Multiple Remote Sensing Data Sources Research Articles

Global structure graph mapping for multimodal change detection

ABSTRACT Multimodal change detection (MCD) combines multiple remote sensing data sources to realize surface change monitoring, which is essential for disaster evaluation and environmental monitoring. However, due to the ‘incomparable’ features in multimodal data, traditional change detection methods for unimodal (homogenous) data no longer apply. To address this issue, this paper proposes a novel MCD method with global structure graph mapping (GSGM) which extracts the ‘comparable’ structural features between multimodal datasets and constructs a global structure graph (GSG) to express the structure information for each of the multi-temporal images, which are then cross-mapped to the other data domain. The change intensity (CI) is determined by measuring the change of GSGs after mapping and the differences between GSGs and mapped GSGs. The forward and backward CI maps (CIMs) are then fused with the latent low-rank representation method (LLRR), and the change map (CM) is obtained by threshold segmentation. Experiments on five multimodal and four unimodal datasets demonstrate the effectiveness and robustness of the proposed method (source code is made available at https://github.com/rshante0426/GSGM).

Mountain Forest Type Classification Based on One-Dimensional Convolutional Neural Network

Convolutional neural networks (CNNs) have demonstrated their efficacy in remote sensing applications for mountain forest classification. However, two-dimensional convolutional neural networks (2D CNNs) require a significant manual involvement in the visual interpretation to obtain continuous polygon label data. To reduce the errors associated with manual visual interpretation and enhance classification efficiency, it is imperative to explore alternative approaches. In this research, we introduce a novel one-dimensional convolutional neural network (1D CNN) methodology that directly leverages field investigation data as labels for classifying mountain forest types based on multiple remote sensing data sources. The hyperparameters were optimised using an orthogonal table, and the model’s performance was evaluated on Mount Emei of Sichuan Province. Comparative assessments with traditional classification methods, namely, a random forest (RF) and a support vector machine (SVM), revealed superior results obtained by the proposed 1D CNN. Forest type classification using the 1D CNN achieved an impressive overall accuracy (OA) of 97.41% and a kappa coefficient (Kappa) of 0.9673, outperforming the U-Net (OA: 94.45%, Kappa: 0.9239), RF (OA: 88.99%, Kappa: 0.8488), and SVM (OA: 88.79%, Kappa: 0.8476). Moreover, the 1D CNN model was retrained using limited field investigation data from Mount Wawu in Sichuan Province and successfully classified forest types in that region, thereby demonstrating its spatial-scale transferability with an OA of 90.86% and a Kappa of 0.8879. These findings underscore the effectiveness of the proposed 1D CNN in utilising multiple remote sensing data sources for accurate mountain forest type classification. In summary, the introduced 1D CNN presents a novel, efficient, and reliable method for mountain forest type classification, offering substantial contributions to the field.

The uncertain geographic context problem (UGCoP) in measuring people’s exposure to green space using the integrated 3S approach

The “mobility turn” in environmental health studies promoted the integrated 3S (GIS, GPS, and remote sensing) approach in the study of urban residents’ exposure to green space and corresponding health outcomes. However, few studies have examined the uncertainty induced by contextual settings when measuring people’s exposure to green space using the conventional and integrated 3S approaches. In this paper, we compared the differences in green space exposure obtained from different geographic contexts using residence-based and mobility-based methods, multiple spatial resolutions, and buffer zones. We collected 7-day GPS trajectories from 208 participants at the 1-minute temporal resolution in Hong Kong. Entire Hong Kong’s green space was delineated using multiple remote sensing data sources at 3 m, 10 m, and 30 m spatial resolutions. Lastly, the residence-based and mobility-based measurements of exposure to green space were calculated for each participant using 100 m, 300 m, and 500 m buffer zones at three spatial resolutions. Descriptive analyses, t-tests, and logistic regression were employed to examine the influence of different contextual settings on different measurements of green space exposure and their health effects. The results indicate multiple significant disparities. The most striking difference is that mobility-based measurements of exposure to green space are significantly higher than those of residence-based measurements, manifesting the uncertain geographic context problem (UGCoP). For future studies, we suggest using mobility-based measurements of exposure to green space, smaller buffer zones, and finer spatial resolutions, which enable more accurate measurements of green space exposure for the study of green space’s health effects.

