Recent technical advances in drones make them increasingly relevant and important tools for forest measurements. However, information on how to optimally set flight parameters and choose sensor resolution is lagging behind the technical developments. Our study aims to address this gap, exploring the effects of drone flight parameters (altitude, image overlap, and sensor resolution) on image reconstruction and successful 3D point extraction. This study was conducted using video footage obtained from flights at several altitudes, sampled for images at varying frequencies to obtain forward overlap ratios ranging between 91 and 99%. Artificial reduction of image resolution was used to simulate sensor resolutions between 0.3 and 8.3 Megapixels (Mpx). The resulting data matrix was analysed using commercial multi-view reconstruction (MVG) software to understand the effects of drone variables on (1) reconstruction detail and precision, (2) flight times of the drone, and (3) reconstruction times during data processing. The correlations between variables were statistically analysed with a multivariate generalised additive model (GAM), based on a tensor spline smoother to construct response surfaces. Flight time was linearly related to altitude, while processing time was mainly influenced by altitude and forward overlap, which in turn changed the number of images processed. Low flight altitudes yielded the highest reconstruction details and best precision, particularly in combination with high image overlaps. Interestingly, this effect was nonlinear and not directly related to increased sensor resolution at higher altitudes. We suggest that image geometry and high image frequency enable the MVG algorithm to identify more points on the silhouettes of tree crowns. Our results are some of the first estimates of reasonable value ranges for flight parameter selection for forestry applications.

Harvester-mounted yield monitor sensors are expensive and require calibration and data cleaning. Therefore, we evaluated six vegetation indices (VI) from unmanned aerial system (Quantix™ Mapper) imagery for corn (Zea mays L.) yield prediction. A field trial was conducted with N sidedress treatments applied at four growth stages (V4, V6, V8, or V10) compared against zero-N and N-rich controls. Normalized difference vegetation index (NDVI) and enhanced vegetation index 2 (EVI2), based on flights at R4, resulted in the most accurate yield estimations, as long as sidedressing was performed before V6. Yield estimations based on earlier flights were less accurate. Estimations were most accurate when imagery from both N-rich and zero-N control plots were included, but elimination of the zero-N data only slightly reduced the accuracy. Use of a ratio approach (VI_Trt/VI_N-rich and Yield_Trt/Yield_N-rich) enables the extension of findings across fields and only slightly reduced the model performance. Finally, a smaller plot size (9 or 75 m² compared to 150 m²) resulted in a slightly reduced model performance. We concluded that accurate yield estimates can be obtained using NDVI and EVI2, as long as there is an N-rich strip in the field, sidedressing is performed prior to V6, and sensing takes place at R3 or R4.

Leaf area index (LAI) is an established structural variable that reflects the three-dimensional (3-D) leaf layering of vegetation in response to environmental inputs. In this context, unmanned aerial system (UAS) based methods present a new approach to such plant-to field-scale LAI assessment for precision agriculture applications. This article used UAS-based light detection and ranging (LiDAR) data and multispectral imagery (MSI) as two modalities to evaluate the LAI of a snap bean field, toward eventual yield modeling and disease risk assessment. LiDAR-derived and MSI-derived metrics were fed to multiple biophysical-based and regression models. The correlation between the derived LAI and field-measured LAI was significant. Six LiDAR-derived metrics were fit in eight models to predict LAI, among which the square root of the laser penetration index achieved the most accurate prediction result ( ${R^2}$ = 0.61, n RMSE = 19%). The MSI-derived models, which contained both structural features and spectral signatures, provided similar predicting effectiveness, with predicted ${R^2}$ ≈0.5 and n RMSE≈22%. We furthermore observed variation in model effectiveness for different cultivars, different cultivar groups, and different UAS flight altitudes, for both the LiDAR and MSI approaches. For data collected at a consistent flight altitude, MSI-derived models could even exceed LiDAR-derived models, in terms of accuracy. This finding could support the possibility of replacing LiDAR with more cost-effective MSI-based approaches. However, LiDAR remains a viable modality, since a LiDAR-derived 3-D model only required a single predictor variable, while an MSI-derived model relied on multiple independent variables in our case.

The use of small unmanned aerial system (UAS)-based structure-from-motion (SfM; photogrammetry) and LiDAR point clouds has been widely discussed in the remote sensing community. Here, we compared multiple aspects of the SfM and the LiDAR point clouds, collected concurrently in five UAS flights experimental fields of a short crop (snap bean), in order to explore how well the SfM approach performs compared with LiDAR for crop phenotyping. The main methods include calculating the cloud-to-mesh distance (C2M) maps between the preprocessed point clouds, as well as computing a multiscale model-to-model cloud comparison (M3C2) distance maps between the derived digital elevation models (DEMs) and crop height models (CHMs). We also evaluated the crop height and the row width from the CHMs and compared them with field measurements for one of the data sets. Both SfM and LiDAR point clouds achieved an average RMSE of ~0.02 m for crop height and an average RMSE of ~0.05 m for row width. The qualitative and quantitative analyses provided proof that the SfM approach is comparable to LiDAR under the same UAS flight settings. However, its altimetric accuracy largely relied on the number and distribution of the ground control points.

