Search

Wheat stripe rust (WSR), a fungal disease capable of inflicting severe crop loss, threatens most of global wheat production. Breeding for genetic resistance is the primary defense against stripe rust infection. Further development of rustresistant wheat varieties depends on the ability to accurately and rapidly quantify rust resilience. In this study we demonstrate the ability of visible through shortwave infrared reflectance spectroscopy to effectively provide high-throughput classification of wheat stripe rust severity and identify important spectral regions for classification accuracy. Random forest models were developed using both leaf-level and canopy-level hyperspectral reflectance observations collected across a breeding population that was scored for WSR severity using 10 and 5 severity classes, respectively. The models were able to accurately diagnose scored disease severity class across these fine scoring scales between 45-52% of the time, which improved to 79-96% accuracy when allowing scores to be off-by-one. The canopy-level model demonstrated higher accuracy and distinct spectral characteristics relative to the leaf-level models, pointing to the use of this technology for field-scale monitoring. Leaf-level model performance was strong despite clear variation in scoring conducted between wheat growth stages. Two approaches to reduce predictor and model complexity, principal component dimensionality reduction and backward feature elimination, were applied here. Both approaches demonstrated that model classification skill could remain high while simplifying high-dimensional hyperspectral reflectance predictors, with parsimonious models having approximately 10 unique components or wavebands. Through the use of a high-resolution infection severity scoring methodology this study provides one of the most rigorous tests of the use of hyperspectral reflectance observations for WSR classification. We demonstrate that machine learning in combination with a few carefully-selected wavebands can be leveraged for precision remote monitoring and management of WSR to limit crop damage and to aid in the selection of resilient germplasm in breeding programs. [ABSTRACT FROM AUTHOR]

Context: Pasture dieback has emerged as a significant threat to the health and productivity of sown pastures in eastern Queensland and northern New South Wales, Australia. Aims: We aimed to address knowledge gaps on spatial spread patterns, recovery trajectories and floristic changes using remote sensing and ground surveys. Methods: We used a time series of high-resolution (12–25 cm) aerial imagery to quantify and compare pasture dieback spread over 7 years in three land-use areas: ungrazed pasture, grazed pasture and rehabilitation following mining. The green leaf index was applied using supervised random forest algorithms to classify areas affected between 2015 and 2021. Flora surveys were conducted to compare impacted and unimpacted areas for the three land uses and validate classifications. Key results: The first emergence of pasture dieback was in ungrazed pasture, and these areas recorded the highest rate of dieback spread at 1.88 ha month−1, compared with 0.54 and 0.19 ha month−1 in rehabilitated and grazed pastures respectively. Field validation showed that dieback-impacted pastures shifted from buffel grass (Cenchrus ciliaris L.), to forb-dominated communities with significantly different species mix, biomass and cover conditions. An analysis of local climate data showed that winter night-time temperatures and rainfall were notably higher than long-term means in the year preceding the first detection of pasture dieback. Conclusions: High resolution aerial imagery and ground surveys can be used to monitor pasture health by employing vegetation indices and random forest classifiers. Implications: Ungrazed pastures and roadside areas should be managed to protect the region from further outbreaks. Pasture dieback has been affecting the health and productivity of pastures in eastern Queensland and northern NSW. We used aerial imagery to map and compare pasture dieback spread patterns in three land use areas: grazed pasture, ungrazed pasture and rehabilitated pasture following coal mining. The fastest rate of spread was recorded in ungrazed pasture followed by rehabilitation and grazed pastures. Flora surveys showed that dieback-impacted sites shifted from buffel grass to forb dominated pastures. [ABSTRACT FROM AUTHOR]

Rubber is a perennial plant grown for natural rubber production, which is used in various global products. Ensuring the sustainability of rubber cultivation is crucial for smallholder farmers and economic development. Accurately predicting rubber yields is necessary to maintain price stability. Remote sensing technology is a valuable tool for collecting spatial data on a large scale. However, for smaller plots of land owned by smallholder farmers, it is necessary to process productivity estimates from high-resolution satellite data that are accurate and reliable. This study examines the impact of spatial factors on rubber yield and evaluates the technical suitability of using grouping analysis with the forest classification and regression (FCR) method. We developed a high-density variable using spatial data from rubber plots in close proximity to each other. Our approach incorporates eight environmental variables (proximity to streamlines, proximity to main river, soil drainage, slope, aspect, NDWI, NDVI, and precipitation) using an FCR model and GIS. We obtained a dataset of 1951 rubber yield locations, which we split into a training set (60%) for model development and a validation set (40%) for assessment using area under the curve (AUC) analysis. The results of the alternative FCR models indicate that Model 1 performs the best. It achieved the lowest root mean square error (RMSE) value of 19.15 kg/ha, the highest R-squared (R2) value (FCR) of 0.787, and also the highest R2 (OLS) value of 0.642. The AUC scores for Model 1, Model 2, and Model 3 were 0.792, 0.764, and 0.732, respectively. Overall, Model 4 exhibited the highest performance according to the AUC scores, while Model 3 performed the poorest with the lowest AUC score. Based on these findings, it can be concluded that Model 1 is the most effective in predicting FCR compared to the other alternative models. [ABSTRACT FROM AUTHOR]

Upland habitats provide vital ecological services, yet they are highly threatened by natural and anthropogenic stressors. Monitoring these vulnerable habitats is fundamental for conservation and involves determining information about their spatial locations and conditions. Remote sensing has evolved as a promising tool to map the distribution of upland habitats in space and time. However, the resolutions of most freely available satellite images (e.g., 10-m resolution for Sentinel-2) may not be sufficient for mapping relatively small features, especially in the heterogeneous landscape—in terms of habitat composition—of uplands. Moreover, the use of traditional remote sensing methods, imposing discrete boundaries between habitats, may not accurately represent upland habitats as they often occur in mosaics and merge with each other. In this context, we used high-resolution (2 m) Pleiades satellite imagery and Random Forest (RF) machine learning to map habitats at two Irish upland sites. Specifically, we investigated the impact of varying spatial resolutions on classification accuracy and proposed a complementary approach to traditional methods for mapping complex upland habitats. Results showed that the accuracy generally improved with finer spatial resolution data, with the highest accuracy values (80.34% and 79.64%) achieved for both sites using the 2-m resolution datasets. The probability maps derived from the RF-based fuzzy classification technique can represent complex mosaics and gradual transitions occurring in upland habitats. The presented approach can potentially enhance our understanding of the spatiotemporal dynamics of habitats over large areas. [ABSTRACT FROM AUTHOR]

Climate change is making water management increasingly difficult due to rising temperatures and unpredictable rainfall patterns, impacting crop water availability and irrigation needs. This study investigated the ability of machine learning and satellite remote sensing to monitor water status and physiology. The research focused on predicting different eco-physiological parameters in an irrigated peach orchard under Mediterranean conditions, utilizing multispectral reflectance data and machine learning algorithms (extreme gradient boosting, random forest, support vector regressor); ground data were acquired from 2021 to 2023 in the south of Italy. The random forest model outperformed in predicting net assimilation (R2 = 0.61), while the support vector machine performed best in predicting electron transport rate (R2 = 0.57), Fv/Fm ratio (R2 = 0.66) and stomatal conductance (R2 = 0.56). Random forest also proved to be the most effective in predicting stem water potential (R2 = 0.62). These findings highlighted the potential of integrating machine learning techniques with high-resolution satellite imagery to assist farmers in monitoring crop health and optimizing irrigation practices, thereby addressing the challenges determined by climate change. [ABSTRACT FROM AUTHOR]

In recent years, the semantic segmentation model has been widely applied in fields such as the extraction of crops due to its advantages such as strong discrimination ability, high accuracy, etc. Currently, there is no standard set of ground true label data for major crops in China, and the visual interpretation process is usually time-consuming and laborious. The sample size also makes it difficult to support the model to learn enough ground features, resulting in poor generalisation ability of the model, which in turn makes the model difficult to apply in fine extraction tasks of large-area crops. In this study, a method to establish a pseudo-label sample set based on the random forest algorithm to train a semantic segmentation model (U-Net) was proposed to perform winter wheat extraction. With the help of the GEE platform, Winter Wheat Canopy Index (WCI) indicators were employed in this method to initially extract winter wheat, and training samples (i.e., pseudo labels) were built for the semantic segmentation model through the iterative process of "generating random sample points—random forest model training—winter wheat extraction"; on this basis, the U-net model was trained with multi-time series remote sensing images; finally, the U-Net model was employed to obtain the spatial distribution map of winter wheat in Henan Province in 2022. The results illustrated that: (1) Pseudo-label data were constructed using the random forest model in typical regions, achieving an overall accuracy of 97.53% under validation with manual samples, proving that its accuracy meets the requirements for U-Net model training. (2) Utilizing the U-Net model, U-Net++ model, and random forest model constructed based on pseudo-label data for 2022, winter wheat mapping was conducted in Henan Province. The extraction accuracy of the three models is in the order of U-Net model > U-Net++ model > random forest model. (3) Using the U-Net model to predict the winter wheat planting areas in Henan Province in 2019, although the extraction accuracy decreased compared to 2022, it still exceeded that of the random forest model. Additionally, the U-Net++ model did not achieve higher classification accuracy. (4) Experimental results demonstrate that deep learning models constructed based on pseudo-labels exhibit higher classification accuracy. Compared to traditional machine learning models like random forest, they have higher spatiotemporal adaptability and robustness, further validating the scientific and practical feasibility of pseudo-labels and their generation strategies, which are expected to provide a feasible technical pathway for intelligent extraction of winter wheat spatial distribution information in the future. [ABSTRACT FROM AUTHOR]

