Rapid lithological mapping using multi-source remote sensing data fusion and automatic sample generation strategy

Assessing the performance of machine learning algorithms for soil salinity mapping in Google Earth Engine platform using Sentinel-2A and Landsat-8 OLI data

Nitrogen plays an important role in improving soil productivity and maintaining ecosystem stability. Mapping and monitoring the soil total nitrogen (STN) content is the basis for modern soil management. The Google Earth Engine (GEE) platform covers a wide range of available satellite remote sensing datasets and can process massive data calculations. We collected 6823 soil samples in Shandong Province, China. The random forest (RF) algorithm predicted the STN content in croplands from 2002 to 2016 in Shandong Province, China on the GEE platform. Our results showed that RF had the coefficient of determination (R2) (0.57), which can predict the spatial distribution of the STN and analyze the trend of STN changes. The remote sensing spectral reflectance is more important in model building according to the variable importance analysis. From 2002 to 2016, the STN content of cropland in the province had an upward trend of 35.6%, which increased before 2010 and then decreased slightly. The GEE platform provides an opportunity to map dynamic changes of the STN content effectively, which can be used to evaluate soil properties in the future long-term agricultural management.

The ecological quality of inland areas is an important aspect of the United Nations Sustainable Development Goals (UN SDGs). The ecological environment of Northwest China is vulnerable to changes in climate and land use/land cover, and the changes in ecological quality in this arid region over the last two decades are not well understood. This makes it more difficult to advance the UN SDGs and develop appropriate measures at the regional level. In this study, we used the Moderate Resolution Imaging Spectroradiometer (MODIS) products to generate remote sensing ecological index (RSEI) on the Google Earth Engine (GEE) platform to examine the relationship between ecological quality and environment in Xinjiang during the last two decades (from 2000 to 2020). We analyzed a 21-year time series of the trends and spatial characteristics of ecological quality. We further assessed the importance of different environmental factors affecting ecological quality through the random forest algorithm using data from statistical yearbooks and land use products. Our results show that the RSEI constructed using the GEE platform can accurately reflect the ecological quality information in Xinjiang because the contribution of the first principal component was higher than 90.00%. The ecological quality in Xinjiang has increased significantly over the last two decades, with the northern part of this region having a better ecological quality than the southern part. The areas with slightly improved ecological quality accounted for 31.26% of the total land area of Xinjiang, whereas only 3.55% of the land area was classified as having a slightly worsen (3.16%) or worsen (0.39%) ecological quality. The vast majority of the deterioration in ecological quality mainly occurred in the barren areas Temperature, precipitation, closed shrublands, grasslands and savannas were the top five environmental factors affecting the changes in RSEI. Environmental factors were allocated different weights for different RSEI categories. In general, the recovery of ecological quality in Xinjiang has been controlled by climate and land use/land cover during the last two decades and policy-driven ecological restoration is therefore crucial. Rapid monitoring of inland ecological quality using the GEE platform is projected to aid in the advancement of the comprehensive assessment of the UN SDGs.

Floods are one of the most devastating natural hazards, which lead to fatal outcomes to socio-economic and human life for several weeks in vast areas. Floods mostly occur in the monsoon season when cloud-free optical data are hardly available; whereas, microwave data has advantages to penetrate the cloud and enables real-time flood mapping. The current study focuses using Sentinel-1 synthetic aperture radar (SAR) images for flood or water inundation mapping in Assam during July – September in 2018, 2019 and 2020. Otsu’s automatic thresholding method was applied on the vertically transmitted vertically received (VV) polarization band Sentinel-1 SAR data to identify the water inundated areas. The Google Earth Engine (GEE) platform was used for data processing, which requires lesser data processing capabilities in the user system and less time than other traditional approaches. The comparison with the Sentinel-2 optical data available for few regions indicated high mapping accuracy (≥87%). The total water inundated area was identified as identified as 1453, 2081 and 3634 km2 in 2018, 2019 and 2020, respectively. All the districts in the floodplain of the Brahmaputra River and three districts in the southern region of Assam experienced three-time flood events in the last 3 years. Overlaying with vegetation map indicated that Vthe flood events severely affected vast cropland, grassland and fores. Overlaying with vegetation map indicated that the flood events severely affected vast cropland, grassland and forest. The Sentinel-1 SAR data and GEE platform allowed rapid flood assessment and enabled real-time mapping, which is particularly important for the decision-makers to develop flood controlling measures, mitigation, relief,and rescue planning.KeywordsFlood mappingBrahmaputra RiverSentinel-1Google Earth Engine (GEE)

