Machine Learning Classification of Mediterranean Forest Habitats in Google Earth Engine Based on Seasonal Sentinel-2 Time-Series and Input Image Composition Optimisation

Remote Sensing, Journal Year: 2021, Volume and Issue: 13(4), P. 586 - 586

Published: Feb. 7, 2021

The sustainable management of natural heritage is presently considered a global strategic issue. Owing to the ever-growing availability free data and software, remote sensing (RS) techniques have been primarily used map, analyse, monitor resources for conservation purposes. need adopt multi-scale multi-temporal approaches detect different phenological aspects vegetation types species has also emerged. time-series composite image approach allows capturing much spectral variability, but presents some criticalities (e.g., time-consuming research, downloading data, required storage space). To overcome these issues, Google Earth engine (GEE) proposed, cloud-based computational platform that users access process remotely sensed at petabyte scales. application was tested in protected area Calabria (South Italy), which particularly representative Mediterranean mountain forest environment. In random (RF), support vector machine (SVM), classification regression tree (CART) algorithms were perform supervised pixel-based based on use Sentinel-2 images. A select best input (seasonal composition strategies, statistical operators, band composition, derived indices (VIs) information) implemented. set accuracy indicators, including overall (OA) multi-class F-score (Fm), computed assess results classifications. GEE proved be reliable powerful tool process. (OA = 0.88 Fm 0.88) achieved using RF with summer composite, adding three VIs (NDVI, EVI, NBR) bands. SVM produced OAs 0.83 0.80, respectively.

Language: Английский

---

Urban Land Use and Land Cover Change Analysis Using Random Forest Classification of Landsat Time Series

Remote Sensing, Journal Year: 2022, Volume and Issue: 14(11), P. 2654 - 2654

Published: June 1, 2022

Efficient implementation of remote sensing image classification can facilitate the extraction spatiotemporal information for land use and cover (LULC) classification. Mapping LULC change pave way to investigate impacts different socioeconomic environmental factors on Earth’s surface. This study presents an algorithm that uses Landsat time-series data analyze change. We applied Random Forest (RF) classifier, a robust method, in Google Earth Engine (GEE) using imagery from 5, 7, 8 as inputs 1985 2019 period. also explored performance pan-sharpening bands besides impact compositions produce high-quality map. used statistical increase multispectral bands’ (Landsat 7–9) spatial resolution 30 m 15 m. In addition, we checked based several spectral indices other auxiliary such digital elevation model (DEM) surface temperature (LST) final accuracy accuracy. compared result our proposed method Copernicus Global Land Cover Layers (CGLCL) map verify algorithm. The results show that: (1) Using pan-sharpened top-of-atmosphere (TOA) products more accurate instead reflectance (SR) alone; (2) LST DEM are essential features classification, them accuracy; (3) produced higher (94.438% overall (OA), 0.93 Kappa, F1-score) than CGLCL (84.4% OA, 0.79 0.50 2019; (4) total agreement between test exceeds 90% (93.37–97.6%), 0.9 (0.91–0.96), 0.85 (0.86–0.95) Kappa values, F1-score, respectively, which is acceptable both Moreover, provide code repository allows classifying 4, within GEE. be quickly easily regions interest mapping.

Language: Английский

---

Land Use and Land Cover Mapping Using Sentinel-2, Landsat-8 Satellite Images, and Google Earth Engine: A Comparison of Two Composition Methods

Remote Sensing, Journal Year: 2022, Volume and Issue: 14(9), P. 1977 - 1977

Published: April 20, 2022

Accurate and real-time land use/land cover (LULC) maps are important to provide precise information for dynamic monitoring, planning, management of the Earth. With advent cloud computing platforms, time series feature extraction techniques, machine learning classifiers, new opportunities arising in more accurate large-scale LULC mapping. In this study, we aimed at finding out how two composition methods spectral–temporal metrics extracted from satellite can affect ability a classifier produce maps. We used Google Earth Engine (GEE) platform create cloud-free Sentinel-2 (S-2) Landsat-8 (L-8) over Tehran Province (Iran) as 2020. Two methods, namely, seasonal composites percentiles metrics, were define four datasets based on series, vegetation indices, topographic layers. The random forest was classification identifying most variables. Accuracy assessment results showed that S-2 outperformed L-8 overall class level. Moreover, comparison indicated percentile both series. At level, improved performance related their better about phenological variation different classes. Finally, conclude methodology GEE an fast way be