Urban Remote Sensing with Spatial Big Data: A Review and Renewed Perspective of Urban Studies in Recent Decades

During the past decades, multiple remote sensing data sources, including nighttime light images, high spatial resolution multispectral satellite images, unmanned drone images, and hyperspectral images, among many others, have provided fresh opportunities to examine the dynamics of urban landscapes. In the meantime, the rapid development of telecommunications and mobile technology, alongside the emergence of online search engines and social media platforms with geotagging technology, has fundamentally changed how human activities and the urban landscape are recorded and depicted. The combination of these two types of data sources results in explosive and mind-blowing discoveries in contemporary urban studies, especially for the purposes of sustainable urban planning and development. Urban scholars are now equipped with abundant data to examine many theoretical arguments that often result from limited and indirect observations and less-than-ideal controlled experiments. For the first time, urban scholars can model, simulate, and predict changes in the urban landscape using real-time data to produce the most realistic results, providing invaluable information for urban planners and governments to aim for a sustainable and healthy urban future. This current study reviews the development, current status, and future trajectory of urban studies facilitated by the advancement of remote sensing and spatial big data analytical technologies. The review attempts to serve as a bridge between the growing “big data” and modern urban study communities.

Geographic Distribution of Desert Locusts in Africa, Asia and Europe Using Multiple Sources of Remote-Sensing Data

In history, every occurrence of a desert locust plague has brought a devastating blow to local agriculture. Analyses of the potential geographic distribution and migration paths of desert locusts can be used to better monitor and provide early warnings about desert locust outbreaks. By using environmental data from multiple remote-sensing data sources, we simulate the potential habitats of desert locusts in Africa, Asia and Europe in this study using a logistic regression model that was developed based on desert locust monitoring records. The logistic regression model showed high accuracy, with an average training area under the curve (AUC) value of 0.84 and a kappa coefficient of 0.75. Our analysis indicated that the temperature and leaf area index (LAI) play important roles in shaping the spatial distribution of desert locusts. A model analysis based on data for six environmental variables over the past 15 years predicted that the potential habitats of desert locust present a periodic movement pattern between 40°N and 30°S latitude. The area of the potential desert locust habitat reached a maximum in July, with a suitable area exceeding 2.77 × 107 km2 and located entirely between 0°N and 40°N in Asia-Europe and Africa. In December, the potential distribution of desert locusts reached its minimum area at 0.68 × 107 km2 and was located between 30°N and 30°S in Asia and Africa. According to the model estimates, desert locust-prone areas are distributed in northern Ethiopia, South Sudan, northwestern Kenya, the southern Arabian Peninsula, the border area between India and Pakistan, and the southern Indian Peninsula. In addition, desert locusts were predicted to migrate from east to west between these areas and in Africa between 10°N and 17°N. Countries in these areas should closely monitor desert locust populations and respond rapidly.

Spatiotemporal Patterns of Ecosystem Service Value Changes and Their Coordination with Economic Development: A Case Study of the Yellow River Basin, China.

By integrating multiple remote sensing data sources this study accurately assesses the spatiotemporal characteristics of changes in ecosystem service values (ESVs) in the Yellow River Basin from 2000 to 2015 through Theil-Sen median trend analysis and the Mann-Kendall test. The stability and continuity of the ESVs were comprehensively characterized using coefficients of variation and the Hurst exponent. The degree of coherence between ESVs and economic growth (represented by gross domestic product GDP) on the same temporal and spatial scales was analyzed using ecological-economic coordination (EEC) models. The results show that (1) from 2001 to 2015 the total ESV and the ESV per unit area in the Yellow River Basin generally showed a U-shaped pattern (decreasing slightly then increasing rapidly). (2) The areas with increasing ESVs made up approximately 55.6% of the total area of the river basin. The areas with a decreasing pattern were mainly in the west and north of the Yellow River Basin. (3) The stability and continuity of the ESVs showed a clustered, compact distribution. (4) The most common level of EEC was slightly uncoordinated followed by slightly coordinated and highly coordinated. The proportion of coordinated areas was relatively higher in cultivated land and the lowest in built-up land.