The increasing trend of industrial development has led to rapid changes in land use over the past few years. The changes in land use in Saraburi Province were studied during three periods: the years 2000, 2010, and 2021, using Landsat satellite imagery processed with the Google Earth Engine (GEE) platform. The Random Forest (RF) machine learning algorithm was employed to classify land use activities into five categories based on the criteria set by the Department of Land Development. The overall accuracy values for land use were 83.59%, 82.42%, and 80.78%, respectively. The changes in land use have clearly affected the urban and built-up area. In 2000, there were only 7,561.15 ha, but by 2021, the area had increased to 34,221.11 ha. Forest areas also saw an increase. In 2000, there were only 79,626.26 ha, which increased to 163,542.05 ha in 2021, reflecting that Saraburi Province places importance on green areas. In contrast, agricultural areas decreased, which was attributed to the increasing number of households, according to the statistical report from the Department of Provincial Administration. In 2000, there were 168,979 households, and by 2021, they had increased to 288,275 households, resulting in a change in land use from agricultural areas to urban and built-up areas, totaling 22,114.95 ha. This information can be used as a guideline for planning development and land use to promote Saraburi Province, aiming towards becoming an eco-industrial city that is livable and has sustainable environmental quality in the future.

Monitoring changes in vegetation cover is important for the restoration of ecosystems in the Gorontalo Regency area. The utilization of remote sensing technology makes it possible to detect the dynamics of changes in vegetation cover spatially and temporally. The Terra MODIS satellite image collection in the study area is available in large numbers and sizes. Therefore, cloud computing-based spatial technology support is needed. Google Earth Engine (GEE) as a geospatial computing device is an alternative to cover this shortfall. The aim of this study is to explore the condition of vegetation cover spatially and temporally using the GEE platform. A total of 43 MODIS images in the study area, recording periods 2000 and 2020, were used to quickly and effectively generate vegetation cover maps. The process of downloading, processing, and analyzing data was automated through the GEE interface. The results of the mapping in 2000 and 2020 are shown by maps of vegetation cover in two classes, namely; vegetation and non-vegetation. The accuracy of the vegetation cover map shows good results, namely an overall accuracy of 0.81 for 2000 and 0.85 for 2020. The area of the non-vegetation class increased by 2815.29 ha, and the vegetation class decreased by 2767.31 ha. The map of spatial changes in vegetation cover in the study area is classified into three classes, namely revegetation, devegetation, and unchanged. Based on these results, the extraction of vegetation cover changes in the study area using the GEE platform can be carried out well.

