Multi-temporal High-resolution Images Research Articles

Application of multi-temporal satellite imagery for urban tree species identification

Accurate tree inventories are critical for urban forest management but challenging to obtain, as many urban trees are on private property (backyards, etc.) and are excluded from public inventories. Here, we examined the feasibility of tree species identification in a large heterogenous urban area (>850 km2) by using multi-temporal PlanetScope images (3.2 m resolution, multi-spectral) and inventory data from more than 20,000 ground observations within the urban forest of the Greater Chicago area. Our approach achieved an overall classification accuracy of 0.60 and 0.71 for 18 species and ten genera, respectively, but varied from moderate to high for certain species (0.59–0.92) and genera (0.61–0.91). In particular, we identified key host tree species (Fraxinus americana, F. pennsylvanica, and Acer saccharinum) for two damaging invasive insects, emerald ash borer (EAB, Agrilus planipennis) and Asian longhorn beetle (ALB, Anoplophora glabripennis), with over 0.80 accuracies. In addition, we demonstrated that including images from the autumn months (September–November), either for a single-season model or a combined multiple-season model, improved the identification accuracy of temperate deciduous trees. Further, the high classification accuracy of support vector machine (SVM) over random forest (RF) and neural network (NN) approaches suggests that future work might benefit from comparing multiple classification methods to select the approach that maximizes species classification accuracy. Our study demonstrated the potential for applying multi-temporal high-resolution images in urban tree classification, which can be used for urban forest management at a large spatial scale.

Unsupervised Landslide Detection From Multitemporal High-Resolution Images Based on Progressive Label Upgradation and Cross-Temporal Style Adaption

Unsupervised Landslide Detection From Multitemporal High-Resolution Images Based on Progressive Label Upgradation and Cross-Temporal Style Adaption

Change Detection of Building Objects in High-Resolution Single-Sensor and Multi-Sensor Imagery Considering the Sun and Sensor’s Elevation and Azimuth Angles

Building change detection is a critical field for monitoring artificial structures using high-resolution multitemporal images. However, relief displacement depending on the azimuth and elevation angles of the sensor causes numerous false alarms and misdetections of building changes. Therefore, this study proposes an effective object-based building change detection method that considers azimuth and elevation angles of sensors in high-resolution images. To this end, segmentation images were generated using a multiresolution technique from high-resolution images after which object-based building detection was performed. For detecting building candidates, we calculated feature information that could describe building objects, such as rectangular fit, gray-level co-occurrence matrix (GLCM) homogeneity, and area. Final building detection was then performed considering the location relationship between building objects and their shadows using the Sun’s azimuth angle. Subsequently, building change detection of final building objects was performed based on three methods considering the relationship of the building object properties between the images. First, only overlaying objects between images were considered to detect changes. Second, the size difference between objects according to the sensor’s elevation angle was considered to detect the building changes. Third, the direction between objects according to the sensor’s azimuth angle was analyzed to identify the building changes. To confirm the effectiveness of the proposed object-based building change detection performance, two building density areas were selected as study sites. Site 1 was constructed using a single sensor of KOMPSAT-3 bitemporal images, whereas Site 2 consisted of multi-sensor images of KOMPSAT-3 and unmanned aerial vehicle (UAV). The results from both sites revealed that considering additional shadow information showed more accurate building detection than using feature information only. Furthermore, the results of the three object-based change detections were compared and analyzed according to the characteristics of the study area and the sensors. Accuracy of the proposed object-based change detection results was achieved over the existing building detection methods.

Land Subsidence Induced by Rapid Urbanization in Arid Environments: A Remote Sensing-Based Investigation