Making plant breeding programs less expensive, fast, practical, and accurate, especially for soybeans, promotes the selection of new soybean genotypes and contributes to the emergence of new varieties that are more efficient in absorbing and metabolizing nutrients. Using spectral information from soybean genotypes combined with nutritional information on secondary macronutrients can help genetic improvement programs select populations that are efficient in absorbing and metabolizing these nutrients. In addition, using machine learning algorithms to process this information makes the acquisition of superior genotypes more accurate. Therefore, the objective of the work was to verify the classification performance of soybean genotypes regarding secondary macronutrients by ML algorithms and different inputs. The experiment was conducted in the experimental area of the Federal University of Mato Grosso do Sul, municipality of Chapadão do Sul, Brazil. Soybean was sown in the 2019/20 crop season, with the planting of 103 F2 soybean populations. The experimental design used was randomized blocks, with two replications. At 60 days after crop emergence (DAE), spectral images were collected with a Sensifly eBee RTK fixed-wing remotely piloted aircraft (RPA), with autonomous takeoff control, flight plan, and landing. At the reproductive stage (R1), three leaves were collected per plant to determine the macronutrients calcium (Ca), magnesium (Mg), and sulfur (S) levels. The data obtained from the spectral information and the nutritional values of the genotypes in relation to Ca, Mg, and S were subjected to a Pearson correlation analysis; a PC analysis was carried out with a k-means algorithm to divide the genotypes into clusters. The clusters were taken as output variables, while the spectral data were used as input variables for the classification models in the machine learning analyses. The configurations tested in the models were spectral bands (SBs), vegetation indices (VIs), and a combination of both. The combination of machine learning algorithms with spectral data can provide important biological information about soybean plants. The classification of soybean genotypes according to calcium, magnesium, and sulfur content can maximize time, effort, and labor in field evaluations in genetic improvement programs. Therefore, the use of spectral bands as input data in random forest algorithms makes the process of classifying soybean genotypes in terms of secondary macronutrients efficient and important for researchers in the field. [ABSTRACT FROM AUTHOR]

Land use and land cover (LULC) analysis is crucial for understanding societal development and assessing changes during the Anthropocene era. Conventional LULC mapping faces challenges in capturing changes under cloud cover and limited ground truth data. To enhance the accuracy and comprehensiveness of the descriptions of LULC changes, this investigation employed a combination of advanced techniques. Specifically, multitemporal 30 m resolution Landsat-8 satellite imagery was utilized, in addition to the cloud computing capabilities of the Google Earth Engine (GEE) platform. Additionally, the study incorporated the random forest (RF) algorithm. This study aimed to generate continuous LULC maps for 2014 and 2020 for the Shrirampur area of Maharashtra, India. A novel multiple composite RF approach based on LULC classification was utilized to generate the final LULC classification maps utilizing the RF-50 and RF-100 tree models. Both RF models utilized seven input bands (B1 to B7) as the dataset for LULC classification. By incorporating these bands, the models were able to influence the spectral information captured by each band to classify the LULC categories accurately. The inclusion of multiple bands enhanced the discrimination capabilities of the classifiers, increasing the comprehensiveness of the assessment of the LULC classes. The analysis indicated that RF-100 exhibited higher training and validation/testing accuracy for 2014 and 2020 (0.99 and 0.79/0.80, respectively). The study further revealed that agricultural land, built-up land, and water bodies have changed adequately and have undergone substantial variation among the LULC classes in the study area. Overall, this research provides novel insights into the application of machine learning (ML) models for LULC mapping and emphasizes the importance of selecting the optimal tree combination for enhancing the accuracy and reliability of LULC maps based on the GEE and different RF tree models. The present investigation further enabled the interpretation of pixel-level LULC interactions while improving image classification accuracy and suggested the best models for the classification of LULC maps through the identification of changes in LULC classes. [ABSTRACT FROM AUTHOR]

Accurate mapping of forest habitats, especially in NATURA sites, is essential information for forest monitoring and sustainable management but also for habitat characterisation and ecosystem functioning. Remote sensing data and spatial modelling allow accurate mapping of the presence and distribution of tree species and habitats and are valuable tools for the long-term assessment of habitat status required by the European Commission. In order to serve the above, the present study aims to propose a methodology to accurately map the spatial distribution of forest habitats in three NATURA2000 sites of Cyprus by employing Sentinel-1 and Sentinel-2 data as well as topographic features using the Google Earth Engine (GEE). A pivotal aspect of the methodology identified was that the best band combination of the Random Forest (RF) classifier achieves the highest performance for mapping the dominant habitats in the three case studies. Specifically, in the Akamas region, eight habitat types have been mapped, in Paphos nine and six in Troodos. These habitat types are included in three of the nine habitat groups based on the EU's Habitat Directive: the sclerophyllous scrub, rocky habitats and caves and forests. The results show that using the RF algorithm achieves the highest performance, especially using Dataset 6, which is based on S2 bands, spectral indices and topographical features, and Dataset 13, which includes S2, S1, spectral indices and topographical features. These datasets achieve an overall accuracy (OA) of approximately 91–94%. In contrast, Dataset 7, which includes only S1 bands and Dataset 9, which combines S1 bands and spectral indices, achieve the lowest performance with an OA of approximately 25–43%. [ABSTRACT FROM AUTHOR]

Machine learning(ML) based techniques for Land Use Land Cover(LULC) classification is crucial for extracting valuable insights from satellite imagery. The impact of severe class imbalance problems on LULC datasets, added to the limited temporal coverage, incomplete spectral information, spatial resolution constraints, and limitations due to sensor characteristics of a single satellite, hinder the efficient capturing of complex image features. The object-based classification using Gray-level co-occurrence matrix(GLCM) and Simple Non-Iterative Clustering(SNIC) captures the textural and spectral information, respectively, to enhance the accuracy in heterogeneous landscapes and overcome the limitations of pixel-based classification, such as the sensitivity towards noise and spectral confusion in mixed pixels at the cost of increased additional computational steps. In this context, the proposed study leverages high-quality fused images with diverse temporal, spectral, and spatial information obtained by fusing Landsat-8 and Sentinel-2 satellite imageries using Spatial-and-Temporal-Adaptive-Reflectance-Fusion-Model(STARFM). Further, to solve the class imbalance problem, a Granular Computing(GrC) based Segmentation and Textural Analysis(GrCSTA) framework is proposed for reducing the number of image primitives and computations required for subsequent image analysis processes in the object-based classification. The GrCSTA framework focuses on extracting the Spatial Granules(Gs) from the fused imagery using Spatial neighborhood Granulation(SNGr), textural indices of the Gs using GLCM, and reduced textural indices using Principal Component Analysis(PCA). Gs and its reduced textural indices are input features to train the Random Forest(RF) classifier. Experimental results demonstrate that the proposed GrCSTA framework achieves comparably higher accuracy than the state-of-the-art models. [ABSTRACT FROM AUTHOR]

Since there has been concern about food security, accurate prediction of wheat yield prior to harvest is a key component. Random Forest (RF) has been used in many classification and regression applications, such as yield estimation, and the performance of RF has improved by tuning its hyperparameters. In this paper, different changes are made to traditional RF for yield estimation, and the performance of RF is evaluated. Accordingly, RFs constructed using various weak learners, as well as a combined RF consisting of different weak learners are assessed by growing weak Gaussian Process Regression (GPR), Decision Tree (DT), Neural Network (NN), and Stepwise Regression (SW) models in the forest. The input data to DTs are also partitioned into leaves in N (e.g., two) dimensional feature space by using clustering in each parent node. In addition, a subset of the training set is randomly sampled with replacement for training a learner in the forest, instead of randomly sampling the whole training set in traditional RF. Using clustering in DTs added flexibility while utilizing NN as a weak learner yielded the most favorable outcomes in our research. The number of input training samples (Itree) to each tree was also identified as a new hyperparameter to the forest, and the prediction results were more influenced by the Itree compared to the known hyperparameters, such as the number of trees in the forest (ntree) and the number of features for each tree (mtry). [ABSTRACT FROM AUTHOR]