Language: Английский

---

Analysis of Land Use and Land Cover Using Machine Learning Algorithms on Google Earth Engine for Munneru River Basin, India

Sustainability, Journal Year: 2021, Volume and Issue: 13(24), P. 13758 - 13758

Published: Dec. 13, 2021

The growing human population accelerates alterations in land use and cover (LULC) over time, putting tremendous strain on natural resources. Monitoring assessing LULC change large areas is critical a variety of fields, including resource management climate research. has emerged as concern for policymakers environmentalists. As the need reliable estimation maps from remote sensing data grows, it to comprehend how different machine learning classifiers perform. primary goal present study was classify Google Earth Engine platform using three algorithms—namely, support vector (SVM), random forest (RF), classification regression trees (CART)—and compare their performance accuracy assessments. area classified via supervised classification. For improved accuracy, NDVI (normalized difference vegetation index) NDWI water indices were also derived included. years 2016, 2018, 2020, multitemporal Sentinel-2 Landsat-8 with spatial resolutions 10 m 30 used ‘Water bodies’, ‘forest’, ‘barren land’, ‘vegetation’, ‘built-up’ major classes. average overall SVM, RF, CART images 90.88%, 94.85%, 82.88%, respectively, 93.8%, 95.8%, 86.4% images. These results indicate that RF outperform both SVM terms accuracy.

Language: Английский

---

Comparison of Land Use Land Cover Classifiers Using Different Satellite Imagery and Machine Learning Techniques

Remote Sensing, Journal Year: 2022, Volume and Issue: 14(19), P. 4978 - 4978

Published: Oct. 6, 2022

Accurate land use cover (LULC) classification is vital for the sustainable management of natural resources and to learn how landscape changing due climate. For accurate efficient LULC classification, high-quality datasets robust methods are required. With increasing availability satellite data, geospatial analysis tools, methods, it essential systematically assess performance different combinations data help select best approach classification. Therefore, this study aims evaluate two commonly used platforms (i.e., ArcGIS Pro Google Earth Engine) with Landsat, Sentinel, Planet) through a case city Charlottetown in Canada. Specifically, three classifiers Pro, including support vector machine (SVM), maximum likelihood (ML), random forest/random tree (RF/RT), utilized develop maps over period 2017–2021. Whereas four Engine, SVM, RF/RT, minimum distance (MD), regression (CART), same period. To identify most classifier, overall accuracy kappa coefficient each classifier calculated throughout all platforms, methods. Change detection then conducted using quantify changes Results show that SVM both Engine presents compared other classifiers. In particular, shows an 89% 91% 94% Planet. Similarly, 87% Landsat 8 92% Sentinel 2. Furthermore, change results 13.80% 14.10% forest areas have been turned into bare urban class, respectively, 3.90% has converted area from 2017 2021, suggesting intensive urbanization. The will provide scientific basis selecting remote sensing imagery maps.

Language: Английский

---

Comparison of Three Machine Learning Algorithms Using Google Earth Engine for Land Use Land Cover Classification

Rangeland Ecology & Management, Journal Year: 2023, Volume and Issue: 92, P. 129 - 137

Published: Nov. 22, 2023

Language: Английский

---

Machine learning algorithm based prediction of land use land cover and land surface temperature changes to characterize the surface urban heat island phenomena over Ahmedabad city, India

Urban Climate, Journal Year: 2022, Volume and Issue: 42, P. 101116 - 101116

Published: Feb. 9, 2022

Language: Английский

---

Vegetation Mapping with Random Forest Using Sentinel 2 and GLCM Texture Feature—A Case Study for Lousã Region, Portugal