The rapid increase in the population of many of the older major cities within the countries of the Saharan-Arabian Desert is steering vast and disorganized urban expansion and in many cases introducing adverse environmental impacts such as soil erosion, rise in groundwater levels, and contamination of shallow aquifers, as well as development of deformational features including land subsidence. Using the rapidly growing city of Riyadh (1992: 467 km2; 2018: 980 km2), the capital of the Kingdom of Saudi Arabia as a test site, we utilized Small Baseline Subset (SBAS) interferometric analyses of 2016 to 2018 Sentinel-1 images together with multi-temporal high-resolution images viewable on Google Earth, GPS, field, land use land cover (LULC), and geological data to assess the distribution and rates of land subsidence and their causal effects. Three main causes of subsidence were identified and assessed: (1) discharge of wastewater effluents from septic systems in newly urbanized areas that lead to an increase in soil moisture, rise in groundwater levels, waterlogging, and wetting and hydrocompaction of dry alluvium loose sediments causing land subsidence (up to −20 mm/y) in wadis and lowlands; (2) the subsurface dissolution of karst formation by wastewater effluents and the collapse of voids and cavities at depth under stresses introduced by heavy construction machinery, causing sagging and land subsidence (up to −5 mm/y); and (3) leveling, compaction, and degradation of municipal and building waste materials in organized landfills and disorganized dump sites that resulted in significant land subsidence (up to −21 mm/y) and differential settling that could jeopardize the stability of structures erected over these sites. Our findings highlight the potential use of the advocated integrated approach to assess the nature and extent of land deformation associated with rapid urban growth in arid lands, and to identify areas most impacted for the purpose of directing and prioritizing remediation efforts.

Present-day gully activity in Sisyphi Cavi, Mars – Flow-like features and block movements

The continuously increasing number of multi-temporal high-resolution images from the surface of Mars offers the possibility for detailed studies of present-day surface activity. In this study we investigated all gullies in the Sisyphi Cavi region (355° E, 71° S) of the south polar region of Mars. This region is influenced by the seasonal deposition of a decimeters-thick translucent slab ice in late autumn/winter and its subsequent sublimation in spring. We mapped all gullies (n = 17.760) and measured their orientations. We also identified gullies with contemporary activity (n = 35) using multi-temporal HiRISE images for martian years (MY) 28 to 34. We observed two different kinds of activity: (1) dark flow-like features, and (2) movement of blocks. For both, sediment was transported from the source region (alcove and/or flanks of channels) down the gully. Using image data from HRSC, CTX, and HiRISE, we monitored the general defrosting of the study region. We also analyzed the maximum daytime surface temperatures of the complete study region based on TES data from MYs 24 to 26. To identify the origin and triggering mechanism of the observed activity, we used: (1) detailed topographic investigations (e.g., slope angles) of two extensively gullied slopes based on two HiRISE-DTMs, (2) identification of small scale displacements with Digital Image Correlation (DIC), and (3) orientation measurements of active gullies and comparison to non-active gullies.We found that for the active gullies studied, activity happens at the end of spring between LS ~ 225° and ~250°. This is consistent with the timing of final stages of defrosting in the region. At this time, some surfaces are already defrosted while others still host the seasonal slab ice cover. For the surfaces with slab ice, dark defrosting spots (and flows, if the surface is inclined) are observed on dark dunes as well as on gully aprons and in gully channels. These spots form when, triggered by basal sublimation generated overpressure, sediment entrained in CO2-gas is transported through cracks within the ice and redeposited onto the frosted surface. We compared and linked both morphologic features (dark dune spots and dark flow-like features) and concluded that these features have comparable or even the same triggering mechanisms. Based on this extensive study, the most plausible mechanism for ongoing gully activity can be divided into two steps: 1) accumulation of material within gully channels via small dry flows on top of the slab ice (comparable to dark dune spots/flows), 2) when a critical mass is reached, the sediment flows down the still frosted gully on top of the sublimating ice or as a mixture of dry material and ice in a catastrophic flow. The triggering factor for the movement of blocks remains unclear, as their timing could not be constrained with the available data. We identified headwall erosion in one gully in the study region, whereas a discrete source could not be identified for the other sites, suggesting multiple failure mechanisms could be active in such gullies. Finally, through volume balance calculations we show that active gullies in Sisyphi Cavi could have been formed within decades to several tens of thousands of MY, but gully-morphology indicates a much-longer period for formation of the entire gully-landform.

