A flood assessment of data scarce region using an open-source 2D hydrodynamic modeling and Google Earth Image: a case of Sabarmati flood, India

The Windows screen-capture tools was used to get the Google Earth (GE) images. Compared with the original remote sensing images, although the image quality was reduced and the spectral information was lacking, it has been able to meet the needs of this study. A method for estimating fractional vegetation coverage (FVC) using GE images based on K-Means algorithm was proposed. Firstly, GE image was preliminarily classified by using K-Means algorithm. Secondly, by visual interpretation, the initial classification results were further clustered into 4 types according to the number and brightness feature of land surface types in the image, low brightness (shadow), medium low brightness (high density vegetation), medium high brightness (sparse vegetation), high brightness (bare ground), the FVC of each category was determined by its characteristics and composition. Finally, weighted by the proportion of pixels in the image, took the weighted sum of the FVC of all categories as the FVC of the image. In addition, the field survey data were used to verify the FVC estimated by the proposed method, the results showed that: the precision of estimated vegetation coverage could reach 80% ~ 90%.

Large-scale mapping of gully-affected areas: An approach integrating Google Earth images and terrain skeleton information

Migrating rivers, consequent paleochannels: The unlikely partners and hotspots of flooding

Despite many studies on land degradation in the Highlands of Northern Ethiopia, quantitative information regarding long-term changes in land use/cover (LUC) is rare. Hence, this study aims to investigate the LUC changes in the Geba catchment (5142 km2), Northern Ethiopia, over 80 years (1935–2014). Aerial photographs (APs) of the 1930s and Google Earth (GE) images (2014) were used. The point-count technique was utilized by overlaying a grid on APs and GE images. The occurrence of cropland, forest, grassland, shrubland, bare land, built-up areas and water body was counted to compute their fractions. A multivariate adaptive regression spline was applied to identify the explanatory factors of LUC and to create fractional maps of LUC. The results indicate significant changes of most types, except for forest and cropland. In the 1930s, shrubland (48%) was dominant, followed by cropland (39%). The fraction of cropland in 2014 (42%) remained approximately the same as in the 1930s, while shrubland significantly dropped to 37%. Forests shrank further from a meagre 6.3% in the 1930s to 2.3% in 2014. High overall accuracies (93% and 83%) and strong Kappa coefficients (89% and 72%) for point counts and fractional maps respectively indicate the validity of the techniques used for LUC mapping.

Land cover change is a major environmental concern in the northwestern highlands of Ethiopia. This study detected land cover transitions over the past 30 years in the Chemoga basin (total area = 118,359 ha). Land cover maps were generated via the supervised classification of Landsat images with the help of the Google Earth (GE) images. A total of 218 unchanged land features sampled from GE images were used as the training datasets. Classification accuracy was evaluated by comparing classified images with 165 field observations during the 2017 field visit. The overall accuracy was 85.4% and the kappa statistic was 0.81, implying that the land classification was satisfactory. Agricultural land is the dominant land cover in the study basin, and increased in extent by 2,337 ha from 1987 to 2017. The second and third most dominant land cover types, grassland and woodland, decreased by 1.9% and 3.6%, respectively, over the past 30 years. The increase in agricultural lands was mostly due to the conversion of grasslands and woodlands, although some agricultural lands changed to Eucalyptus plantations and human settlements. The results revealed that the expansion of built-up space and agricultural lands was the major driver of fragmentation of the landscape, and degradation of natural resources in the Chemoga basin, Ethiopia.

We propose and demonstrate a fast, robust method for using satellite images to localize an Unmanned Aerial Vehicle (UAV). Previous work using satellite images has large storage and computation costs and is unable to run in real time. In this work, we collect Google Earth (GE) images for a desired flight path offline and an autoencoder is trained to compress these images to a low-dimensional vector representation while retaining the key features. This trained autoencoder is used to compress a real UAV image, which is then compared to the precollected, nearby, autoencoded GE images using an inner-product kernel. This results in a distribution of weights over the corresponding GE image poses and is used to generate a single localization and associated covariance to represent uncertainty. Our localization is computed in 1% of the time of the current standard and is able to achieve a comparable RMSE of less than 3 m in our experiments, where we robustly matched UAV images from six runs spanning the lighting conditions of a single day to the same map of satellite images.