How Green Is Green? Modeling Urban Greenness Exposure in Environmental Health Research

Exposure to greenness has several health benefits, yet analyses are limited by uncertainty about accuracy of greenness metrics, often derived from different remotely-sensed data sources and using different spatial methods. This research (1) assesses the strengths and weaknesses of multiple greenness data sources and metrics for application in environmental health research, and (2) develops and tests an alternative greenness exposure metric that can be applied worldwide.Methods:We analyzed 5 data sources: Landsat time series, Landsat 8, Sentinel-2, RapidEye, and the Green View Index (GVI), derived from Google Street View. These data sets span various time series, resolutions and costs, and represent aerial and perspective views. We compared the Normalized Difference Vegetation Index (NDVI) using different imagery types and the GVI for various buffer distances around postal codes and examined sensitivity to spatial metrics and data sources. Based on these analyses, we constructed a spatially-weighted greenness metric combining data from Sentinel-2 and the GVI that incorporates neighbourhood street-level and at-home greenness.Results:NDVI showed correlations of between 0.65 and 0.85 among satellite types and demonstrated significant inter-variability. GVI showed low correlations with all other data types (0.25-0.40), suggesting an important new source of greenness data. Metrics were spatially sensitive, particularly at small distances and high resolutions, but lacked temporal sensitivity. Initial analyses suggest that the proposed metric is superior to traditional measures that overestimate neighbourhood greenness exposure and underestimate at-home exposure.Conclusions:This research is the first comparison of multiple remote sensing data sources and metrics, including both aerial and novel perspective views. It presents an alternative greenness exposure metric based on freely accessible data sources that may be applied in public health research internationally.

Assessing the Variability of Corn and Soybean Yields in Central Iowa Using High Spatiotemporal Resolution Multi-Satellite Imagery

The utility of remote sensing data in crop yield modeling has typically been evaluated at the regional or state level using coarse resolution (>250 m) data. The use of medium resolution data (10–100 m) for yield estimation at field scales has been limited due to the low temporal sampling frequency characteristics of these sensors. Temporal sampling at a medium resolution can be significantly improved, however, when multiple remote sensing data sources are used in combination. Furthermore, data fusion approaches have been developed to blend data from different spatial and temporal resolutions. This paper investigates the impacts of improved temporal sampling afforded by multi-source datasets on our ability to explain spatial and temporal variability in crop yields in central Iowa (part of the U.S. Corn Belt). Several metrics derived from vegetation index (VI) time-series were evaluated using Landsat-MODIS fused data from 2001 to 2015 and Landsat-Sentinel2-MODIS fused data from 2016 and 2017. The fused data explained the yield variability better, with a higher coefficient of determination (R2) and a smaller relative mean absolute error than using a single data source alone. In this study area, the best period for the yield prediction for corn and soybean was during the middle of the growing season from day 192 to 236 (early July to late August, 1–3 months before harvest). These findings emphasize the importance of high temporal and spatial resolution remote sensing data in agricultural applications.

Multiple remote sensing data sources to assess spatio-temporal patterns of fire incidence over Campos Amazônicos Savanna Vegetation Enclave (Brazilian Amazon)

Fire activity plays an important role in the past, present and future of Earth system behavior. Monitoring and assessing spatial and temporal fire dynamics have a fundamental relevance in the understanding of ecological processes and the human impacts on different landscapes and multiple spatial scales. This work analyzes the spatio-temporal distribution of burned areas in one of the biggest savanna vegetation enclaves in the southern Brazilian Amazon, from 2000 to 2016, deriving information from multiple remote sensing data sources (Landsat and MODIS surface reflectance, TRMM pluviometry and Vegetation Continuous Field tree cover layers). A fire scars database with 30 m spatial resolution was generated using a Landsat time series. MODIS daily surface reflectance was used for accurate dating of the fire scars. TRMM pluviometry data were analyzed to dynamically establish time limits of the yearly dry season and burning periods. Burned area extent, frequency and recurrence were quantified comparing the results annually/seasonally. Additionally, Vegetation Continuous Field tree cover layers were used to analyze fire incidence over different types of tree cover domains. In the last seventeen years, 1.03millionha were burned within the study area, distributed across 1432 fire occurrences, highlighting 2005, 2010 and 2014 as the most affected years. Middle dry season fires represent 86.21% of the total burned areas and 32.05% of fire occurrences, affecting larger amount of higher density tree surfaces than other burning periods. The results provide new insights into the analysis of burned areas of the neotropical savannas, spatially and statistically reinforcing important aspects linked to the seasonality patterns of fire incidence in this landscape.