Landslides Mapped using Satellite Data in the Western Ghats of India After Excess Rainfall During August 2018

Excess rainfall during August 2018 triggered numerous landslides in the Western Ghats region of India covering the states of Kerala, Karnataka and Tamil Nadu. These landslides caused widespread damage to property, loss of life and adversely affected various land resources. In this article, we present an inventory of landslide prepared from the analysis of multitemporal high-resolution images acquired before and after the rainfall event from Resourcesat-2, WorldView-2, GF-2, SPOT-6 and 7, Pleiades-1, Kompsat-3 and Sentinel-2 Earth observation satellites. A total of 6970 landslides with a cumulative area of 22.6 sq. km were mapped for this rainfall event. Majority of landslides have occurred in Kerala (5191), followed by Karnataka (993) and Tamil Nadu (606). Landslides are mostly debris slide and debris flow type with entrainment along the channels. Results show that landslides (83.2%) are triggered by very high rainfall. Also, very high rainfall has resulted in 14.9% of landslides even though slopes are moderate, mainly in the Kodagu district of Karnataka.

Object-oriented change detection and damage assessment using high-resolution remote sensing images, Tangjiao Landslide, Three Gorges Reservoir, China

This paper presents a new approach for tracking land cover changes and assessing damages caused by individual reservoir landslides using high-resolution images. The object-oriented change detection (OOCD) approach has recently become more popular than traditional pixel-oriented methods for high-resolution image analysis. However, few studies have applied the OOCD approach to the land cover change detection (LCCD) and the damage assessments of individual landslide. An OOCD approach was applied to the multi-temporal high-resolution images taken in 2002, 2005, 2010 and 2013 throughout the Tangjiao Landslide in the TGRA. The proposed OOCD approach contains three major steps: land-cover-type analysis, high-resolution image classification using an object-oriented classification method and LCCD via comparing the classified geographic objects in different temporal images. The object-oriented classification results show that the overall classification accuracies of the 2005, 2010 and 2013 images were greater than 92%, with Kappa Index of Agreement values of at least 89%. The IKONOS image taken in 2002 is an exception to both of these values. Land cover change maps suggest that various damages occurred throughout the study area from 2002 to 2013. Large woodland areas were submerged at the front of the study area. In addition, 21,365 m2 buildings and a 300-m-long road were damaged between 2005 and 2010. Furthermore, a large surface crack was observed from 2010 to 2013. The results suggest that the OOCD approach is able to effectively and accurately identify damage characteristics on the individual landslide. The results also show that the instability of Tangjiao Landslide is controlled by the geological conditions and seriously affected by the reservoir water-level fluctuation and seasonal rainfall.

High Resolution Mapping of Cropping Cycles by Fusion of Landsat and MODIS Data

Multiple cropping, a common practice of intensive agriculture that grows crops multiple times in the agricultural land in one growing season, is an effective way to fulfill the food demand given limited cropland areas. Deriving cropping cycles from satellite data provides the spatial distribution of cropping intensities that allows for monitoring of the multiple cropping activities over large areas. Although efforts have been made to map cropping cycles at 500 m or coarser resolution, producing cropping cycle maps at high resolution remain challenging because data from single satellite sensor do not provide sufficient spatiotemporal observations. In this paper, we generate dense time series of satellite data at 30 m resolution by fusion of Landsat and MODIS data, and derive the cropping cycles from the fused time series data. The method achieves overall accuracies of 92.5% and 89.2%, respectively, for two typical regions of multiple cropping in China using samples identified based on satellite time series data, and an overall accuracy of 81.2% for four subregions using all samples identified based on multi-temporal high resolution images. The mapped crop cycles show to be reasonable geographically and agree with the national census data. The fusion approach provides a feasible way to map cropping cycles at 30 m resolution and enables improved depiction of the spatial distribution of multiple cropping.

A Feasibility Study of Sea Ice Motion and Deformation Measurements Using Multi-Sensor High-Resolution Optical Satellite Images

Sea ice motion and deformation have generally been measured using low-resolution passive microwave or mid-resolution radar remote sensing datasets of daily (or few days) intervals to monitor long-term trends over a wide polar area. This feasibility study presents an application of high-resolution optical images from operational satellites, which have become more available in polar regions, for sea ice motion and deformation measurements. The sea ice motion, i.e., Lagrangian vector, is measured by using a maximum cross-correlation (MCC) technique and multi-temporal high-resolution images acquired on 14–15 August 2014 from multiple spaceborne sensors on board Korea Multi-Purpose Satellites (KOMPSATs) with short acquisition time intervals. The sea ice motion extracted from the six image pairs of the spatial resolutions were resampled to 4 m and 15 m yields with vector length measurements of 57.7 m root mean square error (RMSE) and −11.4 m bias and 60.7 m RMSE and −13.5 m bias, respectively, compared with buoy location records. The errors from both resolutions indicate more accurate measurements than from conventional sea ice motion datasets from passive microwave and radar data in ice and water mixed surface conditions. In the results of sea ice deformation caused by interaction of individual ice floes, while free drift patterns of ice floes were delineated from the 4 m spatial resolution images, the deformation was less revealing in the 15 m spatial resolution image pairs due to emphasized discretization uncertainty from coarser pixel sizes. The results demonstrate that using multi-temporal high-resolution optical satellite images enabled precise image block matching in the melting season, thus this approach could be used for expanding sea ice motion and deformation dataset, with an advantage of frequent image acquisition capability in multiple areas by means of many operational satellites.