ABSTRACTRapidEye satellite images with high spatial resolution, affordable prices and having Red-Edge band have high potential for time series issues, especially in vegetation studies. Despite these beneficial properties, RapidEye images with 5 m spatial resolution are not sufficiently useful for some applications. According to this problem, enhancing the spatial resolution of RapidEye images can significantly improve the results of the subsequent processes on these images. Fusion of high spatial resolution with high spectral resolution images is known as an effective way to enhance the quality of multispectral remotely sensed images. Unfortunately, the lack of panchromatic band with high spatial resolution has been faced the procedure of improving the spatial resolution of RapidEye images with major problems. In this paper, we have proposed using the free Google Earth (GE) images which have high spatial resolution and high-coverage of land surface to enhance the spatial information of RapidEye images. A simulated panchromatic image has been generated by three band GE image and with three different methods: Mean, principal component analysis (PCA) and weighted average of GE image bands. In the last method, the weights are extracted from the spectral response curve of the satellite which captured the GE image. The simulated panchromatic image has been utilized for pansharpening of RapidEye image in five well-known methods: Brovey, Gram-Schmidt (GS), intensity-hue-saturation (IHS), Pansharp1 and Pansharp2. The most important point is finding the GE image with lowest lag time with RapidEye image. By satisfying this condition, the experiments illuminated that the proposed method can effectively enhance the spatial quality of RapidEye image. Also, this study presented that Pansharp2 method, which used simulated panchromatic image generated by the spectral response curve information, has revealed the best results of RapidEye image pansharpening.

Application of an improved U-Net with image-to-image translation and transfer learning in peach orchard segmentation

The topographic skeleton is the primary expression and intuitive understanding of topographic relief. This study integrated a topographic skeleton into deep learning for terrain reconstruction. Firstly, a topographic skeleton, such as valley, ridge, and gully lines, was extracted from a global digital elevation model (GDEM) and Google Earth Image (GEI). Then, the Conditional Generative Adversarial Network (CGAN) was used to learn the elevation sequence information between the topographic skeleton and high-precision 5 m DEMs. Thirdly, different combinations of topographic skeletons extracted from 5 m, 12.5 m, and 30 m DEMs and a 1 m GEI were compared for reconstructing 5 m DEMs. The results show the following: (1) from the perspective of the visual effect, the 5 m DEMs generated with the three combinations (5 m DEM + 1 m GEI, 12.5 m DEM + 1 m GEI, and 30 m DEM + 1 m GEI) were all similar to the original 5 m DEM (reference data), which provides a markedly increased level of terrain detail information when compared to the traditional interpolation methods; (2) from the perspective of elevation accuracy, the 5 m DEMs reconstructed by the three combinations have a high correlation (>0.9) with the reference data, while the vertical accuracy of the 12.5 m DEM + 1 m GEI combination is obviously higher than that of the 30 m DEM + 1 m GEI combination; and (3) from the perspective of topographic factors, the distribution trends of the reconstructed 5 m DEMs are all close to the reference data in terms of the extracted slope and aspect. This study enhances the quality of open-source DEMs and introduces innovative ideas for producing high-precision DEMs. Among the three combinations, we recommend the 12.5 m DEM + 1 m GEI combination for DEM reconstruction due to its relative high accuracy and open access. In regions where a field survey of high-precision DEMs is difficult, open-source DEMs combined with GEI can be used in high-precision DEM reconstruction.

