Features In Satellite Imagery Research Articles

Multi-sensor satellite imagery reveals spatiotemporal changes in peatland water table after restoration

Remote sensing (RS) has been suggested as a tool to spatially monitor the status of peatland ecosystem functioning after restoration. However, there have been only a few studies in which post-restoration hydrological changes have been quantified with RS-based modelling. To address this gap, we developed an approach to assess post-restoration spatiotemporal changes in the peatland water table (WT) with optical (Sentinel-2 and Landsat 7–9) and radar (Sentinel-1) imagery. We tested the approach in eleven northern boreal peatlands (six restored, and five control sites) impacted by forestry drainage in northern Finland using Google Earth Engine cloud computing capabilities. We constructed a random forest regression model with spatiotemporal field-measured WT data as a dependent variable and satellite imagery features as independent variables. To assess the spatiotemporal changes, we constructed representative maps for situations before and after restoration, separately for early summer high-water and midsummer low-water conditions. To further quantify temporal changes during 2015–2023 and to test their statistical significance, we conducted a bootstrap hypothesis test for the areas near the restoration measures and similar areas in the control sites. The regression model had a relatively good fit and explanatory capacity (overall R2 = 0.71, RMSE = 6.01 cm), while there were notable site-specific variations. The WT maps showed that the post-restoration changes were not uniform and concentrated near the restoration measures. The bootstrap test showed that the WT increased more in the restored areas (4.7–8.8 cm) than in the control areas (0.1–5.2 cm). Our results indicate that restoration impact on surface hydrology can be quantified with multi-sensor satellite imagery and a machine learning approach in treeless peatlands.

Predicting Short-Term Intensity Change in Tropical Cyclones Using a Convolutional Neural Network

Abstract This study details a two-method, machine learning approach to predict current and short-term intensity change in global tropical cyclones (TCs), “D-MINT” and “D-PRINT.” D-MINT and D-PRINT use infrared imagery and environmental scalar predictors, while D-MINT also employs microwave imagery. Results show that current TC intensity estimates from D-MINT and D-PRINT are more skillful than three established intensity estimation methods routinely used by operational forecasters for North Atlantic and eastern and western North Pacific TCs. Short-term intensity predictions are validated against five operational deterministic guidances at 6-, 12-, 18-, and 24-h lead times. D-MINT and D-PRINT are less skillful than NHC and consensus TC intensity predictions in North Atlantic and eastern North Pacific TCs, but are more skillful than the other guidances for at least half of the lead times. In western North Pacific, north Indian Ocean, and Southern Hemisphere TCs, D-MINT is more skillful than the JTWC and other individual TC intensity forecasts for over half of the lead times. When probabilistically predicting TC rapid intensification (RI), D-MINT is more skillful in North Atlantic and western North Pacific TCs than three operationally used RI guidances, but less skillful for yes–no RI forecasts. In addition, this work demonstrates the importance of microwave imagery, as D-MINT is more skillful than D-PRINT. Since D-MINT and D-PRINT are convolutional neural network models interrogating two-dimensional structures within TC satellite imagery, this study also demonstrates that those features can yield better short-term predictions than existing scalar statistics of satellite imagery in operational models. Finally, a diagnostics tool is revealed to aid the attribution of the D-MINT/D-PRINT intensity predictions. Significance Statement This study develops a method to predict current and short-term forecasts of tropical cyclone (TC) intensity using artificial intelligence. The resultant models use a convolutional neural network (CNN) that can identify two-dimensional features in satellite imagery that are indicative of TC intensity and future intensity change. The performance results indicate that in several TC basins, the CNN approach is generally more skillful than alternative satellite-based estimates of TC intensity as well as operational short-term forecasts of deterministic intensity change and of similar skill to probabilistic rapid intensification forecasts.

Deep Learning on Synthetic Data Enables the Automatic Identification of Deficient Forested Windbreaks in the Paraguayan Chaco

The Paraguayan Chaco is one of the most rapidly deforested areas in Latin America, mainly due to cattle ranching. Continuously forested windbreaks between agricultural areas and forest patches within these areas are mandatory to minimise the impact that the legally permitted logging has on the ecosystem. Due to the large area of the Paraguayan Chaco, comprehensive in situ monitoring of the integrity of these landscape elements is almost impossible. Satellite-based remote sensing offers excellent prerequisites for large-scale land cover analyses. However, traditional methods mostly focus on spectral and texture information while dismissing the geometric context of landscape features. Since the contextual information is very important for the identification of windbreak gaps and central forests, a deep learning-based detection of relevant landscape features in satellite imagery could solve the problem. However, deep learning methods require a large amount of labelled training data, which cannot be collected in sufficient quantity in the investigated area. This study presents a methodology to automatically classify gaps in windbreaks and central forest patches using a convolutional neural network (CNN) entirely trained on synthetic imagery. In a two-step approach, we first used a random forest (RF) classifier to derive a binary forest mask from Sentinel-1 and -2 images for the Paraguayan Chaco in 2020 with a spatial resolution of 10 m. We then trained a CNN on a synthetic data set consisting of purely artificial binary images to classify central forest patches and gaps in windbreaks in the forest mask. For both classes, the CNN achieved an F1 value of over 70%. The presented method is among the first to use synthetically generated training images and class labels to classify natural landscape elements in remote sensing imagery and therewith particularly contributes to the research on the detection of natural objects such as windbreaks.