Forests are an important part natural ecosystems, by for example providing food, fiber, habitat, and biodiversity, all of which contribute to stable natural systems. Assessing and modeling the structure and characteristics of forests, e.g., Leaf Area Index (LAI), volume, biomass, etc., can lead to a better understanding and management of these resources. In recent years, Terrestrial Laser Scanning (TLS) has been recognized as a tool that addresses many of the limitations of manual and traditional forest data collection methods. In this study, we propose a density-based approach for estimating the LAI in a structurally-complex forest environment, which contains variable and diverse structural attributes, e.g., non-circular stem forms, dense canopy and below-canopy vegetation cover, and a diverse species composition. In addition, 242 TLS scans were collected using a portable low-cost scanner, the Compact Biomass Lidar (CBL), in the Hawaii Volcanoes National Park (HAVO), Hawaii Island, USA. LAI also was measured for 242 plots in the site, using an AccuPAR LP-80 ceptometer. The first step after cleaning the point cloud involved detecting the higher forest canopy in the light detection and ranging (lidar) point clouds, using normal change rate assessment. We then estimated Leaf Area Density (LAD), using a voxel-based approach, and divided the canopy point cloud into five layers in the Z (vertical) direction. These five layers subsequently were divided into voxels in the X direction, where the size of these voxels were obtained based on inter-quartile analysis and the number of points in each voxel. We hypothesized that the intensity returned to the lidar system from woody materials, like branches, would be higher than from leaves, due to the liquid water absorption feature of the leaves and higher reflectance for woody material at the 905 nm laser wavelength. We also differentiated between foliar and woody materials using edge detection in the images from projected point clouds and evaluated the density of these regions to support our hypothesis. Density of points, or the number of points divided by the volume of a grid, in a 3D grid size of 0.1 m, was calculated for each of the voxels. The grid size was determined by investigating the size of the branches in the lower portion of the canopy. Subsequently, we fitted a Kernel Density Estimator (KDE) to these values, with the threshold set based on half of the area under the curve in each of the density distributions. All the grids with a density below the threshold were labeled as leaves, while those grids above the threshold were identified as non-leaves. Finally, we modeled LAI using the point densities derived from the TLS point clouds and the listed analysis steps. This model resulted in an $R^2$ value of 0.88. We also estimated the LAI directly from lidar data using the point densities and calculating LAD, which is defined as the total one-sided leaf area per unit volume. LAI can be obtained as the sum of the LAD values in all the voxels. The accuracy of LAI estimation was 90%, with an RMSE value of 0.31, and an average overestimation of $9\%$ in TLS-derived LAI, when compared to field-measured LAI. Algorithm performance mainly was affected by the vegetation density and complexity of the canopy structures. It is worth noting that, since the LAI values cannot be considered spatially independent throughout all the plots in this site, we performed semivariogram analysis on the field-measured LAI data. This analysis showed that the LAI values can be assumed to be independent in plots that are at least 30 m apart. As a result, we divided the data into six subsets in which the plots were 30 m spaced. The $R^2$ values for these subsets, based on modeling of the field-measured LAI using leaf point density values, ranged between 0.84–0.96. The results bode well for using this method for efficient, automatic, and accurate/precise estimation of LAI values in complex forest environments, using a low-cost, rapid-scan TLS.

Remote sensing provides high accuracy/precision for quantifying forest biophysical parameters needed for ecological management. Although the significant impact of Bidirectional Scattering Distribution Functions (BSDF) on remote sensing of vegetation is well-known, current forest metrics derived from sensor data seldom take leaf BSDF into account, and despite the importance of BSDF effects, leaf directional scattering measurements are almost nonexistent. Previous studies have been limited in the spectral coverage and resolution of observed electromagnetic radiation (EMR) and lacked models to interpolate all source-sensor angles beyond measurements. This study captured deciduous broadleaf Bidirectional Reflectance Distribution Functions (BRDFs) from the visible through shortwave-infrared (SWIR) spectral regions (350-2500 nm) and accurately modeled the BRDF for extension to any illumination angle, viewing zenith, or azimuthal angle. We measured bi-conical directional reflectance factor (BCRF) of leaves from three species of large trees, Norway maple (Acer platanoides), American sweetgum (Liquidambar styraciflua), and northern red oak (Quercus rubra). We then fit the data through nonlinear regression to physical, microfacet BRDF models, resulting in normalized root mean square errors of less than 8%, averaged across all wavelengths (excluding low signal-to-noise spectral regions). We extracted leaf physical parameters, including the index of refraction and a relative physical roughness from the microfacet models delineating the three species. The implications for forestry remote sensing are important, as rigorous models to represent leaves allow for the creation of more accurate forest scenes for radiative transfer modeling. Such accuracy enables higher fidelity sensor evaluations and data processing algorithms.