The image spectral data, particularly hyperspectral data, has been proven as an efficient data source for mapping of the spatial variability of soil organic carbon (SOC). Multispectral satellite data are readily available and cost-effective sources of spectral data compared to costly and technically demanding processing of hyperspectral data. Moreover, their continuous acquisition allows to develop a composite from time-series, increasing the spatial coverage of SOC maps. In this study, an evaluation of the prediction ability of models assessing SOC using real multispectral remote sensing data from different platforms was performed. The study was conducted on a study plot (1.45 km²) in the Chernozem region of South Moravia (Czechia). The adopted methods included field sampling and predictive modeling using satellite multispectral Sentinel-2, Landsat-8, and PlanetScope data, and multispectral UAS Parrot Sequoia data. Furthermore, the performance of a soil reflectance composite image from Sentinel-2 data was analyzed. Aerial hyperspectral CASI 1500 and SASI 600 data was used as a reference. Random forest, support vector machine, and the cubist regression technique were applied in the predictive modeling. The prediction accuracy of models using multispectral data, including Sentinel-2 composite, was lower (RPD range from 1.16 to 1.65; RPIQ range from 1.53 to 2.17) compared to the reference model using hyperspectral data (RPD = 2.26; RPIQ = 3.34). The obtained results show very similar prediction accuracy for all spaceborne sensors (Sentinel-2, Landsat-8, and PlanetScope). However, the spatial correlation between the reference mapping results obtained from the hyperspectral data and other maps using multispectral data was moderately strong. UAS sensors and freely available satellite multispectral data can represent an alternative cost-effective data source for remote SOC mapping on the local scale.

One of the best preconditions for the sufficient monitoring of peat bog ecosystems is the collection, processing, and analysis of unique spatial data to understand peat bog dynamics. Over two seasons, we sampled groundwater level (GWL) and soil moisture (SM) ground truth data at two diverse locations at the Rokytka Peat bog within the Sumava Mountains, Czechia. These data served as reference data and were modeled with a suite of potential variables derived from digital surface models (DSMs) and RGB, multispectral, and thermal orthoimages reflecting topomorphometry, vegetation, and surface temperature information generated from drone mapping. We used 34 predictors to feed the random forest (RF) algorithm. The predictor selection, hyperparameter tuning, and performance assessment were performed with the target-oriented leave-location-out (LLO) spatial cross-validation (CV) strategy combined with forward feature selection (FFS) to avoid overfitting and to predict on unknown locations. The spatial CV performance statistics showed low (R² = 0.12) to high (R² = 0.78) model predictions. The predictor importance was used for model interpretation, where temperature had strong impact on GWL and SM, and we found significant contributions of other predictors, such as Normalized Difference Vegetation Index (NDVI), Normalized Difference Index (NDI), Enhanced Red-Green-Blue Vegetation Index (ERGBVE), Shape Index (SHP), Green Leaf Index (GLI), Brightness Index (BI), Coloration Index (CI), Redness Index (RI), Primary Colours Hue Index (HI), Overall Hue Index (HUE), SAGA Wetness Index (TWI), Plan Curvature (PlnCurv), Topographic Position Index (TPI), and Vector Ruggedness Measure (VRM). Additionally, we estimated the area of applicability (AOA) by presenting maps where the prediction model yielded high-quality results and where predictions were highly uncertain because machine learning (ML) models make predictions far beyond sampling locations without sampling data with no knowledge about these environments. The AOA method is well suited and unique for planning and decision-making about the best sampling strategy, most notably with limited data.

This study aimed to examine the potential of convolutional neural networks (CNNs) for the detection of individual trees infested by bark beetles in a multispectral high-resolution dataset acquired by an unmanned aerial system (UAS). We compared the performance of three CNN architectures and the random forest (RF) model to classify the trees into four categories: pines, sbbd (longer infested trees when needles turn yellow), sbbg (trees under green attack) and non-infested trees (sh). The best performance was achieved by the Nez4c3b CNN (kappa 0.80) and Safaugu4c3b CNN (kappa 0.76) using only RGB bands. The main misclassifications were between sbbd and sbbg because of the similar spectral responses. Merging sbbd and sbbg into a more general class of infested trees made the selection of model type less important. All tested model types, including RF, were able to detect infested trees with an F-score of the class over 0.90. Nevertheless, the best overall metrics were achieved again by the Safaugu3c3b model (kappa 0.92) and Nez3cb model (kappa 0.87) using only RGB bands. The performance of both models is comparable, but the Nez model has a higher learning rate for this task. Based on our findings, we conclude that the Nez and Safaugu CNN models are superior to the RF models and transfer learning models for the identification of infested trees and for distinguishing between different infestation stages. Therefore, these models can be used not only for basic identification of infested trees but also for monitoring the development of bark beetle disturbance.

Data on the real extent of soil that is degraded by erosion represent important information for the purposes of conservation policy. However, this type of data is rarely available for large areas. A remote-sensing-based method for identifying of eroded areas at the regional scale has been tested using a combination of time series of free access Sentinel-2 image data, airborne orthoimages and ground-truth data. The unsupervised classification ISODATA of the Sentinel-2A images has been performed. The minimum distance method has been applied for the assignment of unsupervised classes to four erosion classes using the ground-truth data. The automatic classification of eroded soils achieved an overall accuracy of 55.2% for three distinguished classes. An accumulated class has been eliminated as no unsupervised classes were assigned to this erosion class. A simplified classification of two classes (strongly eroded and other soils) reached an accuracy of 80.9%. The overall accuracy of the simplified classification increased to 86.9% after the visual refinement using orthoimages. This study shows the potential of the tested approach to produce valuable data on actual soil degradation by erosion. The limitations of the method are related to the soil cover variability, masking effect of clouds, vegetation or litter and the spectral separability of individual classes.

This study presents a complex empirical image-based radiometric calibration method for a Tetracam μMCA multispectral frame camera. The workflow is based on a laboratory investigation of the camera’s radiometric properties combined with vicarious atmospheric correction using an empirical line. The effect of the correction is demonstrated on out-of-laboratory field campaign data. The dark signal noise behaviour was investigated based on the exposure time and ambient temperature. The vignette effect coupled with nonuniform quantum efficiency was studied with respect to changing exposure times and illuminations to simulate field campaign conditions. The efficiency of the proposed correction workflow was validated by comparing the reflectance values that were extracted from a fully corrected image and the raw data of the reference spectroscopy measurement using three control targets. The Normalized Root Mean Square Errors (NRMSE) of all separate bands ranged from 0.24 to 2.10%, resulting in a significant improvement of the NRMSE compared to the raw data. The results of a field experiment demonstrated that the proposed correction workflow significantly improves the quality of multispectral imagery. The workflow was designed to be applicable to the out-of-laboratory conditions of UAV imaging campaigns in variable natural conditions and other types of multiarray imaging systems.