Leveraging autoencoders in change vector analysis of optical satellite images

Various applications in remote sensing demand automatic detection of changes in optical satellite images of the same scene acquired over time. This paper investigates how to leverage autoencoders in change vector analysis, in order to better delineate possible changes in a couple of co-registered, optical satellite images. Let us consider both a primary image and a secondary image acquired over time in the same scene. First an autoencoder artificial neural network is trained on the primary image. Then the reconstruction of both images is restored via the trained autoencoder so that the spectral angle distance can be computed pixelwise on the reconstructed data vectors. Finally, a threshold algorithm is used to automatically separate the foreground changed pixels from the unchanged background. The assessment of the proposed method is performed in three couples of benchmark hyperspectral images using different criteria, such as overall accuracy, missed alarms and false alarms. In addition, the method supplies promising results in the analysis of a couple of multispectral images of the burned area in the Majella National Park (Italy).

Evaluating the impacts of major cyclonic catastrophes in coastal Bangladesh using geospatial techniques

Cyclonic catastrophes frequently devastate coastal regions of Bangladesh that host around 35 million people which represents two-thirds of the total population. They have caused many problems like agricultural crop loss, forest degradation, damage to built-up areas, river and shoreline changes that are linked to people’s livelihood and ecological biodiversity. There is an absence of a comprehensive assessment of the major cyclonic disasters of Bangladesh that integrates geospatial technologies in a single study. This study aims to integrate geospatial technologies with major disasters and compares them, which has not been tried before. This paper tried to identify impacts that occurred in the coastal region by major catastrophic events at a vast level using different geospatial technologies. It focuses to identify the impacts of major catastrophic events on livelihood and food production as well as compare the impacts and intensity of different disasters. Furthermore, it compared the losses among several districts and for that previous and post-satellite images of disasters that occurred in 1988, 1991, 2007, 2009, 2019 were used. Classification technique like machine learning algorithm was done in pre- to post-disaster images. For quantifying change in the indication of different factors, indices including NDVI, NDWI, NDBI were developed. “Change vector analysis” equation was performed in bands of the images of pre- and post-disaster to identify the magnitude of change. Also, crop production variance was analyzed to detect impacts on crop production. Furthermore, the changes in shallow to deep water were analyzed. There is a notable change in shallow to deep water bodies after each disaster in Satkhira and Bhola district but subtle changes in Khulna and Bagerhat districts. Change vector analysis revealed greater intensity in Bhola in 1988 and Satkhira in 1991. Furthermore, over the years 2007 and 2009 it showed medium and deep intense areas all over the region. A sharp decrease in Aus rice production is witnessed in Barishal in 2007 when cyclone “Sidr” was stricken. The declination of potato production is seen in Khulna district after the 1988 cyclone. A huge change in the land-use classes from classified images like water body, Pasture land in 1988 and water body, forest in 1991 is marked out. Besides, a clear variation in the settlement was observed from the classified images. This study explores the necessity of using more geospatial technologies in disastrous impacts assessment around the world in the context of Bangladesh and, also, emphasizes taking effective, proper and sustainable disaster management and mitigation measures to counter future disastrous impacts.

Rapid detection of land cover change in tropical savanna environment using conditional change vector analysis on remote sensing data in Moyo watershed, Sumbawa Regency, West Nusa Tenggara Province, Indonesia

In environmental management, land cover change is a crucial aspect. The area of tropical savanna environments is vulnerable to land degradation. This study aimed to rapidly detect land cover changes in a tropical savanna environment based on remote sensing data. Conditional change detection was performed using the Change Vector Analysis (CVA) with input parameters such as the Enhanced Vegetation Index (EVI) and Normalized Difference Soil Index (NDSI). The results showed that during the period 2015 to 2019, changes were detected in the Moyo watershed every year. From 2015 to 2016, the Moyo River Basin was dominated by changes with a change magnitude of less than 0.088, which was 63% of the Moyo River Basin area. From 2016 to 2017, the changes were dominated by the change magnitude value of 0.063, which was 58.6% of the Moyo River Basin area. From 2017 to 2018, changes were dominated by the change magnitude value of 0.084 of 55.26% of the Moyo watershed area. From 2018 to 2019, the change was dominated by the change magnitude value of 0.057, which was 47.57% of the Moyo watershed area. The direction of land cover change was dominated by Q2 in 2016, Q4 in 2017 and 2018, and Q2 and Q4 in 2019. These changes generally occurred in the Moyo watershed middle and downstream parts, which are grasslands. The use of the Conditional Change Vector Analysis (CCVA) approach in a tropical savanna environment can detect changes and the direction of change with an accuracy of about 70%.