Introduction: Land use classification plays a critical role in analyzing land use/cover change (LUCC). Remote sensing land use classification based on machine learning algorithm is one of the hot spots in current remote sensing technology research. The diversity of surface objects and the complexity of their distribution in mixed mining and agricultural areas have brought challenges to the classification of traditional remote sensing images, and the rich information contained in remote sensing images has not been fully utilized. Methods: A quantitative difference index was proposed quantify and select the texture features of easily confused land types, and a random forest (RF) classification method with multi-feature combination classification schemes for remote sensing images was developed, and land use information of the mine-agriculture compound area of Peixian in Xuzhou, China was extracted. Results: The quantitative difference index proved effective in reducing the dimensionality of feature parameters and resulted in a reduction of the optimal feature scheme dimension from 57 to 22. Among the four classification methods based on the optimal feature classification scheme, the RF algorithm emerged as the most efficient with a classification accuracy of 92.38% and a Kappa coefficient of 0.90, which outperformed the support vector machine (SVM), classification and regression tree (CART), and neural network (NN) algorithm. Conclusion: The findings indicate that the quantitative differential index is a novel and effective approach for discerning distinct texture features among various land types. It plays a crucial role in the selection and optimization of texture features in multispectral remote sensing imagery. Random forest (RF) classification method, leveraging a multi-feature combination, provides a fresh method support for the precise classification of intricate ground objects within the mine-agriculture compound area. [ABSTRACT FROM AUTHOR]

Coastal dunes are considered some of the most threatened and vulnerable habitats in the European Union. Mapping the spatial distribution of these habitats is an essential task for their conservation. Advances in Unoccupied Aerial Vehicles (UAVs) facilitate the flexible acquisition of high-resolution imagery for identifying detailed spatial distributions of habitats within dune systems. This study aimed to assess the effectiveness of UAV remote sensing for mapping these habitat types. Specifically, we determined the impact of temporally acquired UAV-derived spectral and topographic information on classification accuracy. The work combined the multi-temporal UAV imagery with field observation data and used the Random Forest machine learning algorithm to classify dune habitats. Results showed that using multi-temporal UAV imagery increased classification accuracy compared to using uni-temporal UAV imagery (92.37% vs. 84.09%, respectively). Also, including topographic information consistently improved accuracy, regardless of the number of image sets used (the highest accuracy increased from 84.81% to 92.57% for a uni-temporal model). Temporal analyses showed that the data acquired in the middle period of the growing season were better than those acquired in the early or late periods. The methodology presented here demonstrates the potential of using UAV data for detailed mapping and monitoring of habitat types. [ABSTRACT FROM AUTHOR]

India is the world's second-largest exporter of agricultural products. Agriculture is India's most populated sector of the economy which plays a key part in the country's overall socio-economic fabric. Crop yield prediction is crucial yet desirable for every country to ensure food security. Machine learning can be a very important decision support tool for the prediction of crop yield which can help in preparing for the future in a much more substantial way. Many studies have been conducted for crop yield prediction using multi-source data, still, the predictions of the crop of the Rajasthan region in India have not been investigated. In the present study well-known machine learning techniques, decision tree, random forest, lasso regression, and support vector regression have been implemented for crop yield prediction using multi-source data. Machine learning techniques were implemented on data set that consists of Vegetation Indices, extracted from remote sensing data and Weather data. Random forest attained the more accurate prediction with the coefficient of determination (R2) of 0.77, root mean square error of 0.39 t/ha and mean absolute error of 0.28 t/ha. The research obtained a good estimation of crop yield prior to the harvest. The crop yield is impacted by the change in environmental factors, this research would assist the farmers to have a prior estimation of the yield. The study also reveals that gradually adding more data to the model increases the overall prediction rate. [ABSTRACT FROM AUTHOR]

Land use cannot be simply understood as land cover. The same land may carry different functions, such as production, living, and ecological applications; the dominant function of land will affect and restrict other uses. Disorderly urbanization and industrialization have led to an intensification of conflicts among the production, living, and ecological functions of land, which is a major constraint on regional sustainable development. This paper took the perspective of land-use function and used multi-source data such as Sentinel remote-sensing imagery, VIIRS night-time light data, and POIs to classify land-use functions on a large scale in the Beijing–Tianjin–Hebei (BTH) urban agglomeration. The specific research process was as follows. Firstly, the BTH region was multi-scale-segmented based on Sentinel remote-sensing data. Then, the spectral, texture, shape, and socio-economic features of each small area after segmentation were extracted. Moreover, a PLES land-use classification system oriented towards land-use function was established, and a series of representative samples were selected. Subsequently, a random forest model was trained using these samples; then, the trained model was used for the large-scale analysis of land use in the entire BTH region. Finally, the spatial distribution patterns and temporal–spatial evolution characteristics of PLES in the BTH region from 2016 to 2021 were analyzed from the macro level to the micro level. [ABSTRACT FROM AUTHOR]

Present approaches in PV (Photovoltaic) detection are known to be scalable to a larger area using machine learning classification and have improved accuracy on a regional scale with deep learning diagnostics. However, it may cause false detection, time, and cost-consuming when regional deep learning models are directly scaled to a larger area, particularly in large-scale, highly urbanized areas. Thus, a novel two-step downscaling methodology integrating machine learning broad spatial partitioning (step-1) and detailed deep learning diagnostics (step-2) is designed and applied in highly urbanized Jiangsu Province, China. In the first step, this methodology selects suitable feature combinations using the recursive feature elimination with distance correlation coefficient (RFEDCC) strategy for the random forest (RF), considering not only feature importance but also feature independence. The results from RF (overall accuracy = 95.52%, Kappa = 0.91) indicate clear boundaries and little noise. Furthermore, the post-processing of noise removal with a morphological opening operation for the extraction result of RF is necessary for the purpose that less high-resolution remote sensing tiles should be applied in the second step. In the second step, tiles intersecting with the results of the first step are selected from a vast collection of Google Earth tiles, reducing the computational complexity of the next step in deep learning. Then, the improved HRNet with high performance on the test data set (Intersection over Union around 94.08%) is used to extract PV plants from the selected tiles, and the results are mapped. In general, for Jiangsu province, the detection rate of the previous PV database is higher than 92%, and this methodology reduces false detection noise and time consumption (around 95%) compared with a direct deep learning methodology. [ABSTRACT FROM AUTHOR]

Weeds have a significant impact on the growth of rice. Accurate information about weed infestations can provide farmers with important information to facilitate the precise use of chemicals. In this study, we utilized visible light images captured by UAVs to extract information about weeds in areas of two densities on farmland. First, the UAV images were segmented using an optimal segmentation scale, and the spectral, texture, index, and geometric features of each segmented object were extracted. Cross-validation and recursive feature elimination techniques were combined to reduce the dimensionality of all features to obtain a better feature set. Finally, we analyzed the extraction effect of different feature dimensions based on the random forest (RF) algorithm to determine the best feature dimensions, and then we further analyzed the classification result of machine learning algorithms, such as random forest, support vector machine (SVM), decision tree (DT), and K-nearest neighbors (KNN) and compared them based on the best feature dimensions. Using the extraction results of the best classifier, we created a zoning map of the weed infestations in the study area. The results indicated that the best feature subset achieved the highest accuracy, with respective overall accuracies of 95.38% and 91.33% for areas with dense and sparse weed densities, respectively, and F1-scores of 94.20% and 90.57. Random forest provided the best extraction results for each machine learning algorithm in the two experimental areas. When compared to the other algorithms, it improved the overall accuracy by 1.74–12.14% and 7.51–11.56% for areas with dense and sparse weed densities, respectively. The F1-score improved by 1.89–17.40% and 7.85–10.80%. Therefore, the combination of object-based image analysis (OBIA) and random forest based on UAV remote sensing accurately extracted information about weeds in areas with different weed densities for farmland, providing effective information support for weed management. [ABSTRACT FROM AUTHOR]

Groundwater availability is one of the key anxieties in most semi-arid regions of Ethiopia. The purpose of this study was to investigate the groundwater potential zone map of the alluvial plain of Gambela. The study applied analytic hierarchy process (AHP) models with four different machine learning algorithms: random forest classifier (RFC), gradient boosting classifier (GBC), decision tree classifier (DTC), and K-neighbor classifier (KNC). The features that are used as predictors include geology, geomorphology, slope, soil, lineament density, drainage density, land use and land cover (LULC), normalized difference vegetation index (NDVI), topographic wetness index (TWI), topographic roughness index (TRI), and rainfall. The final output of the groundwater potential zone was classified as low, moderate, high, and very high potential zones. The authentication through receiver operating curve (ROC) shows 78.2, 93.4, 92.5, 72.4, and 87.7% values of area under the curve (AUC) for AHP, RFC, GBC, DTC, and KNC, respectively. The results show that RFC and GBC are the best groundwater potential zone (GWPZ) map estimator. The study also shows that rainfall and geomorphology are the primary factors influencing the GWPZ. The outcome might promote improved management alternatives in other areas of the country with a comparable climate. [ABSTRACT FROM AUTHOR]