ATTRIBUTION AND CHARACTERISATION OF SCLEROPHYLL FORESTED LANDSCAPES OVER LARGE AREAS

Abstract. This paper presents a methodology for the attribution and characterisation of Sclerophyll forested landscapes over large areas. First we define a set of woody vegetation data primitives (e.g. canopy cover, leaf area index (LAI), bole density, canopy height), which are then scaled-up using multiple remote sensing data sources to characterise and extract landscape woody vegetation features. The advantage of this approach is that vegetation landscape features can be described from composites of these data primitives. The proposed data primitives act as building blocks for the re-creation of past woody characterisation schemes as well as allowing for re-compilation to support present and future policy and management and decision making needs. Three main research sites were attributed; representative of different sclerophyll woody vegetated systems (Box Iron-bark forest; Mountain Ash forest; Mixed Species foothills forest). High resolution hyperspectral and full waveform LiDAR data was acquired over the three research sites. At the same time, land management agencies (Victorian Department of Environment, Land Water and Planning) and researchers (RMIT, CRC for Spatial Information and CSIRO) conducted fieldwork to collect structural and functional measurements of vegetation, using traditional forest mensuration transects and plots, terrestrial lidar scanning and high temporal resolution in-situ autonomous laser (VegNet) scanners. Results are presented of: 1) inter-comparisons of LAI estimations made using ground based hemispherical photography, LAI 2200 PCA, CI-110 and terrestrial and airborne laser scanners; 2) canopy height and vertical canopy complexity derived from airborne LiDAR validated using ground observations; and, 3) time-series characterisation of land cover features. 1. Accuracy targets for remotely sensed LAI products to match within ground based estimates are ± 0.5 LAI or a 20% maximum (CEOS/GCOS) with new aspirational targets of 5%). In this research we conducted a total of 67 ground-based method-to-method pairwise comparisons across 11 plots in five sites, incorporating the previously mentioned LAI methods. Out of the 67 comparisons, 29 had an RMSE ≥ 0.5 LAIe. This has important implications for the validation of remotely sensed products since ground based techniques themselves exhibit LAI variations greater than internationally recommended guidelines for satellite product accuracies. 2. Two methods of canopy height derivation are proposed and tested over a large area (4 Million Ha). 99th percentile maximum height achieved a RMSE of 6.6%, whilst 95th percentile dominant height a RMSE = 10.3%. Vertical canopy complexity (i.e. the number of forest layers of strata) was calculated as the local maxima of vegetation density within the LiDAR canopy profile and determined using a cubic spline smoothing of Pgap. This was then validated against in-situ and LiDAR observations of canopy strata with an RMSE 0.39 canopy layers. 3. Preliminary results are presented of landcover characterisation using LandTrendr analysis of Landsat LEDAPS data. kNN is then used to link these features to a dense network of 800 field plots sites.