Two extreme rain events over the Atacama Desert and their consequences on the surface: Insights from Change Vector Analysis applied to Landsat 8 OLI imagery

<p>In March of 2015 there was probably the most studied rain event that ever occurred in the Atacama Desert. Three days of heavy rain impacted the southern region, with peak amounts of 85 mm locally. Different approaches have been used to study this event, including field observations, isotopic analysis and examination of InSAR data. During February of 2019 there was another rain event in the northern Atacama Desert, during which over 160 mm of rain fell on the eastern part of the Atacama, and the influence on the surface is still unknown. This study examines both events. The two study areas have different relationships to the rain: the 2015 event is analyzed within the area in which it rained, whereas the 2019 study area is 60 km away from the heavy rain, connected by surface water drainage. Results of particular interest are the variable responses of the different types of surface materials (e.g., varying classes of terrain roughness and mineralogy) and the identification of locations of erosion and deposition.</p> <p>We examine multispectral satellite imagery from the Landsat 8 satellite, an approach that has some advantages over other methods. Advantages include its free access, a longer historical record that may allow examination of more events, and the existence of observations at multiple wavelengths which allows evaluation of mineral phase changes due to the rain, vegetation increment and changes in the type of material.</p> <p>In this work we apply Change Vector Analysis (CVA) (Bruzzone and Fernandez, 2000) to Landsat 8 OLI images to, first, validate the multispectral satellite CVA results using as ground truth the InSAR permanent coherence loss from the 2015 event. Then we apply the method to identify changes due to the 2019 rain event. We compared these results to our field observations.</p> <p>Our results indicate that: 1) CVA applied to Landsat bracketing the 2015 rain event identifies the depositional and erosional areas, correlating well to permanent changes detected by InSAR coherence loss. 2) Surface materials react variably, and some categories of materials changed more due to a rain than others. 3) Spectral analysis and CVA do not detect mineralogic phase responses documented by surface data. 4) Wind driven changes were also detected in some areas. 5) Field observations reveal that erosion and deposition are always well identified by the algorithm as long as the extent of change is larger than the pixel size. 6) The distribution of changes is dependent on surface slope.</p> <p><strong>Reference</strong></p> <p>Lorenzo Bruzzone and Diego Fernández Prieto. Automatic Analysis of the Difference Image for Unsupervised Change Detection. Technical Report 3, 2000. DOI: 10.1109/36.843009</p>

Updating of Land Cover Maps and Change Analysis Using GlobeLand30 Product: A Case Study in Shanghai Metropolitan Area, China

Accurate land cover mapping and change analysis is essential for natural resource management and ecosystem monitoring. GlobeLand30 is a global land cover product from China with 30 m resolution that provides reliable data for many international scientific programs. Few studies have focused on systematically implementing this global land cover product in regional studies. Therefore, this paper presents an object-based extended change vector analysis (ECVA_OB) and transfer learning method to update the reginal land cover map using GlobeLand30 product. The method is designed to highlight small and subtle changes through the concept of uncertain area analysis. Updating is carried out by classifying changed objects using a change-detection-based transfer learning method. Land cover changes are analyzed and the factors affecting updating results are explored. The method was tested with data from Shanghai, China, a city that has experienced significant changes in the past decade. The experimental results show that: (1) the change detection and classification accuracy of the proposed method are 83.30% and 78.77%, respectively, which are significantly better than the values obtained for the multithreshold change vector analysis (MCVA) and the multithreshold change vector analysis and support vector machine (MCVA + SVM) methods; (2) the updated results agree well with GlobeLand30 2010, especially for cultivated land and artificial surfaces, indicating the effectiveness of the proposed method; (3) the most significant changes over the past decade in Shanghai were from cultivated land to artificial surfaces, and the total area containing artificial surfaces in Shanghai increased by about 55% from 2000 to 2011. The factors affecting the updating results are also discussed, which be attributed to the classification accuracy of the base image, extended change vector analysis, and object-based image analysis.