Abstract:International audienceSatellite-borne Synthetic Aperture Radar (SAR) has been used for characterizing and mapping in two relevant ice-free areas in the South Shetland Islands. The objective has been to identify and characterize land surface covers that mainly include periglacial and glacial landforms, using fully polarimetric SAR C band RADARSAT-2 data, on Fildes Peninsula that forms part of King George Island, and Ardley Island. Polarimetric parameters obtained from the SAR data, a selection of field based tr… Show more
“…The low uncertainty of the SVM, in turn, was probably associated with the efficient optimization of its hyperparameters in caret. To the best of our knowledge, only SVM among the ML models tested in this work had been used for classifying geomorphology in the Antarctic Peninsula region (Schmid et al, 2012, 2017). Applying the SVM in Vega Island, we obtained a considerably better accuracy than the above authors, who used this algorithm to classify geomorphology in islands of Maritime Antarctica.…”
Section: Discussionmentioning
confidence: 99%
“…Such landforms encompass patterned ground (Luoto & Hjort, 2005), rock glaciers (Brenning et al, 2007), loess (Zhao et al, 2017), glacial deposits (Zhang et al, 2019), floodplains (Woznicki et al, 2019), alluvial fans (Pipaud & Lehmkuhl, 2017), bedrock outcrops (Scarpone et al, 2017), seamounts (Lawson et al, 2017) and even Martian craters (Urbach & Stepinski, 2009). However, in Antarctica, digital geomorphological mappings are very scarce, with only a few records of ML approaches in the Antarctic Peninsula and surrounding islands (Schmid et al, 2012(Schmid et al, , 2017.…”
The detailed geomorphology of ice‐free landscapes of Antarctica is key to understanding how their highly fragile environments respond to climate change, at different temporal and spatial scales. Despite the recent advances in geomorphological studies of ice‐free areas, machine learning applications to produce landform maps are still scarce on the Antarctic continent. In this study, we evaluated the predictive performance of different supervised machine learning algorithms to produce digital geomorphological maps in Vega Island—Antarctic Peninsula region. We tested six different models: average artificial neural networks, C5.0 decision tree, random forest, support vector machine, supervised self‐organizing map and weighted k‐nearest neighbours. We used an initial set of 54 geomorphometric and spectral predictors, from which redundant variables with Pearson correlation coefficient >|0.95| were removed, and only the most important predictors for each model were selected using recursive feature elimination. For training, we ran each model 100 times and predictions were assessed by the kappa and global accuracy values. The best predictors were the Red Edge 6 and SWIR 11 bands, roughness concentration index, elevation and drainage density. The decision trees C5.0 and random forest had the best performance, with average validation kappa of 0.85 ± 0.03 and 0.84 ± 0.03, respectively, evidencing excellent prediction. Despite the similar performance, random forest showed greater uncertainty degree and accuracy when classifying complex landforms, attesting to its great robustness. From sensitivity and specificity values, we observed that the glaciers and talus showed higher accuracy, whereas cryoplanated platforms and scree slopes had the worst classification. The presented methodology optimized the classification by selecting the most important predictors, assessing accuracy and evaluating uncertainty. The results indicated that machine learning methods have great potential to produce geomorphological mappings in the Antarctic ice‐free areas, as a promising tool to provide detailed information on remote, harsh polar environments.