Simple SummaryForecasting dengue cases often face challenges from (1) time-effectiveness due to time-consuming satellite data downloading and processing, (2) weak spatial representation due to data dependence on administrative unit-based statistics or weather station-based observations, and (3) stagnant accuracy without historical dengue cases. With the advance of the geospatial big data cloud computing in Google Earth Engine and deep learning, this study proposed an efficient framework of dengue prediction at an epidemiological week basis using geospatial big data analysis in Google Earth Engine and Long Short Term Memory modeling. We focused on the dengue epidemics in the Federal District of Brazil during 2007–2019. Based on Google Earth Engine and epidemiological calendar, we computed the weekly composite for each dengue driving factor, and spatially aggregated the pixel values into dengue transmission areas to generate the time series of driving factors. A multi-step-ahead Long Short Term Memory modeling was used, and the time-differenced natural log-transformed dengue cases and the time series of driving factors were considered as outcomes and explantary factors, respectively, with two modeling scenarios (with and without historical cases). The performance is better when historical cases were used, and the 5-weeks-ahead forecast has the best performance.Timely and accurate forecasts of dengue cases are of great importance for guiding disease prevention strategies, but still face challenges from (1) time-effectiveness due to time-consuming satellite data downloading and processing, (2) weak spatial representation capability due to data dependence on administrative unit-based statistics or weather station-based observations, and (3) stagnant accuracy without the application of historical case information. Geospatial big data, cloud computing platforms (e.g., Google Earth Engine, GEE), and emerging deep learning algorithms (e.g., long short term memory, LSTM) provide new opportunities for advancing these efforts. Here, we focused on the dengue epidemics in the urban agglomeration of the Federal District of Brazil (FDB) during 2007–2019. A new framework was proposed using geospatial big data analysis in the Google Earth Engine (GEE) platform and long short term memory (LSTM) modeling for dengue case forecasts over an epidemiological week basis. We first defined a buffer zone around an impervious area as the main area of dengue transmission by considering the impervious area as a human-dominated area and used the maximum distance of the flight range of Aedes aegypti and Aedes albopictus as a buffer distance. Those zones were used as units for further attribution analyses of dengue epidemics by aggregating the pixel values into the zones. The near weekly composite of potential driving factors was generated in GEE using the epidemiological weeks during 2007–2019, from the relevant geospatial data with daily or sub-daily temporal resolution. A multi-step-ahead LSTM model was used, and the time-differenced natural log-transformed dengue cases were used as outcomes. Two modeling scenarios (with and without historical dengue cases) were set to examine the potential of historical information on dengue forecasts. The results indicate that the performance was better when historical dengue cases were used and the 5-weeks-ahead forecast had the best performance, and the peak of a large outbreak in 2019 was accurately forecasted. The proposed framework in this study suggests the potential of the GEE platform, the LSTM algorithm, as well as historical information for dengue risk forecasting, which can easily be extensively applied to other regions or globally for timely and practical dengue forecasts.

Modelling inland Arctic bathymetry from space using cloud-based machine learning and Sentinel-2

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

Land use land cover change (LULC) significantly impacts urban sustainability, urban planning, climate change, natural resource management, and biodiversity. The Chattogram Metropolitan Area (CMA) has been going through rapid urbanization, which has impacted the LULC transformation and accelerated the growth of urban sprawl and unplanned development. To map those urban sprawls and natural resources depletion, this study aims to monitor the LULC change using Landsat satellite imagery from 2003 to 2023 in the cloud-based remote sensing platform Google Earth Engine (GEE). LULC has been classified into five distinct classes: waterbody, build-up, bare land, dense vegetation, and cropland, employing four machine learning algorithms (random forest, gradient tree boost, classification & regression tree, and support vector machine) in the GEE platform. The overall accuracy (kappa statistics) and the receiver operating characteristic (ROC) curve have demonstrated satisfactory results. The results indicate that the CART model outperforms other LULC models when considering efficiency and accuracy in the designated study region. The analysis of LULC conversions revealed notable trends, patterns, and magnitudes across all periods: 2003–2013, 2013–2023, and 2003–2023. The expansion of unregulated built-up areas and the decline of croplands emerged as primary concerns. However, there was a positive indication of a significant increase in dense vegetation within the study area over the 20 years.

