAR), and the fraction of vegetation cover (FCOVER). However, the majority of software packages for processing DHP data are based on a graphical user interface, making programmatic analysis difficult. Meanwhile, few natively support analysis of RAW image formats, while none incorporate the propagation or provision of uncertainties. To address these limitations, we present HemiPy, an open‐source Python module for deriving forest biophysical variables and uncertainties from DHP images in an automated manner. We assess HemiPy using simulated hemispherical images, in addition to multiannual time‐series and litterfall data from several forested National Ecological Observatory Network (NEON) sites, as well as comparison against the CAN‐EYE software package. Multiannual time‐series of PAI, FIPAR and FCOVER demonstrate HemiPy's outputs realistically represent expected temporal patterns. Comparison against litterfall data reveals reasonable accuracies are achievable, with RMSE values close to the error of ~1 unit typically attributed to optical LAI measurement approaches. HemiPy's PAI, FIPAR and FCOVER outputs demonstrate good agreement with CAN‐EYE. Consistent with previous studies, when compared to simulated hemispherical images, better agreement is observed for PAI derived using gap fraction near the hinge angle of 57.5° only, as opposed to values derived using gap fraction over a wider range of zenith angles. HemiPy should prove a useful tool for processing DHP images, and its open‐source nature means that it can be adopted, extended and further refined by the user community.

Snow and ice are important hydrological resources. Their minimum spatial extent over land, here referred to as annual minimum snow/ice (MSI) cover, plays a very important role as an indicator of long-term changes and baseline capacity for surface water storage. Data from Moderate Resolution Imaging Spectroradiometer (MODIS) on Terra satellite for the period of 2000–2014 were utilized in this study. The level-2 MODIS swath imagery for bands B1 to B7 was employed and the 500-m bands B3–B7 were spatially downscaled to a 250-m swath grid. The imagery is available daily with multiple overpasses. This allows for more accurate identification of annual minimum in comparison to high-resolution imagery (e.g., Landsat, ASTER, etc.) available at much coarser temporal rates. Atmospherically corrected 10-day clear-sky composites converted into normalized surface reflectance over the warm season (April 1 to September 20) were employed to identify persistent snow and ice presence. Results were compared with our previous results derived from the MODIS Circumpolar Arctic clear-sky composites, generated for the end of melting season, and showed smaller MSI extent by 24%, on average. Produced MSI distributions were also compared with the permanent snow and ice maps available from 6 global land cover datasets: (i) Global Land Cover GLC-2000, (ii & iii) European Space Agency's (ESA) Globcover circa 2005 and 2009, (iv–vi) land cover maps derived under the ESA Climate Change Initiative (CCI) for 2000, 2005, and 2010. Significant biases were discovered between various land cover datasets and our results. For example, GLC-2000 overestimated snow/ice extent by 194% (325,400 km²) for the Canadian Arctic. The biases over the entire landmass (excluding Greenland) are 135% (3.7 × 10⁵ km²), 113% (3.0 × 10⁵ km²), 89% (2.2 × 10⁵ km²), and 28% (0.8 × 10⁵ km²) between our results and GLC-2000, ESA Globcover 2005, ESA Globcover 2009, and ESA CCI datasets, correspondingly. The derived MSI extent was compared with Randolph Glacier Inventory (RGI) 4.0 and showed much better consistency (ranging from 1% to 15%).

Lichen is an important food source for caribou in Canada. Lichen mapping using remote sensing (RS) images could be a challenging task, however, as lichens generally appear in unevenly distributed, small patches, and could resemble surficial features. Moreover, collecting lichen labeled data (reference data) is expensive, which restricts the application of many robust supervised classification models that generally demand a large quantity of labeled data. The goal of this study was to investigate the potential of using a very-high-spatial resolution (1-cm) lichen map of a small sample site (e.g., generated based on a single UAV scene and using field data) to train a subsequent classifier to map caribou lichen over a much larger area (~0.04 km² vs. ~195 km²) and a lower spatial resolution image (in this case, a 50-cm WorldView-2 image). The limited labeled data from the sample site were also partially noisy due to spatial and temporal mismatching issues. For this, we deployed a recently proposed Teacher-Student semi-supervised learning (SSL) approach (based on U-Net and U-Net++ networks) involving unlabeled data to assist with improving the model performance. Our experiments showed that it was possible to scale-up the UAV-derived lichen map to the WorldView-2 scale with reasonable accuracy (overall accuracy of 85.28% and F1-socre of 84.38%) without collecting any samples directly in the WorldView-2 scene. We also found that our noisy labels were partially beneficial to the SSL robustness because they improved the false positive rate compared to the use of a cleaner training set directly collected within the same area in the WorldView-2 image. As a result, this research opens new insights into how current very high-resolution, small-scale caribou lichen maps can be used for generating more accurate large-scale caribou lichen maps from high-resolution satellite imagery.

Relating ground photographs to UAV orthomosaics is a key linkage required for accurate multi-scaled lichen mapping. Conventional methods of multi-scaled lichen mapping, such as random forest models and convolutional neural networks, heavily rely on pixel DN values for classification. However, the limited spectral range of ground photos requires additional characteristics to differentiate lichen from spectrally similar objects, such as bright logs. By applying a neural network to tiles of a UAV orthomosaics, additional characteristics, such as surface texture and spatial patterns, can be used for inferences. Our methodology used a neural network (UAV LiCNN) trained on ground photo mosaics to predict lichen in UAV orthomosaic tiles. The UAV LiCNN achieved mean user and producer accuracies of 85.84% and 92.93%, respectively, in the high lichen class across eight different orthomosaics. We compared the known lichen percentages found in 77 vegetation microplots with the predicted lichen percentage calculated from the UAV LiCNN, resulting in a R² relationship of 0.6910. This research shows that AI models trained on ground photographs effectively classify lichen in UAV orthomosaics. Limiting factors include the misclassification of spectrally similar objects to lichen in the RGB bands and dark shadows cast by vegetation.