“…The low uncertainty of the SVM, in turn, was probably associated with the efficient optimization of its hyperparameters in caret. To the best of our knowledge, only SVM among the ML models tested in this work had been used for classifying geomorphology in the Antarctic Peninsula region (Schmid et al, 2012, 2017). Applying the SVM in Vega Island, we obtained a considerably better accuracy than the above authors, who used this algorithm to classify geomorphology in islands of Maritime Antarctica.…”
Section: Discussionmentioning
confidence: 99%
“…Such landforms encompass patterned ground (Luoto & Hjort, 2005), rock glaciers (Brenning et al, 2007), loess (Zhao et al, 2017), glacial deposits (Zhang et al, 2019), floodplains (Woznicki et al, 2019), alluvial fans (Pipaud & Lehmkuhl, 2017), bedrock outcrops (Scarpone et al, 2017), seamounts (Lawson et al, 2017) and even Martian craters (Urbach & Stepinski, 2009). However, in Antarctica, digital geomorphological mappings are very scarce, with only a few records of ML approaches in the Antarctic Peninsula and surrounding islands (Schmid et al, 2012(Schmid et al, , 2017.…”
The detailed geomorphology of ice‐free landscapes of Antarctica is key to understanding how their highly fragile environments respond to climate change, at different temporal and spatial scales. Despite the recent advances in geomorphological studies of ice‐free areas, machine learning applications to produce landform maps are still scarce on the Antarctic continent. In this study, we evaluated the predictive performance of different supervised machine learning algorithms to produce digital geomorphological maps in Vega Island—Antarctic Peninsula region. We tested six different models: average artificial neural networks, C5.0 decision tree, random forest, support vector machine, supervised self‐organizing map and weighted k‐nearest neighbours. We used an initial set of 54 geomorphometric and spectral predictors, from which redundant variables with Pearson correlation coefficient >|0.95| were removed, and only the most important predictors for each model were selected using recursive feature elimination. For training, we ran each model 100 times and predictions were assessed by the kappa and global accuracy values. The best predictors were the Red Edge 6 and SWIR 11 bands, roughness concentration index, elevation and drainage density. The decision trees C5.0 and random forest had the best performance, with average validation kappa of 0.85 ± 0.03 and 0.84 ± 0.03, respectively, evidencing excellent prediction. Despite the similar performance, random forest showed greater uncertainty degree and accuracy when classifying complex landforms, attesting to its great robustness. From sensitivity and specificity values, we observed that the glaciers and talus showed higher accuracy, whereas cryoplanated platforms and scree slopes had the worst classification. The presented methodology optimized the classification by selecting the most important predictors, assessing accuracy and evaluating uncertainty. The results indicated that machine learning methods have great potential to produce geomorphological mappings in the Antarctic ice‐free areas, as a promising tool to provide detailed information on remote, harsh polar environments.
“…Sampling was J o u r n a l P r e -p r o o f carried out within areas representing different geomorphological features and surface covers. On DI and FP sampling was constrained to areas where soil forming processes was minimal such as on weathered bedrock with vegetation cover and fine volcanic deposits where bare soil and sparse vegetation was present (Schmid et al, 2017). In contrast, at CP, where there is an abundance of vegetation, samples were taken in areas that had a high vegetation cover and soil forming processes were more advanced.…”
Section: Description Of the Study Sites Soils And Sampling Point Loca...mentioning
“…Remote sensing techniques offer a great potential to identify relief and landscape features, as well as detect changes in areas where access is difficult and little or no data are available [3]. Synthetic Aperture Radar (SAR) sensors are useful for determining physical properties of terrestrial land covers such as structure and roughness [4]. Furthermore, in environments encountered in the Antarctic Peninsula region, using SAR data is of advantage because microwaves function in all-weather conditions.…”
Section: Introductionmentioning
confidence: 99%
“…Considering the sensor characteristics and spatial resolution, preliminary satisfactory results were obtained to map complex geomorphological features such as periglacial landforms. This was initially tested for the area of Fildes Peninsula where satisfactory results were obtained as shown in previous work [4]. However, for the wider area the ground information and other ancillary sources are considered limited, at this moment, to carry out a robust validation.…”
Ice-free areas within the Northern Antarctic Peninsula region are of interest for studying changes occurring to surface covers, including those related to glacial coverage, raised beach deposits and periglacial processes and permafrost. The objective of this work is to map the main surface covers within ice-free areas of King George Island, the largest island of the South Shetlands archipelago, using fully polarimetric RADARSAT-2 SAR data. Surface covers such as rock outcrops and glacial till, stone fields, patterned ground, and sand and gravel deposits form the most representative classes and account for 84 km 2 of the ice-free areas on the island. A distribution of complex geomorphological features and landforms was obtained, being some of them considered indicators of periglacial processes and presence of permafrost.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.