Accurate and reliable lithological mapping through satellite-borne remote sensing data and image classification approaches has a critical role since it can automatically and promptly identify lithological units over large areas. Most available Pixel-Object Based comparative classification studies have been applied to land use land cover (LULC) studies; however, this research aims to evaluate and compare the performance of these digital classification methods in the field of geological mapping in semi-arid areas, by integrating spectral bands and neo-bands, particularly the Minimum noise fraction (MNF) and the principal component analysis (PCA), of Sentinel-2A satellite imagery, to map the southern of Skhour Rehamna which is located at the western Moroccan Meseta. The analysis results from two different methods, namely, pixel-based image analysis (PBIA) with k-nearest neighbour (K-NN) and Random Forest (RF) machine learning algorithms (MLAs), and Geographic Object-Based Image Analysis (GEOBIA) were assessed and compared. PBIA method involved selection of training areas whether it was k-NN or RF MLAs, and produced lithological maps that exhibit “salt and pepper” effects as well as problems associated to delineating accurate lithological boundaries, while GEOBIA approach involved multi-resolution segmentation step where scale, shape and compactness parameters should be adjusted as accurate as possible, in order to segment the image into homogeneous and meaningful regions so that the resulted samples were classified using Standard Nearest Neighbour algorithm. Therefore, the resulting lithological maps were assessed by comparing both techniques using confusion matrix, overall accuracy (OA) and Kappa coefficient (K). The results show that the GEOBIA approach had higher overall agreement (83.46% OA and 0.76 K) than RF (81.92% OA and 0.72 K) and k-NN (80.79% OA and 0.70 K) PBIA approaches. Overall, the results clearly indicate the potential of GEOBIA technique for lithological mapping applications to produce more realistic maps.

Drought is one of the frequently observed natural hazard resulting from precipitation deficit and increased evapotranspiration caused by high temperatures. Remote sensing indices are used to analyze spatio-temporal distribution of drought conditions and identify drought severity. In this study, we analyzed the spatio-temporal distribution of drought conditions in Turkey from February 2000 to January 2019 by using different drought indices produced from MODIS satellite data in Google Earth Engine (GEE) platform. Vegetation Health Index (VHI), Normalized Multiband Drought Index(NMDI) and Normalized Difference Drought Index (NDDI) maps in country level for different years and months of the related years were utilized to assess the drought conditions. Time series were also created for some specific locations to deeply analyze the drought conditions in 20-year period. Our results show that MODIS derived drought indices provide useful geospatial information to assess drought conditions in country level. Moreover, GEE platform is very handy and rapid tool to reach related satellite images and conduct remote sensing analysis of huge and long term date efficiently. Geospatial big data could be successfully accessed and processed in this platform not only for drought monitoring but also for other environmental monitoring applications.

Ordos Plateau is one of the primary sources of sediment in the Yellow River, and changes in regional soil erosion directly affect the ecological status of the lower reaches of the Yellow River. Many recent studies have been published using remote sensing (RS) and geographic information systems (GIS) to evaluate soil erosion. In this study, much satellite remote sensing data in the Google Earth Engine (GEE) can better track soil erosion protection, which is significant in guiding the ecological protection and restoration of the Ordos Plateau and the Yellow River basin. In this study, we used GEE to observe the changes in soil erosion in the Ordos area from 2013 to 2021. The Theil–Sen procedure and Mann–Kendall significance test methods were used to evaluate the trend of land erosion in the Ordos area from 2013 to 2021. Based on GEE, the RUSLE is applied to evaluate soil erosion and analyze the changing trend. As a result, (1) we found that the annual change of soil and water loss in the Ordos Plateau showed three stages: 2013–2015, 2016–2018, and 2018–2021. After 2018, soil loss decreased from 14 × 1017 Mg in 2018 to 4 × 1017 Mg in 2021, which indicates that the environmental restoration project started in 2018 has achieved encouraging results. (2) The results showed that 40.9% of the regional soil erosion trend showed a significant decline, and 45.7% of the regional soil erosion trend showed a slight decline. Only 13.3% of the regional soil erosion is on the rise. (3) The test results of different land use types show that 87.3% of soil erosion occurs in bare and cultivated land. Because the terrain of Ordos is relatively flat, 95.39–96.17% of soil erosion occurs in areas with a slope of 0 to 5. (4) The reliability of the RUSLE model based on the GEE platform is proved by regression model verification of observation data and model prediction results. (5) GEE’s cloud-based features can provide data and scripts to users in developing countries which lack sufficient observation data or the necessary computing resources to develop these data. The results show that GEE has robust analysis and processing ability, can analyze a large amount of data, and can provide efficient digital products for soil erosion research.