Mapping gullies over large areas requires detailed aerophotos. Google Earth (GE) provides free access to high resolution satellite imagery, but is the quality good enough to map gullies reliably? The aim of this work was to evaluate the accuracy with which gullies in the West Gippsland region could be mapped using GE images. The area comprised the LaTrobe, Thomson and Avon catchments and extended over 11130 km 2 in Victoria, south-eastern Australia. GE images available for the West Gippsland area at the time of the study were from three sources: Cnes/Spot Image of 2.5 m ground resolution, Digital Globe of 0.6 m resolution, and GeoEye of 0.5m resolution. Gullies were identified and digitized from GE images, then transferred to GIS. After digitization of a pilot area (192 km 2 ), on both forested and agricultural land, an initial field survey was conducted in December 2010 to improve interpretation of GE images. The pilot study showed that large canopy cover in plantations and native forest precluded observation of gullies beneath. Gullies under forest canopy could only be recognized on areas felled or burnt before the image was taken. Following the pilot area evaluation, gully mapping in West Gippsland was restricted to agricultural land; forest and plantation areas, covering 64 % of the catchment, were excluded. A second, more extensive field survey was conducted in January 2011 on 39 transects to assess the accuracy (absence/presence) of the gully map on agricultural land. Gullies totaling 2385 km were mapped in agricultural areas across the region. Most gullies (87%) were located along drainage lines and were connected to streams. It was sometimes difficult to separate gully from streambank erosion. Following field observations, streams were defined as drainage lines of third or greater order (Strahler method), whereas incised first and second order drainage lines were classified as gullies. Gully density on agricultural land increased from West to East across the West Gippsland region, varying from 0.59 km/km 2 in the LaTrobe catchment, to 0.65 km/km 2 in the Thomson and 0.86 km/km 2 in the Avon catchment. The field survey showed that 26% of gullies observed were not mapped from GE images, whereas 13% of mapped gullies were not confirmed by the field survey. Mapping errors were correlated to image resolution, with higher errors associated with coarser resolution images. During the survey, 12 representative cross-sections of gullies were measured by recording the maximum depth, width, and taking a perpendicular photograph. Gullies were generally small and inactive, having a median cross section of 2.7 m 2 (1.7-4.1 m 2 interquartile range). By assuming an exponential decay of gully wall retreat in the gully stabilization phase, current gully erosion rate of active gullies was assessed at 0.02 m 2 /y. Together with the revised gully network extent, suspended sediment load originated by gully erosion in the region was estimated at 10.6 kt/y. This estimate is higher than reported in previous research, due in part to the higher gully density found in this study, as well as to differences in defining gully and streams and in erosion rate estimates. Conditions in the West Gippsland region were not ideal for appraisal of gully erosion using GE images because of the large extent of forest areas and the small gully system occurring in agricultural land. A major limitation of the method was that image resolution was too coarse to distinguish between active or inactive gully areas. Despite the limits highlighted, the use of GE images allowed appraisal of gully extent over a very large area in relatively short time and at no cost for image acquisition. An application of the method in agricultural catchments with older, larger gully systems, such as frequently found in South-east Australia, would be likely to result in much lower errors than found this case study. We conclude that interpretation of GE images for rapid appraisal of gully extent on large areas is a useful approach particularly where old, well developed gully systems are prevalent in agricultural/cleared land. Further work, confirmed by field survey verification, would be useful.

The Sample can drive classification algorithms, thus is a prerequisite for accurate classification. Coastal areas are located in the transitional zone between land and sea, requiring more samples to describe diverse land covers. However, there are scarce studies sharing their sample datasets, leading to a repeat of the time-consuming and laborious sampling procedure. To alleviate the problem, we share a sample set with a total of 16,444 sample points derived from a study of mapping mangroves of China. The sample set contains a total of 10 categories, which are described as follows. 1) The mangroves refer to “true mangroves” (excluding the associate mangrove species). In sampling mangroves, we used the data from the China Mangrove Conservation Network (CMCN, http://www.china-mangrove.org/), a non-governmental organization aiming to promote mangrove ecosystems. The CMCN provides an interactive map that can be annotated by volunteers with text or photos to record mangrove status at a location. Although the locations were shifted due to coordinate system differences and positioning errors, mangroves could be found around the mangrove locations depicted by the CMCN’s map on Google Earth images. There is a total of 1887 mangrove samples. 2) The cropland is dominated by paddy rice. We collected a total 1383 points according to its neat arrangement based on Google Earth images. 3) Coastal forests neighboring mangroves are mostly salt-tolerant, such as Cocos nucifera Linn., Hibiscus tiliaceus Linn., and Cerbera manghas Linn. We collected a total 1158 samples according to their distance to the shoreline based on Google Earth images. 4) Terrestrial forests are forests far from the shoreline, and are intolerant to salt. By visual inspection on Google Earth, we sampled 1269 points based on their appearances and distances to the shoreline. 5) For the grass category, we collected 1282 samples by visual judgement on Google Earth. 6) Saltmarsh, dominated by Spartina alterniflora, covering large areas of tidal flats in China. We collected 2065 samples according to Google Earth images. 7) The tidal flats category was represented by 1517 samples, which were sampled using the most recent global tidal flat map for 2014–2016 and were visually corrected. 8) The “sand or rock” category refers to sandy and pebble beaches or rocky coasts exposed to air, which are not habitats of mangroves. We collected 1622 samples on Google Earth based on visual inspection. 9) For the permanent water category, samples were first randomly sampled from a threshold result of NDWI (> 0.2), and then were visually corrected. A total of 2056 samples were obtained. 10) As to the artificial impervious surfaces category, we randomly sampled from a threshold result corresponding to normal difference built-up index (NDBI) (> 0.1), and corrected them based on Google Earth. The artificial impervious surface category was represented by 2205 samples. This sample dataset covers the low-altitude coastal area of five Provinces (Hainan, Guangdong, Fujian, Zhejiang, and Taiwan), one Autonomous region (Guangxi), and two Special Administrative Regions (Macau and Hong Kong) (see “study_area.shp” in the zip for details). It can be used to train models for coastal land cover classification, and to evaluate classification results. In addition to mangroves, it can also be used in identifying tidal flats, mapping salt marsh, extracting water bodies, and other related applications.Compared with the V1 version, we added a validation dataset for mangrove maps (Mangrove map validation dataset.rar), and thus can evaluate mangrove maps under the same dataset, which benefit the comparison of different mangrove maps. The validation dataset contains 10 shp files, in which each shp file contains 600 mangrove samples (cls_new field = 1) and 600 non-mangrove samples (cls_new field = 0).Compared with the V2 version, we added two classes of forest near water and grass near water, in addition to suppress the prevalent misclassified patches due to the spectral similarity between mangroves and those classes.