Image Fusion-Based Land Cover Change Detection Using Multi-Temporal High-Resolution Satellite Images

Change detection is usually treated as a problem of explicitly detecting land cover transitions in satellite images obtained at different times, and helps with emergency response and government management. This study presents an unsupervised change detection method based on the image fusion of multi-temporal images. The main objective of this study is to improve the accuracy of unsupervised change detection from high-resolution multi-temporal images. Our method effectively reduces change detection errors, since spatial displacement and spectral differences between multi-temporal images are evaluated. To this end, a total of four cross-fused images are generated with multi-temporal images, and the iteratively reweighted multivariate alteration detection (IR-MAD) method—a measure for the spectral distortion of change information—is applied to the fused images. In this experiment, the land cover change maps were extracted using multi-temporal IKONOS-2, WorldView-3, and GF-1 satellite images. The effectiveness of the proposed method compared with other unsupervised change detection methods is demonstrated through experimentation. The proposed method achieved an overall accuracy of 80.51% and 97.87% for cases 1 and 2, respectively. Moreover, the proposed method performed better when differentiating the water area from the vegetation area compared to the existing change detection methods. Although the water area beneath moderate and sparse vegetation canopy was captured, vegetation cover and paved regions of the water body were the main sources of omission error, and commission errors occurred primarily in pixels of mixed land use and along the water body edge. Nevertheless, the proposed method, in conjunction with high-resolution satellite imagery, offers a robust and flexible approach to land cover change mapping that requires no ancillary data for rapid implementation.

OBJECT MANIFOLD ALIGNMENT FOR MULTI-TEMPORAL HIGH RESOLUTION REMOTE SENSING IMAGES CLASSIFICATION

Abstract. Multi-temporal remote sensing images classification is very useful for monitoring the land cover changes. Traditional approaches in this field mainly face to limited labelled samples and spectral drift of image information. With spatial resolution improvement, “pepper and salt” appears and classification results will be effected when the pixelwise classification algorithms are applied to high-resolution satellite images, in which the spatial relationship among the pixels is ignored. For classifying the multi-temporal high resolution images with limited labelled samples, spectral drift and “pepper and salt” problem, an object-based manifold alignment method is proposed. Firstly, multi-temporal multispectral images are cut to superpixels by simple linear iterative clustering (SLIC) respectively. Secondly, some features obtained from superpixels are formed as vector. Thirdly, a majority voting manifold alignment method aiming at solving high resolution problem is proposed and mapping the vector data to alignment space. At last, all the data in the alignment space are classified by using KNN method. Multi-temporal images from different areas or the same area are both considered in this paper. In the experiments, 2 groups of multi-temporal HR images collected by China GF1 and GF2 satellites are used for performance evaluation. Experimental results indicate that the proposed method not only has significantly outperforms than traditional domain adaptation methods in classification accuracy, but also effectively overcome the problem of “pepper and salt”.

Automatic cloud-free image generation from high-resolution multitemporal imagery

The aim of this paper is to document the automatic reconstruction of clouds and their cast shadows for the generation of spontaneous cloud-free images from high-resolution multitemporal images. To apply the proposed technique, a cloud-free reference image, which has the same position as a target image acquired at a different time, is required. First, the cloud region in the target image is detected based on integration of thick and peripheral cloud candidate regions. Next, the detected cloud region is restored using the pixel values of the target image by considering their location relative to the reference images. Finally, the pixel values of the restored image are separately normalized to the values of the reference image to generate a natural-looking cloud-free image. Multitemporal KOMPSAT-2 high-resolution images are used to construct study sites for evaluation of the proposed method in diverse cloud-cover cases. The experimental results show that the proposed method can automatically generate cloud-free images from high-resolution multitemporal images with reasonable qualitative and quantitative performance.