Predicting Rapid Intensification in North Atlantic and Eastern North Pacific Tropical Cyclones Using a Convolutional Neural Network

Abstract This study develops a probabilistic model based on a convolutional neural network to predict rapid intensification (RI) in both North Atlantic and eastern North Pacific tropical cyclones (TCs). Coined “I-RI,” an advantage of using a convolutional neural network to predict RI is that it is designed to learn from spatial fields, like two-dimensional satellite imagery, as well as scalar features. The resulting model RI probability output is validated against two operational RI guidances—an empirical and a deterministic method—to assess skill at predicting RI over 12-, 24-, 36-, 48-, and 72-h lead times. Results indicate that in North Atlantic TCs, AI-RI is more skillful at predicting RI over 12- and 24-h lead times compared to both operational RI guidances. In eastern North Pacific TCs, AI-RI is more skillful than the empirical operational RI guidance at most RI thresholds, but less skillful than the deterministic RI guidance at all thresholds. For TCs north of 15°N, where the deterministic skill was lower, AI-RI was more skillful than the deterministic operational guidance for over half of the RI thresholds. It is also found that AI-RI struggles to reach the higher RI probabilities produced by both of the operational RI guidances in both basins. This work demonstrates that the two-dimensional structures within the satellite imagery of TCs and the evolution of these structures identified using the difference in satellite images, captured by a convolutional neural network, yield better 12–24-h indicators of RI than existing scalar assessments of satellite brightness temperature. Significance Statement The purpose of this study is to develop a method to predict tropical cyclone rapid intensification using artificial intelligence. The developed model uses a convolutional neural network, which can identify features in satellite imagery that are indicative of rapid intensification. The results suggest that, compared with current operational rapid intensification models, a convolutional neural network approach is generally more skillful at predicting rapid intensification.

Towards objective identification and tracking of convective outflow boundaries in next-generation geostationary satellite imagery

Abstract. Sudden wind direction and speed shifts from outflow boundaries (OFBs) associated with deep convection significantly affect weather in the lower troposphere. Specific OFB impacts include rapid variation in wildfire spread rate and direction, the formation of convection, aviation hazards, and degradation of visibility and air quality due to mineral dust aerosol lofting. Despite their recognized importance to operational weather forecasters, OFB characterization (location, timing, intensity, etc.) in numerical models remains challenging. Thus, there remains a need for objective OFB identification algorithms to assist decision support services. With two operational next-generation geostationary satellites now providing coverage over North America, high-temporal- and high-spatial-resolution satellite imagery provides a unique resource for OFB identification. A system is conceptualized here designed around the new capabilities to objectively derive dense mesoscale motion flow fields in the Geostationary Operational Environmental Satellite 16 (GOES-16) imagery via optical flow. OFBs are identified here by isolating linear features in satellite imagery and backtracking them using optical flow to determine if they originated from a deep convection source. This “objective OFB identification” is tested with a case study of an OFB-triggered dust storm over southern Arizona. The results highlight the importance of motion discontinuity preservation, revealing that standard optical flow algorithms used with previous studies underestimate wind speeds when background pixels are included in the computation with cloud targets. The primary source of false alarms is the incorrect identification of line-like features in the initial satellite imagery. Future improvements to this process are described to ultimately provide a fully automated OFB identification algorithm.

