Search

Machine learning can offer an efficient method to identify and map caves, sinkholes, and other cave-like features (i.e. sinkholes, rockshelters, voids) using remotely sensed imagery. While there exists a body of work applying machine learning for sinkhole identification, little work exists for caves. In the densely forested and rugged Maya Lowlands, developing such a methodology can help archaeologists to identify previously unknown caves that may contain important archaeological materials. Here, we introduce a proof-of-concept project that uses random forest and lidar-derived landscape morphometrics to map caves and other cave-like features in northwest Belize. Several undocumented caves and cave-like features were identified in our study area based on model results. Next steps towards making a more robust version of this model include the addition of more training data and integration of a larger number of morphologic parameters. Based on the results described here as well as those in cited works focused on caves, we proposed machine learning as a first step in cave and cave-like feature identification, followed then by fieldwork and ground-truthing. [ABSTRACT FROM AUTHOR]

Precise estimation of Sea Surface Salinity (SSS) is of prime importance to know marine physical and biochemical processes. In this manuscript we have utilized the Random Forest (RF) and Support Vector Regression (SVR) to estimate the SSS concentration. Particularly, we have utilized the global SSS values provided by the Copernicus Marine Services. We have also extracted the Aqua MODIS and Sentinel-3 data from the Google Earth Engine platform. The complete range of SSS varies approximately from 6 psu to 38 psu which enables the global application of the dataset. We have also studied the importance of different wavelength bands and their significance to SSS using the RF model. It is found that wavelengths around 400 nm have their highest significance due to the sensitiveness to Chromophoric Dissolved Organic Matter (CDOM) and other related constituents. For the regression analysis we have also utilized the Laplacian kernel in the SVR. It is found that the SVR with Laplace kernel outperforms the RF and SVR with Radial Basis Function (RBF) kernel. The trained algorithms are used to estimate the SSS content over the Bay of Bengal and the Arabian Sea regions which show interesting variations in SSS content due to diverse climatological conditions. [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]

Datasets containing recorded landslide and non-landslide samples can greatly influence the performance of machine learning (ML) models, which are commonly used in landslide susceptibility mapping (LSM). However, the non-landslide samples cannot be directly obtained. In this study, a pattern-based approach is proposed to improve the LSM process, constructing unsupervised machine learning (UML) – supervised machine learning (SML) collaborative models in which the non-landslide samples can be reasonably selected. Two UML models, the Gaussian mixture model (GMM) and K-means, are introduced to sample the non-landslide datasets with four sampling selections (abbreviated as A, B, C and D, respectively). Then non-landslide patterns recognised by the UML models are learned by the random forest (RF). A new sensitivity index, accuracy improvement ratio (AIR), is defined to evaluate the superiority of these sampling selections. Compared with the GMM-RF model, the K-means-RF model is more capable of recognising non-landslide patterns and providing sufficient and reliable non-landslide samples. The sampling selection A of the K-means-RF with an AIR value of 2.3 is regarded as the best selection. The results indicate that the UML-SML model based on the pattern-based approach offers an effective strategy to find the non-landslide samples and has a better solution to the LSM. [ABSTRACT FROM AUTHOR]

The rapid expansion of exotic saltcedar along riparian corridors has dramatically altered the landscape structure and ecological function of riparian habitats in the western United States. The development of accurate and reproducible mapping methods with remote sensing plays an indispensable role in the timely monitoring of saltcedar, re-evaluating its ecological functions, and establishing effective control measures. The utmost challenge for achieving this goal is manifested as the lack of time series of remote sensing images to capture the saltcedar phenology adequately. To this end, the newly available Landsat 9 images, combined with its counterpart of Landsat 8, offer a precious opportunity to compensate for the temporal image shortage. To understand Landsat 9 in the saltcedar classification and to discover helpful information for its application, this study presents the first attempt to classify saltcedar using intra-annual Landsat 8 and Landsat 9 images. We adopted two machine learning algorithms, support vector machine (SVM) and random forest (RF), to compare the performance of Landsat 9 and Landsat 8 for intra-annual saltcedar classification. In addition, we investigated the respective contribution of each spectral band to the overall performance and identified the optimal time window for saltcedar classification. The results indicated that the difference in classification performance between Landsat 9 and Landsat 8 was insignificant. The shortwave infrared bands associated with both Landsat 8 & 9 have contributed most to the process of saltcedar identification. Image acquired in July, November, and December yielded better results than other months for saltcedar classification. It is concluded that Landsat 8 & 9 constellation has the potential to refine saltcedar classification accuracy on larger spatial and temporal scales. [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]