The exploration of buried mineral deposits is required to generate innovative approaches and the integration of multi-source geoscientific datasets. Mining geochemistry methods have been generated based on the theory of multi-formational geochemical dispersion haloes. Satellite remote sensing data is a form of surficial geoscience datasets and can be considered as big data in terms of veracity and volume. The different alteration zones extracted using remote sensing methods have not been yet categorized based on the mineralogical and geochemical types (MGT) of anomalies and cannot discriminate blind mineralization (BM) from zone dispersed mineralization (ZDM). In this research, an innovative approach was developed to optimize remote sensing-based evidential variables using some constructed mining geochemistry models for a machine learning (ML)-based copper prospectivity mapping. Accordingly, several main steps were implemented and analyzed. Initially, the MGT model was executed by studying the distribution of indicator elements of lithogeochemical data extracted from 50 copper deposits from Commonwealth of Independent States (CIS) countries to identify the MGT of geochemical anomalies associated with copper mineralization. Then, the geochemical zonality model was constructed using the database of the porphyry copper deposits of Iran and Kazakhstan to evaluate the geochemical anomalies related to porphyry copper mineralization (e.g., the Saghari deposit located around the Chah-Musa deposit, Toroud-Chah Shirin belt, central north Iran). Subsequently, the results of mining geochemistry models were used to produce the geochemical evidential variable by vertical geochemical zonality (Vz) (Pb × Zn/Cu × Mo) and to optimize the remote sensing-based evidential variables. Finally, a random forest algorithm was applied to integrate the evidential variables for generating a provincial-scale prospectivity mapping of porphyry copper deposits in the Toroud-Chah Shirin belt. The results of this investigation substantiated that the machine learning (ML)-based integration of multi-source geoscientific datasets, such as mining geochemistry techniques and satellite remote sensing data, is an innovative and applicable approach for copper mineralization prospectivity mapping in metallogenic provinces. [ABSTRACT FROM AUTHOR]

Early prediction of crop production by remote sensing (RS) may help to plan the harvest and ensure food security. This study aims to improve the quantification of yield, grain protein concentration (GPC), and nitrogen (N) output in winter wheat with RS imagery. Ground-truth wheat traits were measured at flowering and harvest in a field experiment combining four N and two water levels in central Spain over 2 years. Hyperspectral and thermal airborne images coincident with Sentinel-1 and Sentinel-2 were acquired at flowering. A parametric linear model using all hyperspectral normalized difference spectral indices (NDSI) and two non-parametric models (artificial neural network and random forest) were used to assess their estimation ability combining NDSIs and other RS indicators. The feasibility of using freely available multispectral satellite was tested by applying the same methodology but using Sentinel-1 and Sentinel-2 bands. Yield estimation obtained the highest R2 value, showing that the visible and short-wave infrared region (VSWIR) had similar accuracy to the hyperspectral and Sentinel-2 imagery (R2 ≈ 0.84). The SWIR bands were important in the GPC estimation with both sensors, whereas N output was better estimated using red-edge-based NDSIs, obtaining satisfactory results with the hyperspectral sensor (R2 = 0.74) and with the Sentinel-2 (R2 = 0.62). When including the Sentinel-2 SWIR index, the NDSI (B11, B3) improved the estimation of N output (R2 = 0.71). Ensemble models based on Sentinel were found to be as reliable as those based on hyperspectral imagery, and including SWIR information improved the quantification of N-related traits. [ABSTRACT FROM AUTHOR]

The estimation of aboveground biomass (AGB) in agroforestry systems using remote sensing has proliferated in the last decades. Similarly, machine learning is also being used in AGB assessments. This study reviews the applications of remote sensing and machine learning for AGB estimation in agroforestry systems (AFS). A detailed review was conducted using 33 recent papers by extracting and comparing information on agroforestry type, data sources, methodology, and model accuracy. Statistical tests were performed to evaluate the differences in performances. High- and very-high-resolution imageries (less than 2 m) are widely used for AGB assessment because they helped to delineate heterogeneous features of AFS. Object-based image analysis yielded classification accuracy of up to 90 percent in some cases. Random Forest, Stochastic Gradient Boosting, and Support Vector Regression are the most common algorithms used for AGB estimation. However, there are no statistically significant differences in the performance between machine learning and other models. Similarly, scholars incorporated spectral indices with spectral bands, texture, and biophysical variables as covariate categories into AGB estimation models. The study finds no significant differences in results (R-squared) by adding more covariate categories. The accuracy of AGB estimates depends upon multiple factors, such as the spectral and spatial resolution, number and types of covariates, methods for AFS delineation and AGB estimation, and types and sizes of AFS. Despite some of the methodological challenges around measuring understory vegetation, advancements in cloud computing like Google Earth Engine and the availability of high-resolution datasets present opportunities for wider use of remote sensing for biomass estimation of AFS. Remote sensing and machine learning have the potential to estimate aboveground biomass over a large area with high accuracy and contribute to carbon monitoring. [ABSTRACT FROM AUTHOR]

Nitrogen (N) and phosphorus (P) are primary indicators of soil nutrients in agriculture. Accurate management of these nutrients is essential for ensuring food security. High-resolution, multi-spectral remote sensing images can provide crucial information for mapping soil nutrients at the field scale. This study compares the capabilities of ZH-1 and Sentinel-2 satellite data, along with different spectral indices, in mapping soil nutrients (total N and Olsen-P) using two machine learning algorithms, random forest (RF) and XGBoost (XGB). Two agricultural fields in Suihua City were selected as the study areas for this investigation. The results showed that Sentinel-2 data performed best in computing the total N content in soil using the RF model (R2 = 0.74, RMSE = 0.10 g/kg). However, for the soil Olsen-P content, the XGBoost model performed better with ZH-1 data (R2 = 0.75, RMSE = 9.79 mg/kg) than the RF model. This study demonstrates that both ZH-1 and Sentinel-2 satellite data perform well in terms of accurately mapping soil total N and Olsen-P contents using machine learning. Due to its higher spectral and spatial resolution, ZH-1 remote sensing data provides more detailed information on soil nutrient content during Olsen-P inversion and exhibits comparable accuracy. [ABSTRACT FROM AUTHOR]

Eutrophication happens when water bodies are enriched by minerals and nutrients. Dense blooms of noxious are the most obvious effect of eutrophication that harms water quality, and by increasing toxic substances damage the water ecosystem. Therefore, it is critical to monitor and investigate the development process of eutrophication. The concentration of chlorophyll-a (chl-a) in water bodies is an essential indicator of eutrophication in them. Previous studies in predicting chlorophyll-a concentrations suffered from low spatial resolution and discrepancies between predicted and observed values. In this paper, we used various remote sensing and ground observation data and proposed a novel machine learning–based framework, a random forest inversion model, to provide the spatial distribution of chl-a in 2 m spatial resolution. The results showed our model outperformed other base models, and the goodness of fit improved by over 36.6% while MSE and MAE decreased by over 15.17% and over 21.26% respectively. Moreover, we compared the feasibility of GF-1 and Sentinel-2 remote sensing data in chl-a concentration prediction. We found that better prediction results can be obtained by using GF-1 data, with the goodness of fit reaching 93.1% and MSE only 3.589. The proposed method and findings of this study can be used in future water management studies and as an aid for decision-makers in this field. [ABSTRACT FROM AUTHOR]

The desert steppe serves as a transitional zone between grasslands and deserts, and long-term monitoring of aboveground biomass (AGB) in the desert steppe is essential for understanding grassland changes. While AGB observation techniques based on multisource remote-sensing data and machine-learning algorithms have been widely applied, research on monitoring methods specifically for the desert steppe remains limited. In this study, we focused on the desert steppe of Inner Mongolia, China, as the study area and used field sampling data, MODIS data, MODIS-based vegetation indices (VI), and environmental factors (topography, climate, and soil) to compare the performance of four commonly used machine-learning algorithms: multiple linear regression (MLR), partial least-squares regression (PLS), random forest (RF), and support vector machine (SVM) in AGB estimation. Based on the optimal model, the spatial–temporal characteristics of AGB from 2000 to 2020 were calculated, and the driving forces of climate change and human activities on AGB changes were quantitatively analyzed using the random forest algorithm. The results are as follows: (1) RF demonstrated outstanding performance in terms of prediction accuracy and model robustness, making it suitable for AGB estimation in the desert steppe of Inner Mongolia; (2) VI contributed the most to the model, and no significant difference was found between soil-adjusted VIs and traditional VIs. Elevation, slope, precipitation, and temperature all had positive effects on the model; (3) from 2000 to 2020, the multiyear average AGB in the study area was 58.34 g/m2, exhibiting a gradually increasing distribution pattern from the inner region to the outer region (from north to south); (4) from 2000 to 2020, the proportions of grassland with AGB slightly and significantly increasing trend in the study area were 87.08% and 5.13%, respectively, while the proportions of grassland with AGB slightly and significantly decreasing trend were 7.76% and 0.05%, respectively; and (5) over the past 20 years, climate change, particularly precipitation, has been the primary driving force behind AGB changes of the study area. This research holds reference value for improving desert steppe monitoring capabilities and the rational planning of grassland resources. [ABSTRACT FROM AUTHOR]