Global-scale remote sensing of mine areas and analysis of factors explaining their extent

Mines are composed of features like open cut pits, water storage ponds, milling infrastructure, waste rock dumps, and tailings storage facilities that are often associated with impacts to surrounding areas. The size and location of mine features can be determined from satellite imagery, but to date a systematic analysis of these features across commodities and countries has not been conducted. We created detailed maps of 295 mines producing copper, gold, silver, platinum group elements, molybdenum, lead-zinc, nickel, uranium or diamonds, representing the dominant share of global production of these commodities. The mapping entailed the delineation and classification of 3,736 open pits, waste rock dumps, water ponds, tailings storage facilities, heap leach pads, milling infrastructure and other features, totalling ~3,633 km2. Collectively, our maps highlight that mine areas can be highly heterogeneous in composition and diverse in form, reflecting variations in underlying geology, commodities produced, topography and mining methods. Our study therefore emphasises that distinguishing between specific mine features in satellite imagery may foster more refined assessments of mine-related impacts. We also compiled detailed annual data on the operational characteristics of 129 mines to show via regression analysis that the sum area of a mine's features is mainly explained by its cumulative production volume (cross-validated R2 of 0.73). This suggests that the extent of future mine areas can be estimated with reasonable certainty based on expected total production volume. Our research may inform environmental impact assessments of new mining proposals, or provide land use data for life cycle analyses of mined products.

Intercropping Classification From GF-1 and GF-2 Satellite Imagery Using a Rotation Forest Based on an SVM

Remote sensing has been widely used in vegetation cover research but is rarely used for intercropping area monitoring. To investigate the efficiency of Chinese Gaofen satellite imagery, in this study the GF-1 and GF-2 of Moyu County south of the Tarim Basin were studied. Based on Chinese GF-1 and GF-2 satellite imagery features, this study has developed a comprehensive feature extraction and intercropping classification scheme. Textural features derived from a Gray level co-occurrence matrix (GLCM) and vegetation features derived from multi-temporal GF-1 and GF-2 satellites were introduced and combined into three different groups. The rotation forest method was then adopted based on a Support Vector Machine (RoF-SVM), which offers the advantage of using an SVM algorithm and that boosts the diversity of individual base classifiers by a rotation forest. The combined spectral-textural-multitemporal features achieved the best classification result. The results were compared with those of the maximum likelihood classifier, support vector machine and random forest method. It is shown that the RoF-SVM algorithm for the combined spectral-textural-multitemporal features can effectively classify an intercropping area (overall accuracy of 86.87% and kappa coefficient of 0.78), and the classification result effectively eliminated salt and pepper noise. Furthermore, the GF-1 and GF-2 satellite images combined with spectral, textural, and multi-temporal features can provide sufficient information on vegetation cover located in an extremely complex and diverse intercropping area.

Integration of classification tree analyses and spatial metrics to assess changes in supraglacial lakes in the Karakoram Himalaya

Alpine glacier responses to climate change reveal increases in significant retreat with corresponding increases in the production of glacier meltwater and development of supraglacial lakes. The rate of occurrence and spatial extent of lakes in the Himalaya are difficult to determine because current spectral-based image analyses of glacier surfaces are limited through anisotropic reflectance and lack of high-quality digital elevation models (DEMs). Additionally, the limitations of multivariate classification algorithms to classify glacier features in satellite imagery have led to an increased interest in non-parametric methods, such as classification and regression trees. This article demonstrates the utility of a semi-automated approach that integrates classification tree-based image segmentation and spatial metrics in an object-oriented analysis to differentiate supraglacial lakes from glacier debris, ice, ice-cliffs, and lateral and medial moraines. We used 2000 and 2004 Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) visible and near-infrared (VNIR) and shortwave infrared (SWIR) imagery to characterize and assess supraglacial conditions on the Baltoro Glacier in the Karakoram Himalaya. Input variables for the image segmentation include ASTER VNIR and SWIR spectral bands, indices (e.g. normalized difference water index (NDWI), normalized difference vegetation index (NDVI), and normalized difference snow index (NDSI)), image band ratios (e.g. NIR/red, middle infrared (MIR)/green, and MIR/red), and DEM derivatives. Classification tree analysis was used to generate initial image segments and it was particularly effective in differentiating water from ice and other glacier surface features. The object-oriented analysis included the use of Boolean logic and squared pixel (SqP) spatial metric to refine the classification tree output. The results of classification tree-based image segmentation show that NDWI is the most important single variable for characterizing glacier surface features followed by NDVI, NIR/red ratio, and green and red spectral bands. Lake features extracted from both images show that there were 131 lakes in 2000 as compared to 157 lakes in 2004. In general, there was a significant increase in the planimetric area of these lakes from 2000 to 2004, and we documented the formation of 26 new lakes. It appears that lake-size increments occur mostly in the lower part of the ablation zone, whereas most of the new lakes are formed in the upper part of the ablation zone. The classification tree outputs are intuitive and the data-derived thresholds eliminate commonly subjective visual determination of such threshold values used in image segmentation. The results of this study show that glacier lakes in high-mountain environments can be characterized without topographic information, which is difficult to obtain from a DEM. Also, the semi-automated method described in this article can potentially eliminate the often laborious visual multitemporal analysis of glacier lake surface change, thereby producing consistent and replicable results needed to assess the trends of alpine glacier response to climate change in the Himalaya.