Remote Sensing, Journal Year: 2022, Volume and Issue: 14(18), P. 4585 - 4585

Published: Sept. 14, 2022

Vegetation mapping requires accurate information to allow its use in applications such as sustainable forest management against the effects of climate change and threat wildfires. Remote sensing provides a powerful resource fundamental data at different spatial resolutions spectral regions, making it an essential tool for vegetation biomass management. Due ever-increasing availability free software, satellites have been predominantly used map, analyze, monitor natural resources conservation purposes. This study aimed map from Sentinel-2 (S2) complex mixed cover Lousã district Portugal. We ten multispectral bands with resolution 10 m, four indices, including Normalized Difference Index (NDVI), Green (GNDVI), Enhanced (EVI), Soil Adjusted (SAVI). After applying principal component analysis (PCA) on S2A bands, texture features, mean (ME), homogeneity (HO), correlation (CO), entropy (EN), were derived first three components. Textures obtained using Gray-Level Co-Occurrence Matrix (GLCM). As result, 26 independent variables extracted S2. defining land classes object-based approach, Random Forest (RF) classifier was applied. The accuracy evaluated by confusion matrix, metrics overall (OA), producer (PA), user (UA), kappa coefficient (Kappa). described classification methodology showed high OA 90.5% 89% mapping. Using GLCM features indices increased up 2%; however, achieved highest (92%), indicating features′ capability detecting variability species stand level. ME CO contribution among textures. GNDVI outperformed other variable importance. Moreover, only especially 11, 12, 2, potential classify 88%. that adding least one feature index into may effectively increase tree discrimination.

Language: Английский

---

A Review of Practical AI for Remote Sensing in Earth Sciences

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(16), P. 4112 - 4112

Published: Aug. 21, 2023

Integrating Artificial Intelligence (AI) techniques with remote sensing holds great potential for revolutionizing data analysis and applications in many domains of Earth sciences. This review paper synthesizes the existing literature on AI sensing, consolidating analyzing methodologies, outcomes, limitations. The primary objectives are to identify research gaps, assess effectiveness approaches practice, highlight emerging trends challenges. We explore diverse including image classification, land cover mapping, object detection, change hyperspectral radar analysis, fusion. present an overview technologies, methods employed, relevant use cases. further challenges associated practical such as quality availability, model uncertainty interpretability, integration domain expertise well solutions, advancements, future directions. provide a comprehensive researchers, practitioners, decision makers, informing at exciting intersection sensing.

Language: Английский

---

Prediction of Future Land Use/Land Cover Changes Using a Coupled CA-ANN Model in the Upper Omo–Gibe River Basin, Ethiopia

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(4), P. 1148 - 1148

Published: Feb. 20, 2023

Land use/land cover change evaluation and prediction using spatiotemporal data are crucial for environmental monitoring better planning management of land use. The main objective this study is to evaluate changes the time period 1991–2022 predict future CA-ANN model in Upper Omo–Gibe River basin. Landsat-5 TM 1991, 1997, 2004, Landsat-7 ETM+ 2010, Landsat-8 (OLI) 2016 2022 were downloaded from USGS Earth Explorer Data Center. A random forest machine learning algorithm was employed LULC classification. classification result evaluated an accuracy assessment technique assure correctness method employing kappa coefficient. Kappa coefficient values indicate that there strong agreement between classified reference data. Using MOLUSCE plugin QGIS model, predicted. Artificial neural network (ANN) cellular automata (CA) methods made available modeling via plugin. Transition potential computed, predicted model. An overall 86.53% value 0.82 obtained by comparing actual with simulated same year. findings revealed 2037, agricultural (63.09%) shrubland (5.74%) showed significant increases, (−48.10%) grassland (−0.31%) decreased. From 2037 2052, built-up area (2.99%) a increase, (−2.55%) decrease. 2052 2067, projected simulation (3.15%) (0.32%) increased, (−1.59%) (−0.56%) decreases. According study’s findings, drivers expansion areas land, which calls thorough investigation additional models give planners policymakers clear information on their effects.

Language: Английский

---