The understating of natural resources and hazards is very fundamental from spatial perspective to reduce loss of resource and human lives. For this, the main aim of the current research is to synthesize the three natural resources (groundwater, wetland, and forest) into one multi-resource (MR) potentiality map and three natural hazards (flood, landslide, and shoreline erosion) into one multi-hazard (MH) susceptibility map. To achieve this goal, several inventories, geo-environmental factors, satellite bands, and indices have been taken as input data sets. Random forest (RF) model has been considered for individual mapping (except shoreline erosion mapping) due to its high precision. For shoreline erosion mapping, digital shoreline analysis system (DSAS) model has been employed. The accuracies of the model have been determined from the area under receiver operating characteristic (AUROC) curve, producer accuracy (PA), and user accuracy (UA). RF model has more than 90% prediction accuracy in different resource and hazard mapping, whereas for the forest mapping, PA and UA are 90% and 80%, respectively. The results of this study for multi-resource mapping show forest (24.05%) and groundwater (22.28%) as the major resources, whereas flood (17.27%) is the most destructive in comparison with other hazards. The multi-resource and multi-hazard maps of the research area provide an important tool to land managers and policymakers for sustainable development and management. [ABSTRACT FROM AUTHOR]

Operational weather satellites, dating back to 1970s, currently provide the best basis for climatological investigations, such as an analysis of changes in the cloud cover. Because clouds are highly dynamic in time, temporally high-resolution data from the geostationary orbit are preferred in order to take variations in the diurnal cycles into account. For such studies, a consistent dataset in space and time is mandatory, but not yet available. Ground-based point measurements of various cloud parameters, such as ceiling, visibility, and cloud type are often sparsely spread and inconsistent, making it difficult to derive reliable spatio-temporal information over large areas. The Meteosat program has generally provided suitable data from over Europe since 1977, but different spatial, spectral, and radiometric resolution of the instruments of the individual satellites, including early-years calibration uncertainties, makes harmonization necessary to finally derive a time series applicable to any kind of climatological study. In this study, a machine learning-based approach has been employed to generate a long-term consistent dataset with high spatio-temporal resolution and extensive coverage over Europe by the harmonization of Meteosat First Generation (MFG) and Meteosat Second Generation (MSG) satellite datasets (1991–2020). A random forest (RF) regressor is trained on the overlap period (2004–2006), where datasets of both satellite generation (MFG and MSG) are available to predict MFG Water Vapour (WV) and Infrared (IR) channels brightness temperature (BT) values based on MSG channels. The aim of the study is to synthesize MFG MVIRI data from MSG SEVIRI to generate a consistent MFG time series. The results indicate a good match of MFG synthesized data with the original MFG data with a mean absolute error of 0.7 K for the WV model and 1.6 K for the IR model, and an out-of-bag (OOB) R² score of 0.98 for both the models. Based on the trained models, the MFG scenes are synthesized from the MSG scenes for the years from 2006 to 2020. The long-term homogeneity of the generated time series is analyzed. The harmonized dataset will be applied to generate a continuous time series on fog and low stratus (FLS) occurrence for a climatological time scale of 30 years. [ABSTRACT FROM AUTHOR]

Lithological information is important in mineral resource exploration, geological observations, mine planning or degradation vulnerability assessment. Currently, lithology mapping can be performed in a fast, inexpensive, and easy way using remote sensing data and machine learning. Remote sensing techniques have become a valuable and promising tool for mapping lithological units and searching for minerals. Typically, the integration of remote sensing data with geophysical data provides a better diagnosis to lithological units than single-source mapping methodologies. Accordingly, this study used a combination of remote sensing and airborne geophysical data utilizing the Random Forest algorithm with small training samples to enhance lithology mapping in Komopa, Papua Province, Indonesia. Geophysical data consisting of magnetic, electromagnetic, and radiometric were added one by one gradually to the remote sensing data, which includes Sentinel 2A, ALOS PALSAR, and DEM (digital elevation model) to compare the accuracy of the classification results from each dataset. The results showed that the model that combined remote sensing data and the three types of geophysical data produced the best classification, with an overall accuracy of 0.81, precision of 0.66, recall of 0.47, and F1 score of 0.52. This fused data can increase the accuracy of the classification results by 8% overall accuracy, 6% precision, 11% recall, and 13% F1 score when compared to the model that only used remote sensing data. [ABSTRACT FROM AUTHOR]

Fire severity mapping can capture heterogeneous fire severity patterns over large spatial extents. Although numerous remote sensing approaches have been established, regional-scale fire severity mapping at fine spatial scales (<5 m) from high-resolution satellite images is challenging. The fire severity of a vast forest fire that occurred in Southwest China was mapped at 2 m spatial resolution by random forest models using Sentinel 2 and GF series remote sensing images. This study demonstrated that using the combination of Sentinel 2 and GF series satellite images showed some improvement (from 85% to 91%) in global classification accuracy compared to using only Sentinel 2 images. The classification accuracy of unburnt, moderate, and high severity classes was significantly higher (>85%) than the accuracy of low severity classes in both cases. Adding high-resolution GF series images to the training dataset reduced the probability of low severity being under-predicted and improved the accuracy of the low severity class from 54.55% to 72.73%. RdNBR was the most important feature, and the red edge bands of Sentinel 2 images had relatively high importance. Additional studies are needed to explore the sensitivity of different spatial scales satellite images for mapping fire severity at fine spatial scales across various ecosystems. [ABSTRACT FROM AUTHOR]

Mapping large wildfires (LW) is essential for environmental applications and enhances the understanding of the dynamics of affected areas. Remote sensing techniques supported by machine learning and time series have been increasingly used in studies addressing this issue and have shown potential for this type of analysis. The main aim of this article is to develop a methodology for mapping LW in northwestern Portugal using a machine learning algorithm and time series from Landsat images. For the burnt area classification, we initially used the Fourier harmonic model to define outliers in the time series that represented pixels of possible burnt areas and, then, we applied the random forest classifier for the LW classification. The results indicate that the harmonic analysis provided estimates with the actual observed values of the NBR index; thus, the pixels classified by random forest were only those that were masked, collaborated in the processing, and reduced possible spectral confusion between targets with similar behaviour. The burnt area maps revealed that ~23.5% of the territory was burnt at least once from 2001 to 2020. The temporal variability of the burnt area indicated that, on average, 6.504 hectares were affected by LW within the 20 years. The annual burnt area varied over the years, with the minimum annual area detected in 2014 (679.5 hectares) and the maximum mapped area detected in 2005 (73,025.1 hectares). We concluded that the process of defining the mask with the outliers considerably reduced the universe of pixels to be classified within each image, which leaves the training of the classifier focused on separating the set of pixels into two groups with very similar spectral characteristics, thus contributing so that the separation of groups with similar spectral behaviour was performed automatically and without great sampling effort. The method showed satisfactory accuracy results with little omission for burnt areas. [ABSTRACT FROM AUTHOR]

Sen2-Agri is a software system that was developed to facilitate the use of multi-temporal satellite data for crop classification with a random forest (RF) classifier in an operational setting. It automatically ingests and processes Sentinel-2 and LandSat 8 images. Our goal was to provide practitioners with recommendations for the best sample size and composition. The study area was located in the Yaqui Valley in Mexico. Using polygons of more than 6000 labeled crop fields, we prepared data sets for training, in which the nine crops had an equal or proportional representation, called Equal or Ratio, respectively. Increasing the size of the training set improved the overall accuracy (OA). Gains became marginal once the total number of fields approximated 500 or 40 to 45 fields per crop type. Equal achieved slightly higher OAs than Ratio for a given number of fields. However, recall and F-scores of the individual crops tended to be higher for Ratio than for Equal. The high number of wheat fields in the Ratio scenarios, ranging from 275 to 2128, produced a more accurate classification of wheat than the maximal 80 fields of Equal. This resulted in a higher recall for wheat in the Ratio than in the Equal scenarios, which in turn limited the errors of commission of the non-wheat crops. Thus, a proportional representation of the crops in the training data is preferable and yields better accuracies, even for the minority crops. [ABSTRACT FROM AUTHOR]