A Study of Enhanced Index-based Built-up Index Based on Landsat TM Imagery

A new method for extraction of built-up land information both in suburban area and in urban district by using Landsat TM/ETM+ imagery is proposed in this paper.Firstly,to suppress the information of bare soil with the middle-infrared(TM7),near-infrared(TM4) and green band(TM2),it is necessary to build a secondary index,which is called normalized difference bareness and built-up index(NDBBI).At the same time,to enhance the information of bare soil from existing indices of bare soil index(BSI)and modified normalized difference water index(MNDWI),another secondary index is built,which is called enhanced bare soil index(EBSI).Finally,the indices of NDBBI,EBSI,SAVI and MNDWI are applied to constructing a new index for delineating built-up land features in satellite imagery,which is called enhanced index-based built-up index(EIBI).The new index(EIBI)can be employed to extract the built-up land information both in suburban area and in urban district.This approach has been successful in Fuzhou and Zhangzhou experimental regions.Built-up features can be extracted objectively and sufficiently with the accuracy above 90%.

A new index for delineating built‐up land features in satellite imagery

A new index derived from existing indices – an index‐based built‐up index (IBI) – is proposed for the rapid extraction of built‐up land features in satellite imagery. The IBI is distinguished from conventional indices by its first‐time use of thematic index‐derived bands to construct an index rather than by using original image bands. The three thematic indices used in constructing the IBI are the soil adjusted vegetation index (SAVI), the modified normalized difference water index (MNDWI) and the normalized difference built‐up index (NDBI). Respectively, these represent the three major urban components of vegetation, water and built‐up land. The new index has been verified using the Landsat ETM+ image of Fuzhou City in southeastern China. The result shows that the IBI can significantly enhance the built‐up land feature while effectively suppressing background noise. A statistical analysis indicates that the IBI possesses a positive correlation with land surface temperature, but negative correlations with the NDVI and the MNDWI.

Variability and forcing of the East Australian Current

The spatial and temporal variability of the East Australian Current (EAC) is investigated using 6 years (1993–1998) of surface geostrophic stream function from an optimal interpolation of altimeter sea surface heights and velocities derived from tracking thermal features in satellite imagery. Variability appears as a series of cyclones and anticyclones propagating southwestward and westward with periods between 90 and 180 days. The behavior of the variability changes over the 6 years. Energy in the mesoscale frequencies moves slowly south and diminishes with more westward propagation in the region where the current separates from the coast. We find no evidence for a consistent forcing of the EAC by mesoscale signals propagating westward from the South Pacific basin. We suggest that the observations are consistent with variability originating between 32°S and 35°S through intrinsic instabilities of the current.

An improved backpropagation neural network for detection of road-like features in satellite imagery

This paper presents an application of backpropagation neural network for the detection of linear structures in remote-sensing images. The purpose of the approach is two-fold. First, to exploit the advantages of a neural network classifier over the tranditional ones. Second, to avoid the strategic phases of enhancement and thresholding. Once the network is learnt, the classification scheme is real-time. Two critical issues in the present approach are the selection of the network architecture and the rate of convergence of learning. Solutions to these two problems are proposed. Experimental results on IRS and SPOT images are presented. Satisfactory classification results have been obtained using the network.

Using neural networks to locate edges and linear features in satellite images

Edges and linear features are manifestations of discontinuities. In geologic applications of satellite imagery, edges and linear features are used to identify faults, lineaments, or lithology changes. Two techniques for detecting these types of features in satellite imagery are skilled human interpretation and mathematical manipulation (e.g. Fourier analysis). The former approach has the advantage of being able to learn, combined with a high degree of fault tolerance. Disadvantages to this approach are that it is labor intensive (slow) and somewhat arbitrary in its decision process. The latter approach is purely mathematical and, when implemented on a computer, is fast and consistent. The disadvantages of the mathematical approach are the lack of both fault tolerance and learning ability. A third approach, applied here, is to use neural networks (NNs). NNs combine the speed and accuracy of computers with the fault tolerance of human beings. NNs are arrays of highly interconnected, simple processing units which excel at pattern recognition. Because of this ability, they are appropriate tools for recognizing edges and linear features in two-dimensional (2-D) data, such as satellite imagery, aeromagnetic, and gravity data. Image interpretation is comprised of two phases. The syntactic phase identifies image primitives such as lines and edges, and the semantic phase resolves the meaning of the lines and edges. This study focused on the first aspect of locating syntactical information in an image. We determined that NNs are of only limited use for detecting linear features, but they are capable of detecting edges at various scales.