Google Earth Engine (GEE) is a versatile cloud platform in which pixel-based (PB) and object-oriented (OO) Land Use–Land Cover (LULC) classification approaches can be implemented, thanks to the availability of the many state-of-art functions comprising various Machine Learning (ML) algorithms. OO approaches, including both object segmentation and object textural analysis, are still not common in the GEE environment, probably due to the difficulties existing in concatenating the proper functions, and in tuning the various parameters to overcome the GEE computational limits. In this context, this work is aimed at developing and testing an OO classification approach combining the Simple Non-Iterative Clustering (SNIC) algorithm to identify spatial clusters, the Gray-Level Co-occurrence Matrix (GLCM) to calculate cluster textural indices, and two ML algorithms (Random Forest (RF) or Support Vector Machine (SVM)) to perform the final classification. A Principal Components Analysis (PCA) is applied to the main seven GLCM indices to synthesize in one band the textural information used for the OO classification. The proposed approach is implemented in a user-friendly, freely available GEE code useful to perform the OO classification, tuning various parameters (e.g., choose the input bands, select the classification algorithm, test various segmentation scales) and compare it with a PB approach. The accuracy of OO and PB classifications can be assessed both visually and through two confusion matrices that can be used to calculate the relevant statistics (producer’s, user’s, overall accuracy (OA)). The proposed methodology was broadly tested in a 154 km2 study area, located in the Lake Trasimeno area (central Italy), using Landsat 8 (L8), Sentinel 2 (S2), and PlanetScope (PS) data. The area was selected considering its complex LULC mosaic mainly composed of artificial surfaces, annual and permanent crops, small lakes, and wooded areas. In the study area, the various tests produced interesting results on the different datasets (OA: PB RF (L8 = 72.7%, S2 = 82%, PS = 74.2), PB SVM (L8 = 79.1%, S2 = 80.2%, PS = 74.8%), OO RF (L8 = 64%, S2 = 89.3%, PS = 77.9), OO SVM (L8 = 70.4, S2 = 86.9%, PS = 73.9)). The broad code application demonstrated very good reliability of the whole process, even though the OO classification process resulted, sometimes, too demanding on higher resolution data, considering the available computational GEE resources.

Abstract. This study aims to use remote sensing techniques to map the urban region of Ankara from the past to the present, assessing the nature, magnitude and direction of changes within the area, including the transformation of LULC classes and explaining the driving forces behind these transformations. The study encompasses three stages. Firstly, Landsat 7 ETM+ images from 2000 and Sentinel-2 satellite images from 2020 were obtained for Ankara city and surroundings through the Google Earth Engine (GEE) platform. Image classification was conducted for both 2000 and 2020 using 'Blue', 'Green', 'Red', 'Vegetation Red Edge1', 'Vegetation Red Edge2', 'Vegetation Red Edge3', 'NIR', 'Vegetation Red Edge4', 'Water vapour', ' SWIR1', 'SWIR2' bands, as well as 'NDWI', 'NDVI', 'NDBI' indices on the GEE platform. LULC was classified using the Random Forest (RF) classifier, which included six classes: urban area, forest, water surfaces, open areas, agricultural areas and roads. Secondly, the LULC maps of the 2000 and 2020 images were classified using RF. The study employed the 'Categorical Change, Pixel Value Change and Time Series Change' methods to determine the transformations between LULC categories. Specifically, the urban change within the study area increased by 70% between 2000 and 2020. Over the past 20 years, from 2000 to 2020, the urban areas in Ankara expanded by 170%. Consequently, accurately determining the nature, magnitude and direction of urban development using remote sensing data offers valuable baseline information for various disciplines related to spatial planning at local and national scales.