BACKGROUND: Ecballium elaterium (common name: squirting cucumber) is an emerging weed problem in hedgerow or superintensive olive groves under no tillage. It colonizes the inter-row area infesting the natural or sown cover crops, and is considered a hard-to-control weed. Research in other woody crops has shown E. elaterium has a patchy distribution, which makes this weed susceptible to design a site-specific control strategy only addressed to E. elaterium patches. Therefore, the aim of this work was to develop a methodology based on the analysis of imagery acquired with an uncrewed aerial vehicle (UAV) to detect and map E. elaterium infestations in hedgerow olive orchards. RESULTS: The study was conducted in two superintensive olive orchards, and the images were taken using a UAV equipped with an RGB sensor. Flights were conducted on two dates: in May, when there were various weeds infesting the orchard, and in September, when E. elaterium was the only infesting weed. UAV-orthomosaics in the first scenario were classified using random forest models, and the orthomosaics from September with E. elaterium as the only weed, were analyzed using an unsupervised algorithm. In both cases, the overall accuracies were over 0.85, and the producer's accuracies for E. elaterium ranged between 0.74 and 1.00. CONCLUSION: These results allow the design of a site-specific and efficient herbicide control protocol which would represent a step forward in sustainable weed management. The development of these algorithms in free and open-source software fosters their application in small and medium farms. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Remote Sensing is the property of Editorial Office of Journal of Remote Sensing & Science Publishing Co. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Predicting maize yield using spectral information, temperature, and different irrigation management through machine learning algorithms provide information in a fast, accurate, and non-destructive way. The use of multispectral sensor data coupled with irrigation management in maize allows further exploration of water behavior and its relationship with changes in spectral bands presented by the crop. Thus, the objective of this study was to evaluate, by means of multivariate statistics and machine learning techniques, the relationship between irrigation management and spectral bands in predicting maize yields. Field experiments were carried out over two seasons (first and second seasons) in a randomized block design with four treatments (control and three additional irrigation levels) and eighteen sample repetitions. The response variables analyzed were vegetation indices (IVs) and crop yield (GY). Measurement of spectral wavelengths was performed with the Sensefly eBee RTK, with autonomous flight control. The eBee was equipped with the Parrot Sequoia multispectral sensor acquiring reflectance at the wavelengths of green (550 nm ± 40 nm), red (660 nm ± 40 nm), red-edge (735 nm ± 10 nm), and NIR (790 nm ± 40 nm). The blue length (496 nm) was obtained by additional RGB imaging. Data were subjected to Pearson correlations (r) between the evaluated variables represented by a correlation and scatter plot. Subsequently, the canonical analysis was performed to verify the interrelationship between the variables evaluated. Data were also subjected to machine learning (ML) analysis, in which three different input dataset configurations were tested: using only irrigation management (IR), using irrigation management and spectral bands (SB+IR), and using irrigation management, spectral bands, and temperature (IR+SB+Temp). ML models used were: Artificial Neural Network (ANN), M5P Decision Tree (J48), REPTree Decision Tree (REPT), Random Forest (RF), and Support Vector Machine (SVM). A multiple linear regression (LR) was tested as a control model. Our results revealed that Random Forest has higher accuracy in predicting grain yield in maize, especially when associated with the inputs SB+IR and SB+IR+Temp. [ABSTRACT FROM AUTHOR]

Remote sensing data provide significant information about surface geological features, but they have not been fully investigated as a tool for delineating mineral prospective targets using the latest advancements in machine learning predictive modeling. In this study, besides available geological data (lithology, structure, lineaments), Landsat-8, Sentinel-2, and ASTER multispectral remote sensing data were processed to produce various predictor maps, which then formed four distinct datasets (namely Landsat-8, Sentinel-2, ASTER, and Data-integration). Remote sensing enhancement techniques, including band ratio (BR), principal component analysis (PCA), and minimum noise fraction (MNF), were applied to produce predictor maps related to hydrothermal alteration zones in Hamissana area, while geological-based predictor maps were derived from applying spatial analysis methods. These four datasets were used independently to train a random forest algorithm (RF), which was then employed to conduct data-driven gold mineral prospectivity modeling (MPM) of the study area and compare the capability of different datasets. The modeling results revealed that ASTER and Sentinel-2 datasets achieved very similar accuracy and outperformed Landsat-8 dataset. Based on the area under the ROC curve (AUC), both datasets had the same prediction accuracy of 0.875. However, ASTER dataset yielded the highest overall classification accuracy of 73%, which is 6% higher than Sentinel-2 and 13% higher than Landsat-8. By using the data-integration concept, the prediction accuracy increased by about 6% (AUC: 0.938) compared with the ASTER dataset. Hence, these results suggest that the framework of exploiting remote sensing data is promising and should be used as an alternative technique for MPM in case of data availability issues. [ABSTRACT FROM AUTHOR]

Land-use and land-cover (LULC) mapping in the complex area is a challenging task due to the mixed vegetation patterns, and rough mountains with fast-flowing rivers. In Vietnam, LULC update is not frequently. In this study, we applied a supervised machine learning (Random forest—RF) approach to mapping LULC in Thanh Hoa province, Vietnam from 2011 to 2015 utilizing multitemporal Normalized Difference Vegetation Index (NDVI) data from MODIS, combined with topographic features. Random forest classification (RFC) reached a total prediction accuracy of 91% and Kappa coefficient (K) of 0.89 across eight LULCs. Besides, the results showed that the features extracted from time-series NDVI comprising the mean of yearly NDVI, the sum of NDVI, and the topography were the important variables controlling the LULC classification. For similar studies on the distribution of LULC, the method proposed in this study could be helpful. [ABSTRACT FROM AUTHOR]

Mapping the spatial distribution of soil organic carbon (SOC) in lands covered by tropical forests is important to understand the relationship and dynamics of SOC in this type of ecosystem. In this study, the Random Forest (RF) algorithm was used to map SOC stocks of topsoil (0–15 cm) in forest lands of the Dominican Republic. The methodology was developed using geospatial datasets available in the Google Earth Engine (GEE) platform combined with a set of 268 soil samples. Twenty environmental covariates were analyzed, including climate, topography, and vegetation. The results indicate that Model A (combining all 20 covariates) was only marginally better than Model B (combining topographic and climatic covariates), and Model C (only combining multispectral remote sensing data derived from Landsat 8 OLI images). Model A and Model B yielded SOC mean values of 110.35 and 110.87 Mg C ha−1, respectively. Model A reported the lowest prediction error and uncertainty with an R2 of 0.83, an RMSE of 35.02 Mg C ha−1. There was a strong dependence of SOC stocks on multispectral remote sensing data. Therefore, multispectral remote sensing proved accurate to map SOC stocks in forest ecosystems in the region. [ABSTRACT FROM AUTHOR]

The utilization of satellite remote sensing images for retrieving aerosol optical parameters has been extensively discussed over the past few decades. While employing machine learning models is indeed a viable approach, a significant portion of these studies still rely on redundant data. Moreover, the discussion regarding aerosol absorption, a crucial factor for determining aerosol radiative impact and distinguishing aerosol components, is limited in current machine learning studies. In this study, we propose a random forest model to retrieve high-precision aerosol properties and their absorption over land from Himawari-8 geostationary satellite images. Remarkably, this model attains a high degree of accuracy in estimating aerosol optical depth (AOD), absorption aerosol optical depth (AAOD), and single scattering albedo (SSA) of heavy air mass using only seven primary predictors (observational radiances or their mathematical combinations, geometries, and wavelength). For AOD, the new random forest model demonstrates excellent performance on an hourly scale (R2 ≥ 0.89, MAE < 0.07, RMSE <0.13), >80% of the samples fall within the expected error (EE) range. Concerning AAOD, the validation indicates that at least 65% of AAODs have a bias of ≤50%, with an R2 exceeding 0.78, MAE ≤ 0.008 and RMSE ≤0.016. SSA also demonstrates a high accuracy (R2 ≥ 0.57, MAE < 0.03, RMSE <0.05), with >70% of the results have an error ≤ 0.03. Through more comprehensive independent spatiotemporal cross validation, it can be determined that the model also offers reliable spatial and temporal predictions. The proposed RF model is capable of learning aerosol properties under most atmosphere scenarios, providing a reasonable conversion from predictors to AOD and AAOD/SSA under high aerosol loadings. The spatial patterns of these parameters suggest that the retrievals show considerable potential in capturing high aerosol loading in East Asia and biomass burning in Southeast Asia. The method introduced in this study offers a new approach to obtaining aerosol properties from geostationary satellite remote sensing, featuring a flexible process, simple inputs, high accuracy, and enhanced robustness. Additionally, it furnishes supplementary insights into aerosol absorption, presenting new possibilities in determining aerosol radiative impact and distinguishing aerosol components. • Much simpler input features for aerosol retrieval using the random forest method. • AAOD and SSA are accurately retrieved from Himawari using the RF model. • Improving AOD retrievals by optimizing the predictors in the RF model. • The retrievals show considerable potential in capturing aerosol absorption. [ABSTRACT FROM AUTHOR]