Forest species mapping using airborne hyperspectral APEX data

Abstract The accurate mapping of forest species is a very important task in relation to the increasing need to better understand the role of the forest ecosystem within environmental dynamics. The objective of this paper is the investigation of the potential of a multi-temporal hyperspectral dataset for the production of a thematic map of the dominant species in the Forêt de Hardt (France). Hyperspectral data were collected in June and September 2013 using the Airborne Prism EXperiment (APEX) sensor, covering the visible, near-infrared and shortwave infrared spectral regions with a spatial resolution of 3 m by 3 m. The map was realized by means of a maximum likelihood supervised classification. The classification was first performed separately on images from June and September and then on the two images together. Class discrimination was performed using as input 3 spectral indices computed as ratios between red edge bands and a blue band for each image. The map was validated using a testing set selected on the basis of a random stratified sampling scheme. Results showed that the algorithm performances improved from an overall accuracy of 59.5% and 48% (for the June and September images, respectively) to an overall accuracy of 74.4%, with the producer’s accuracy ranging from 60% to 86% and user’s accuracy ranging from 61% to 90%, when both images (June and September) were combined. This study demonstrates that the use of multi-temporal high-resolution images acquired in two different vegetation development stages (i.e., 17 June 2013 and 4 September 2013) allows accurate (overall accuracy 74.4%) local-scale thematic products to be obtained in an operational way.

Evaluation of Coastline Changes under Human Intervention Using Multi-Temporal High-Resolution Images: A Case Study of the Zhoushan Islands, China

Continued sea-level rise and coastal development have led to considerable concerns on coastline changes along inhabited islands. Analysis of long-term coastline changes of islands is however limited due to unavailable data and the cost of field work. In this study, high-resolution images taken from 1970–2011 at an interval of about 10 years and topographic maps were collected to determine coastline changes and their drivers in the Zhoushan Islands, China. Results show that nearly all inhabited islands appeared to have noteworthy seaward expansion during the past four decades. Coastline change rates varied among islands, and the annual change rate of Zhoushan Island (the main island) reached 12.83 ± 0.17 m/year during the same period. Since 2003, the study area has been dominated by artificial coast. The proportion of harbor/port and urban/industrial coast has significantly increased, while rocky coasts and shelter-farm coasts have shrunk greatly. Preliminary analysis of drivers for these coastline changes across the Zhoushan Islands highlights the roles of human policies during different periods as well as location, which were the dominant factors controlling the great spatial and temporal complexity of coastline changes of the major islands. Sediment supply from the Yangtze River decreased after the completion of the Three Gorges Dam in 2003; however, the Zhoushan coast rapidly accreted seaward during the last decade and the artificial siltation, coastal engineering, and harbor dredging materials could be responsible for the observed coastline changes. Pressured by rapid development of the port industry, the Zhoushan coast may face unprecedented challenges in coastal use in the near future. This research provides the basic background information for future studies on coastal protection and management.