Monitoring and mapping complex urban features (e.g. roads and buildings) from remotely sensed data multispectral and hyperspectral has gained enormous research interest. Accurate ground truth allows for high quality assessment of classified images and to verify the produced map. Ground truth can be acquired from: field using the handheld Global Positioning System (GPS) device and from Images with high resolution extracted from Google Earth in additional to field. Ground truth or training samples could be achieved from VHR satellite images such as QuickBird, Ikonos, Geoeye-1 and Wordview images. Archived images are costly for researchers in developing countries. Images from GE with high spatial resolution are free for public and can be used directly producing large scale maps, in producing LULC mapping and training samples. Google Earth (GE) provides free access to high resolution satellite imagery, but is the quality good enough to map urban areas. Costal of the Red sea, Marsa Alam could be mapped using GE images. The main objective of this research is exploring the accuracy assessment of producing large scale maps from free Google Earth imagery and to collect ground truth or training samples in limited geographical extend. This research will be performed on Marsa Alam city or located on the western shore of the Red Sea, Red sea Governorate, Egypt. Marsa Alam is located 274 km south of Hurghada. The proposed methodology involves image collection taken into consideration the resolution of collected photographs which depend on the height of view. After that, image rectification using suitable rectification methods with different number and distributions of GCPs and CPs. Database and Geographic information systems (GIS) layers were created by on-screen vectorization based on the requirement of large scale maps. Attribute data have been collected from the field. The obtained results show that the planmetric accuracy of the produced map from Google Earth Images met map scale 10 000 according to (National Map Accuracy Standards).The collect ground truth or training samples from GE images and field help in accuracy assessment of classification process.

Provision of land use and forest density maps in semi-arid areas of Iran using Sentinel-2 satellite images and vegetation indices

The evolution of remote sensing technologies has improved scientific study and research. The often high cost of satellite images has led to research into alternative remotely sensed data for various analysis. Google earth data being cheap and readily available has been employed in many analyses including textural analysis with positive results. The application of OBIA to google earth image was employed in this study to assess the predictive ability of rubber tree diameter at breast height towards carbon modelling. Out of a total of 190 manually delineated tree crowns, 102 trees were found to have a 1 to 1 matching with segmented crowns on the Google Earth images were used. For the whole study area over- segmentation value was 0.43 (43% error) and the under-segmentation was 0.32 (32% error) with the D-Value computed as 0.38 (38% error) which means that the segmentation accuracy is 62%. Models developed from the segmentation process and field data were linear, quadratic and cubic models with R2 of 0.014, 0.137 and 0.139 respectively. Primarily, these poor R2 values are due to the fact that Google earth images have poor spectral values, red and infrared portions are absent which affect the clear crown detection of the tree canopies. The tree canopies are equally highly clustered, therefore with poor spectral values individual tree detection using OBIA procedure achieves very little success in the diameter at breast height prediction.

Automatic classification of Google Earth images for a larger scale monitoring of bush encroachment in South Africa