Wildfires significantly disturb ecosystems by altering forest structure, vegetation ecophysiology, and soil properties. Understanding the complex interactions between topographic and climatic conditions in post-wildfire recovery is crucial. This study investigates the interplay between topography, climate, burn severity, and years after fire on vegetation recovery across dominant land cover types (evergreen forest, shrubs, and grassland) in the Pacific Northwest region. Using Moderate Resolution Imaging Spectroradiometer (MODIS) data, we estimated vegetation recovery by calculating the incremental Enhanced Vegetation Index (EVI) change during post-fire years. A machine learning algorithm, random forest (RF), was employed to map relationships between the input features (elevation, slope, aspect, precipitation, temperature, burn severity, and years after fire) and the target (incremental EVI recovery) for each land cover type. Variable importance analysis and partial dependence plots were generated to understand the influence of individual features. The observed and predicted incremental EVI values showed good matches, with R 2 values of 0.99 for training and 0.89–0.945 for testing. The study found that climate variables, specifically precipitation and temperature, were the most important features overall, while elevation played the most significant role among the topographic factors. Partial dependence plots revealed that lower precipitation tended to cause a reduction in vegetation recovery for varying temperature ranges across land cover types. These findings can aid in developing targeted strategies for post-wildfire forest management, considering the varying responses of different land cover types to topographic, climatic, and burn severity factors. • Factors impacting post-fire vegetation recovery were analyzed with random forest. • Climate factors had the most influence in controlling the recovery process. • Topographic controls were mostly evident in evergreen forests. [ABSTRACT FROM AUTHOR]

Rain type classification into convective and stratiform is an essential step required to improve quantitative precipitation estimations by remote sensing instruments. Previous studies with Micro Rain Radar (MRR) measurements and subjective rules have been performed to classify rain events. However, automating this process by using machine learning (ML) models provides the advantages of fast and reliable classification with the possibility to classify rain minute by minute. A total of 20,979 min of rain data measured by an MRR at Das in northeast Spain were used to build seven types of ML models for stratiform and convective rain type classification. The proposed classification models use a set of 22 parameters that summarize the reflectivity, the Doppler velocity, and the spectral width (SW) above and below the so-called separation level (SL). This level is defined as the level with the highest increase in Doppler velocity and corresponds with the bright band in stratiform rain. A pre-classification of the rain type for each minute based on the rain microstructure provided by the collocated disdrometer was performed. Our results indicate that complex ML models, particularly tree-based ensembles such as xgboost and random forest which capture the interactions of different features, perform better than simpler models. Applying methods from the field of interpretable ML, we identified reflectivity at the lowest layer and the average spectral width in the layers below SL as the most important features. High reflectivity and low SW values indicate a higher probability of convective rain. [ABSTRACT FROM AUTHOR]

Satellite image data deliver consistent and frequent information for crop yield estimation over large areas. Hyperspectral narrowbands are more sensitive spectrally to changes in crop growth than multispectral broadbands but few studies quantified the gains in the former over the later. The PRecursore IperSpettrale della Missione Applicativa (PRISMA) mission offers narrow (≤10 nm) band capability across the full optical range. The multispectral broadband Sentinel-2 mission carries four experimental red-edge and near infrared (NIR) hyperspectral narrow (≤20 nm) bands. We compared the performance of PRISMA and Sentinel-2 spectral bands at important phases of crop development (vegetative, reproductive, maturity) in estimating field-level biomass and yield for corn, rice, soybean, and wheat. We selected three data-driven methods: two-band vegetation indices (TBVIs), partial least squares regression (PLSR), and random forest (RF). The PRISMA and Sentinel-2 models on average explained approximately 20% more variability in biomass and yield with RF than TBVIs and PLSR. The mean RMSE of the PRISMA RF models was 0.42 and 0.17 kg m−2, which was lower than the Sentinel-2 RF models (0.48 and 0.18 kg m−2). Multidate image (seasonal) model performance was generally higher than single-date image model performance. PRISMA shortwave infrared narrowbands and Sentinel-2 red-edge and near infrared bands were among the top-performing spectral regions. The results highlight potential complementarity between the PRISMA and Sentinel-2 missions for predicting crop biomass and yield. The results also show the benefits, limitations, and pitfalls of hyperspectral imaging in agricultural monitoring, which is important for upcoming operational hyperspectral missions, such as ESA CHIME and NASA Surface Biology and Geology (formerly HysPIRI). [ABSTRACT FROM AUTHOR]

Invasive species threaten natural ecosystems globally, displacing native species and causing biodiversity loss. In coastal areas with Mediterranean climate around the world, iceplant (Carpobrotus edulis) has become highly invasive, forming large monospecific zones that compete for resources with native plant species, including threatened or endangered species. Despite the widespread impact of iceplant across coastal areas with a Mediterranean climate, there is no precise information on where it is and how much it has spread. This study focuses on mapping and quantifying iceplant extent along the coast of Santa Barbara County in California, USA, by leveraging machine learning methods to identify iceplant in images from the 2020 National Agriculture Imagery Program (NAIP) archive at 0.6 m/pixel resolution, creating the most extensive assessment to date of this invasive species. Results include a map showing iceplant locations in 2020 with overall accuracy of 87.11% ± 2.45% (95% confidence interval). The estimated iceplant coverage in our region of study is 2.2 ± 0.42 km2 (95% confidence interval). Additionally, this study's use of open data and reproducible data analysis and validation workflow opens the door for the methods presented to be adapted and applied across California and all other Mediterranean climatic regions. In addition, the developed approach will accelerate monitoring over time to comprehend the spread and mitigation of iceplant invasions. • Random forest model produces high-accuracy map of invasive iceplant in southern CA. • Public NAIP data used to detect invasive plant species C. edulis at high-resolution. • Iceplant covered 2.2 km2 (±0.42 km2 95% conf.int.) of SB County's coast in 2020. • First-order texture and metadata as features increase model accuracy efficiently. • Integrated cloud computing and data access enable efficient, reproducible analyses. [ABSTRACT FROM AUTHOR]

Fire susceptibility modeling is crucial for sustaining and managing forests among many other valuable land resources. With 56% of its area covered by forests, Arkansas is known as the "natural state". About 1000 wildfires occurred and burned more than 10,000 acres each year during 1981–2018. In this paper, we use remote-sensing-based machine learning methods to address the natural and anthropogenic factors influencing wildfires and model fire susceptibility in Arkansas. Among the 15 explored variables, potential evapotranspiration, soil moisture, Palmer drought severity index, and dry season precipitation were recognized as the most significant factors contributing to the fire density. The obtained R-squared values are significant, with 0.99 for training the model and 0.92 for the validation. The results show that the Ouachita National Forest and the Ozark Forest, in west-central and west Arkansas, respectively, have the highest susceptibility to wildfires. The southern part of Arkansas has low-to-moderate fire susceptibility, while the eastern part of the state has the lowest fire susceptibility. These new results for Arkansas demonstrate the potency of remote-sensing-based random forest in predicting fire susceptibility at the state level that can be adapted to study fires in other states and help with fire preparedness to reduce loss and save the precious environment. [ABSTRACT FROM AUTHOR]

Copyright of Remote Sensing for Natural Resources is the property of Remote Sensing for Natural Resources Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The occurrence of drought is related to complicated interactions between many factors, such as precipitation, temperature, evapotranspiration and vegetation. In this study, the relationships between drought and precipitation, temperature, vegetation and evapotranspiration were investigated with a random forest (RF), and a new combined drought monitoring index (CDMI) was constructed. The effectiveness of the CDMI in monitoring drought in Shaanxi Province was verified by the in situ 1 ∼ 12-month standardized precipitation index (SPI); relative soil moisture (RSM) and four other commonly used remote sensing drought monitoring indices. The results show that CDMI is more correlated with the SPI and RSM than the four indices. Moreover, the spatial distributions of drought for the CDMI and RSM are similar. Therefore, the CDMI can be used to monitor droughts in Shaanxi Province, and machine learning can explore the relationships between various factors and establish a drought index without knowledge of the causal mechanisms of these factors. [ABSTRACT FROM AUTHOR]