Present-day seasonal gully activity in a south polar pit (Sisyphi Cavi) on Mars

The large amount of multi-temporal high-resolution images acquired in the last few years offers the opportunity to identify morphological changes associated with recent geologic activity on the surface of Mars. In this study we focus on a single gully in Sisyphi Cavi, located in the south polar region at 1.44°E and 68.54°S. The gully incises the gullied equator-facing slope of an isolated polar pit within an infilled impact crater. It is important to notice that the following investigations describe the activity and modifications of an existing gully and not the formation of the gully itself. High-resolution image data analyses show new deposits at the terminus of the gully channel and on the gully apron within spring (after solar longitudes of 236°) of martian years (MY) 29 and 31. Our morphological investigations show that the identified new deposits were formed by dark flows through the entire gully deposited on top of the apron between solar longitudes (LS)∼218° and ∼226°. Thermal data show a temperature increase between LS ∼218° and ∼226°. Near-infrared spectral data show relatively constant band strengths of CO2 ice and H2O ice in this time range. After the formation of the dark flows (after LS∼226°), temperatures increase rapidly from ∼180K to >∼270K at LS∼250°. At this time, spectral data indicate that all volatiles on the surface sublimated. However, an earlier beginning of sublimation when the dark flows were observed (between LS∼218° and ∼226°) is likely, due to the fact that the instruments can only show the last phase of sublimation (decrease of volatile band strengths). Spectral modeling shows that from winter to mid-spring, the surface of the studied area is covered by translucent CO2 slab-ice contaminated by minor amounts of H2O ice and dust. Furthermore, our spectral modeling indicates that the dark material most likely flows on top of the CO2 slab-ice cover. Three different scenarios were proposed to explain the identified dark flows, including (1) flows supported by liquid H2O, (2) flows supported by CO2 gas, and (3) dry flows. On the basis of our study we find that scenario (1) is unlikely because of the very low temperatures. While scenario (2) is consistent with the observed beginning of CO2 ice sublimation in the study area, it is unlikely because of the limitation of the activity to only one gully compared to surrounding gullies that share the same morphologies, slope angles, and volatile contents. Also with scenario (3), dry flows, the activity of only one gully is difficult to explain. Thus, we propose a mixture of scenario (2 and 3), dry flows supported by the ongoing sublimation of CO2 ice within the gully, to be the most plausible scenario, when the observed active gully comprises different source material than the surrounding gullies, i.e., a higher content of probably sand-sized material from outcrops located in the alcove.

Building Change Detection From Multitemporal High-Resolution Remotely Sensed Images Based on a Morphological Building Index

In this study, urban building change detection is investigated, considering that buildings are one of the most dynamic structures in urban areas. To this aim, a novel building change detection approach for multitemporal high-resolution images is proposed based on a recently developed morphological building index (MBI), which is able to automatically indicate the presence of buildings from high-resolution images. In the MBI-based change detection framework, the changed building information is decomposed into MBI, spectral, and shape conditions. A variation of the MBI is a basic condition for the indication of changed buildings. Besides, the spectral information is used as a mask since the change of buildings is primarily related to the spectral variation, and the shape condition is then used as a post-filter to remove irregular structures such as noise and road-like narrow objects. The change detection framework is carried out based on a threshold-based processing at both the feature and decision levels. The advantages of the proposed method are that it does not need any training samples and it is capable of reducing human labor, considering the fact that the current building change detection methods are totally reliant on visual interpretation. The proposed method is evaluated with a QuickBird dataset from 2002 and 2005 covering Hongshan District of Wuhan City, China. The experiments show that the proposed change detection algorithms can achieve satisfactory correctness rates (over 80%) with a low level of total errors (less than 10%), and give better results than the supervised change detection using the support vector machine (SVM).

Land Cover Classification and Information Extraction of Multi-temporal High Resolution Images

Land Cover Classification and Information Extraction of Multi-temporal High Resolution Images

Land Cover Classification of Cloud-Contaminated Multitemporal High-Resolution Images

We show how methods proposed in the statistical community dealing with missing data may be applied for land cover classification, where optical observations are missing due to clouds and snow. The proposed method is divided into two stages: 1) cloud/snow classification and 2) training and land cover classification. The purpose of the cloud/snow classification stage is to determine which pixels are missing due to clouds and snow. All pixels in each optical image are classified into the classes cloud, snow, water, and vegetation using a suitable classifier. The pixels classified as cloud or snow are labeled as missing, and this information is used in the subsequent training and classification stage, which deals with classification of the pixels into various land cover classes. For land cover classification, we apply the maximum-likelihood (assuming normal distributions), -nearest neighbor, and Parzen classifiers, all modified to handle missing features. The classifiers are evaluated on Landsat (both Thematic Mapper and Enhanced Thematic Mapper Plus) images covering a scene at about 900 m a.s.l. in the Hardangervidda mountain plateau in Southern Norway, where 4869 in situ samples of the land cover classes water, ridge, leeside, snowbed, mire, forest, and rock are obtained. The results show that proper modeling of the missing pixels improves the classification rate by 5%-10%, and by using multiple images, we increase the chance of observing the land cover type substantially. The nonparametric classifiers handle nonignorable missing-data mechanisms and are therefore particularly suitable for remote sensing applications where the pixels covered by snow and cloud may depend on the land cover type.