ATTRIBUTION AND CHARACTERISATION OF SCLEROPHYLL FORESTED LANDSCAPES OVER LARGE AREAS

This paper presents a methodology for the attribution and characterisation of Sclerophyll forested landscapes over large areas. First we define a set of woody vegetation data primitives (e.g. canopy cover, leaf area index (LAI), bole density, canopy height), which are then scaled-up using multiple remote sensing data sources to characterise and extract landscape woody vegetation features. The advantage of this approach is that vegetation landscape features can be described from composites of these data primitives. The proposed data primitives act as building blocks for the re-creation of past woody characterisation schemes as well as allowing for re-compilation to support present and future policy and management and decision making needs. <br><br> Three main research sites were attributed; representative of different sclerophyll woody vegetated systems (Box Iron-bark forest; Mountain Ash forest; Mixed Species foothills forest). High resolution hyperspectral and full waveform LiDAR data was acquired over the three research sites. At the same time, land management agencies (Victorian Department of Environment, Land Water and Planning) and researchers (RMIT, CRC for Spatial Information and CSIRO) conducted fieldwork to collect structural and functional measurements of vegetation, using traditional forest mensuration transects and plots, terrestrial lidar scanning and high temporal resolution in-situ autonomous laser (VegNet) scanners. <br><br> Results are presented of: 1) inter-comparisons of LAI estimations made using ground based hemispherical photography, LAI 2200 PCA, CI-110 and terrestrial and airborne laser scanners; 2) canopy height and vertical canopy complexity derived from airborne LiDAR validated using ground observations; and, 3) time-series characterisation of land cover features. <br><br> 1. Accuracy targets for remotely sensed LAI products to match within ground based estimates are ± 0.5 LAI or a 20% maximum (CEOS/GCOS) with new aspirational targets of 5%). In this research we conducted a total of 67 ground-based method-to-method pairwise comparisons across 11 plots in five sites, incorporating the previously mentioned LAI methods. Out of the 67 comparisons, 29 had an RMSE ≥ 0.5 LAIe. This has important implications for the validation of remotely sensed products since ground based techniques themselves exhibit LAI variations greater than internationally recommended guidelines for satellite product accuracies. <br><br> 2. Two methods of canopy height derivation are proposed and tested over a large area (4 Million Ha). 99th percentile maximum height achieved a RMSE of 6.6%, whilst 95th percentile dominant height a RMSE = 10.3%. Vertical canopy complexity (i.e. the number of forest layers of strata) was calculated as the local maxima of vegetation density within the LiDAR canopy profile and determined using a cubic spline smoothing of Pgap. This was then validated against in-situ and LiDAR observations of canopy strata with an RMSE 0.39 canopy layers. <br><br> 3. Preliminary results are presented of landcover characterisation using LandTrendr analysis of Landsat LEDAPS data. kNN is then used to link these features to a dense network of 800 field plots sites.

Estimation and uncertainty analyses of grassland biomass in Northern China: Comparison of multiple remote sensing data sources and modeling approaches

Accurate estimation of grassland biomass and its dynamics are crucial not only for the biogeochemical dynamics of terrestrial ecosystems, but also for the sustainable use of grassland resources. However, estimations of grassland biomass on large spatial scale usually suffer from large variability and mostly lack quantitative uncertainty analyses. In this study, the spatial grassland biomass estimation and its uncertainty were assessed based on 265 field measurements and remote sensing data across Northern China during 2001–2005. Potential sources of uncertainty, including remote sensing data sources (DATsrc), model forms (MODfrm) and model parameters (biomass allocation, BMallo, e.g. root:shoot ratio), were determined and their relative contribution was quantified. The results showed that the annual grassland biomass in Northern China was 1268.37±180.84Tg (i.e., 532.02±99.71g/m2) during 2001–2005, increasing from western to eastern area, with a mean relative uncertainty of 19.8%. There were distinguishable differences among the uncertainty contributions of three sources (BMallo>DATsrc>MODfrm), which contributed 52%, 27% and 13%, respectively. This study highlighted the need to concern the uncertainty in grassland biomass estimation, especially for the uncertainty related to BMallo.

Characterizing forest species composition using multiple remote sensing data sources and inventory approaches

The purpose of the study was to evaluate tree species composition estimated using combinations of different remotely sensed data with different inventory approaches for a forested area in Norway. Basal area species composition was estimated as both species proportions and main species by using data from airborne laser scanning (ALS) and airborne (multispectral and hyperspectral) imagery as auxiliary information in combination with three different inventory approaches: individual tree crown (ITC) approach; semi-individual tree crown (SITC) approach; and area-based approach (ABA). The main tree species classification obtained an overall accuracy higher than 86% for all ABA alternatives and for the two other inventory approaches (ITC and SITC) when combining ALS and hyperspectral imagery. The correlation between estimated species proportions and species proportions measured in the field was higher for coniferous species than for deciduous species and increased with the spectral resolution used. Especially, the ITC approach provided more accurate information regarding the proportion of deciduous species that occurred only in small proportions in the study area. Furthermore, the species proportion estimates of 83% of the plots deviated from field measured species proportions by two-tenths or less. Thus, species composition could be accurately estimated using the different approaches and the highest levels of accuracy were attained when ALS was used in combination with hyperspectral imagery. The accuracies obtained using the ABA in combination with only ALS data were encouraging for implementation in operational forest inventories.