Detection and monitoring invasive species can provide valuable ecological information to guide management decisions. Multispectral imagery remote sensing may be an ideal tool to address this problem by providing accurate and affordable repeat imagery. However, developing training datasets for remote sensing imagery can be riddled with issues such as matching a training site to a single pixel in target imagery or mixed pixels caused by low invader densities. The multitarget multiple-instance spectral match filter (MTMI-SMF) algorithm accounts for such errors and has low computation costs. Here, we evaluated MTMI-SMF for invasive species detection with National Agriculture Imagery Program (NAIP), RapidEye, and Landsat imagery using Brazilian peppertree (Schinus terebinthifolia) in the Everglades National Park as a case study. MTMI-SMF detected Brazilian peppertree with an area under the curve (AUC) of 0.85 for Landsat, 0.82 for RapidEye, and 0.71 NAIP imagery. Invader classification was developed using a threshold calculated from MTMI-SMF image detection confidence values and compared to a random forest classification. MTMI-SMF and random forest classifications had similar overall accuracies with NAIP imagery performing the lowest (an average of 73.6% and 88.9%, respectively) and Landsat performing the highest (an average of 90.9% and 91.4%, respectively). However, due to the relatively rare nature of Brazilian peppertree across this landscape, the overall classification accuracy was not fully representative of detection capabilities. MTMI-SMF outperformed random forest for user accuracy and had comparable producer accuarcy across all three image sources. Classifications primarily relied on the brighter signature of Brazilian peppertree due to the limited number of spectral bands in the multispectral imagery. As a result, confusion was caused by heterogeneous vegetation communities or shrub habitats, which most likely could be resolved with the increased availability of satellite hyperspectral datasets. Our study indicates that MTMI-SMF, a machine-learning approach that allows flexibility in training data, can detect highly problematic invaders using multispectral imagery. This pipeline could be readily adapted to other invaders and ecosystems, as well as different remote sensing image sources, and has low computational requirements. Our study demonstrates that remote sensing technologies and multiple-instance learning algorithms can provide managers with critical tools for tackling the ever-growing, costly, and challenging problem of tracking invasive species' plant spread. • Multitarget multiple-instance spectral match filter minimizes errors in data variability. • MTMI-SMF accurately detected an invader using NAIP, RapidEye, and Landsat imagery. • MTMI-SMF and random forest performed similarly but differed at low invader densities. • Landsat classification performed best due to increase in spectral discrimination. • MTMI-SMF handles imperfect, messy data with low computational requirements. [ABSTRACT FROM AUTHOR]

Biodiversity datasets with high spatial resolution are critical prerequisites for river protection and management decision-making. However, traditional morphological biomonitoring is inefficient and only provides several site estimates, and there is an urgent need for new approaches to predict biodiversity on fine spatial scales throughout the entire river systems. Here, we combined the environmental DNA (eDNA) and remote sensing (RS) technologies to develop a novel approach for predicting the spatial distribution of aquatic insects with high spatial resolution in a disturbed subtropical Dongjiang River system of southeast China. First, we screened thirteen RS-based vegetation indices that significantly correlated with the eDNA-inferred richness of aquatic insects. In particular, the green normalized difference vegetation index (GNDVI) and normalized difference red-edge2 (NDRE2) were closely related to eDNA-inferred richness. Second, using the gradient boosting decision tree, our data showed that the spatial pattern of eDNA-inferred richness could achieve a high spatial resolution to 500 m reach and accurate prediction of more than 80%, and the prediction efficiency of the headwater streams (Strahler stream order = 1) was slightly higher than the downstream (Strahler stream order >1). Third, using the random forest algorithm, the spatial distribution of aquatic insects could reach a prediction rate of over 70% for the presence or absence of specific genera. Overall, this study provides a new approach to achieving high spatial resolution prediction of the distribution of aquatic insects, which supports decision-making on river diversity protection under climate changes and human impacts. [Display omitted] • EDNA and remote sensing datasets are combined to predict species distribution. • Thirteen RS-based vegetation indices are closely related to eDNA-inferred richness. • Aquatic insects can achieve spatial distribution prediction within 500 m reach. [ABSTRACT FROM AUTHOR]

Machine learning algorithms such as Random Forest (RF) are being increasingly applied on traditionally geographical topics such as population estimation. Even though RF is a well performing and generalizable algorithm, the vast majority of its implementations is still 'aspatial' and may not address spatial heterogenous processes. At the same time, remote sensing (RS) data which are commonly used to model population can be highly spatially heterogeneous. From this scope, we present a novel geographical implementation of RF, named Geographical Random Forest (GRF) as both a predictive and exploratory tool to model population as a function of RS covariates. GRF is a disaggregation of RF into geographical space in the form of local sub-models. From the first empirical results, we conclude that GRF can be more predictive when an appropriate spatial scale is selected to model the data, with reduced residual autocorrelation and lower Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) values. Finally, and of equal importance, GRF can be used as an effective exploratory tool to visualize the relationship between dependent and independent variables, highlighting interesting local variations and allowing for a better understanding of the processes that may be causing the observed spatial heterogeneity. [ABSTRACT FROM AUTHOR]

Forests are a global environmental priority that need to be monitored frequently and at large scales. Satellite images are a proven useful, free data source for regular global forest monitoring but these images often have missing data in tropical regions due to climate driven persistent cloud cover. Remote sensing and statistical approaches to filling these missing data gaps exist and these can be highly accurate, but any interpolation method results are uncertain and these methods do not provide measures of this uncertainty. We present a new two-step spatial stochastic random forest (SS-RF) method that uses random forest algorithms to construct Beta distributions for interpolating missing data. This method has comparable performance with the traditional remote sensing compositing method, and additionally provides a probability for each interpolated data point. Our results show that the SS-RF method can accurately interpolate missing data and quantify uncertainty and its applicability to the challenge of monitoring forest using free and incomplete satellite imagery data. We propose that there is scope for our SS-RF method to be applied to other big data problems where a measurement of uncertainty is needed in addition to estimates. [ABSTRACT FROM AUTHOR]

Hulun Lake is facing significant water quality degradation, necessitating effective monitoring for safety. Traditional methods lack the necessary spatial and temporal coverage, underscoring the need for a remote sensing model. In this study, we utilized the Landsat 8 OLI dataset, incorporating cross-section monitoring and field sampling data comprehensively. Employing the random forest algorithm, we constructed a remote sensing inversion model for six water quality parameters in Hulun Lake: chlorophyll-a (Chl-a), total nitrogen (TN), total phosphorus (TP), ammonia nitrogen (NH 3 −N), chemical oxygen demand (COD), and dissolved oxygen (DO). The model was applied to the non-freezing period of Hulun Lake from 2016 to 2021, exhibiting commendable performance and generating high-resolution maps. Time series analysis revealed that during the study period, the pollution levels of TN, TP, and COD in Hulun Lake were extremely serious, exceeding the Class V water standard of China's surface water environmental quality standard. Regional analysis indicated lower pollutant concentrations in the central lake area compared to the lake inlet. The inflowing rivers with high pollution adversely impacted Hulun Lake's water quality. To ensure the continued health of Hulun Lake's water quality, it is imperative to monitor lake water quality attentively and implement necessary measures to prevent further deterioration. This study holds crucial importance for shaping and executing ecological protection and restoration strategies for Hulun Lake. • High accuracy and stability achieved in the inversion of water quality using RF. • Temporal-spatial analysis revealed the distribution and variation characteristics. • Valuable references provided for water resource monitoring and management. [ABSTRACT FROM AUTHOR]

Riparian vegetation composition and channel morphology are susceptible to long-term alterations caused by external stressors, including climate-change-induced droughts and engineered infrastructures. The objectives of this study were to (1) quantify trends in riparian vegetation and channel/floodplain morphology over large spatial (∼290 km) and temporal scales (∼30 years) and (2) investigate the relationships between hydroclimatic drivers and changes in riparian vegetation and channel morphology. We implemented a random forest classifier via a machine learning technique in Google Earth Engine. The study area was a 290 km reach of the Rio Grande located in New Mexico, USA. We used the combination of remotely sensed data and products (e.g., Landsat imagery, Normalized Difference Vegetation Index (NDVI), and land cover) to characterize vegetation, vegetation cover changes, and river morphology shifts from 1984 to 2020. The trend analysis revealed increased vegetated areas and NDVI (0.0004/yr) during long-term drought. The channel experienced a reduction in width associated with vegetation encroachment and the formation of stable vegetated islands. The streamflow hydrograph characteristics were positively correlated with vegetation cover and channel morphology. Our study contributes novel insights into the long-term riparian ecosystem dynamics under drought stress, informing drought impact mitigation and ecosystem management in arid and semi-arid regions. • Changes in riparian vegetation, channel morphology, and hydrology have cascading impacts on the river's ecological function. • The vegetated area has increased, and the channel has narrowed along the Rio Grande during persistent drought. • Maintenance of channel width is challenging once vegetation is established. [ABSTRACT FROM AUTHOR]