This study aims to predict channel unit types (CUTs) by combining remotely sensed data with morphological variables using machine learning algorithms (random forest, support vector machines, multiple adaptive regression splines, extreme gradient boosting and adaptive boosting) within the Upper Ogun River Basin, Southwestern Nigeria. In achieving the aim of this study, we identified the most important variable(s) in CUT discrimination using the random forest – recursive feature elimination (RF-RFE). A total of 249 cross-sections across 83 reaches were sampled during the fieldwork. Landsat 8 and Sentinel-1 bands were retrieved for days the fieldwork was carried and mosaiced using the Google Earth Engine platform. The RF-RFE identified five top variables (accuracy: 0.79 ± 0.14; kappa: 0.39) discriminating the CUT as dimensionless stream power, slope, width, wetted perimeter and Band 4. In essence, there is much hope in the use of remote sensing in CUT mapping at the reach scale. [ABSTRACT FROM AUTHOR]

One of the main applications of machine learning (ML) in remote sensing (RS) is the pixel-level classification of satellite images into land cover types. Although classes with different spectral signatures can be easily separated, e.g. aquatic and terrestrial land cover types, others have similar spectral signatures and are hard to separate using only the information within a single pixel. This work focused on the separation of two cover types with similar spectral signatures, cocoa agroforest and forest, over an area in Pará, Brazil. For this, we study the training and application of several ML algorithms on datasets obtained from a single composite image, a time-series (TS) composite obtained from the same location and by preprocessing the TS composite using simple TS preprocessing techniques. As expected, when ML algorithms are applied to a dataset obtained from a composite image, the median producer's accuracy (PA) and user's accuracy (UA) in those two classes are significantly lower than the median overall accuracy (OA) for all classes. The second dataset allows the ML models to learn the evolution of the spectral signatures over 5 months. Compared to the first dataset, the results indicate that ML models generalize better using TS data, even if the series are short and without any preprocessing. This generalization is further improved in the last dataset. The ML models are subsequently applied to an area with different geographical bounds. These last results indicate that, out of seven classifiers, the popular random forest (RF) algorithm ranked fourth, while XGBoost (×GB) obtained the best results. The best OA, as well as the best PA/UA balance, were obtained by performing feature construction using the M3GP algorithm and then applying XGB to the new extended dataset. [ABSTRACT FROM AUTHOR]

Snow and cloud discrimination is a main factor contributing to errors in satellite-based snow cover. To address the error, satellite-based snow cover performs snow reclassification tests on the cloud pixels of the cloud mask, but the error still remains. Machine Learning (ML) has recently been applied to remote sensing to calculate satellite-based meteorological data, and its utility has been demonstrated. In this study, snow and cloud discrimination errors were analyzed for GK-2A/AMI snow cover, and ML models (Random Forest and Deep Neural Network) were applied to accurately distinguish snow and clouds. The ML-based snow reclassified was integrated with the GK-2A/AMI snow cover through post-processing. We used the S-NPP/VIIRS snow cover and ASOS in situ snow observation data, which are satellite-based snow cover and ground truth data, as validation data to evaluate whether the snow/cloud discrimination is improved. The ML-based integrated snow cover detected 33–53% more snow compared to the GK-2A/AMI snow cover. In terms of performance, the F1-score and overall accuracy of the GK-2A/AMI snow cover was 73.06% and 89.99%, respectively, and those of the integrated snow cover were 76.78–78.28% and 90.93–91.26%, respectively. [ABSTRACT FROM AUTHOR]