Landsat-8 vs. Sentinel-2: examining the added value of sentinel-2’s red-edge bands to land-use and land-cover mapping in Burkina Faso

The availability of freely available moderate-to-high spatial resolution (10–30 m) satellite imagery received a major boost with the recent launch of the Sentinel-2 sensor by the European Space Agency. Together with Landsat, these sensors provide the scientific community with a wide range of spatial, spectral, and temporal properties. This study compared and explored the synergistic use of Landsat-8 and Sentinel-2 data in mapping land use and land cover (LULC) in rural Burkina Faso. Specifically, contribution of the red-edge bands of Sentinel-2 in improving LULC mapping was examined. Three machine-learning algorithms – random forest, stochastic gradient boosting, and support vector machines – were employed to classify different data configurations. Classification of all Sentinel-2 bands as well as Sentinel-2 bands common to Landsat-8 produced an overall accuracy, that is 5% and 4% better than Landsat-8. The combination of Landsat-8 and Sentinel-2 red-edge bands resulted in a 4% accuracy improvement over that of Landsat-8. It was found that classification of the Sentinel-2 red-edge bands alone produced better and comparable results to Landsat-8 and the other Sentinel-2 bands, respectively. Results of this study demonstrate the added value of the Sentinel-2 red-edge bands and encourage multi-sensoral approaches to LULC mapping in West Africa.     

Updating Landsat-based forest cover maps with MODIS images using multiscale spectral-spatial-temporal superresolution mapping

With the high deforestation rates of global forest covers during the past decades, there is an ever-increasing need to monitor forest covers at both fine spatial and temporal resolutions. Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat series images have been used commonly for satellite-derived forest cover mapping. However, the spatial resolution of MODIS images and the temporal resolution of Landsat images are too coarse to observe forest cover at both fine spatial and temporal resolutions. In this paper, a novel multiscale spectral-spatial-temporal superresolution mapping (MSSTSRM) approach is proposed to update Landsat-based forest maps by integrating current MODIS images with the previous forest maps generated from Landsat image. Both the 240 m MODIS bands and 480 m MODIS bands were used as inputs of the spectral energy function of the MSSTSRM model. The principle of maximal spatial dependence was used as the spatial energy function to make the updated forest map spatially smooth. The temporal energy function was based on a multiscale spatial-temporal dependence model, and considers the land cover changes between the previous and current time. The novel MSSTSRM model was able to update Landsat-based forest maps more accurately, in terms of both visual and quantitative evaluation, than traditional pixel-based classification and the latest sub-pixel based super-resolution mapping methods The results demonstrate the great efficiency and potential of MSSTSRM for updating fine temporal resolution Landsat-based forest maps using MODIS images.     

Measures of spatio-temporal accuracy for time series land cover data

Remote sensing is a useful tool for monitoring changes in land cover over time. The accuracy of such time-series analyses has hitherto only been assessed using confusion matrices. The matrix allows global measures of user, producer and overall accuracies to be generated, but lacks consideration of any spatial aspects of accuracy. It is well known that land cover errors are typically spatially auto-correlated and can have a distinct spatial distribution. As yet little work has considered the temporal dimension and investigated the persistence or errors in both geographic and temporal dimensions. Spatio-temporal errors can have a profound impact on both change detection and on environmental monitoring and modelling activities using land cover data. This study investigated methods for describing the spatio-temporal characteristics of classification accuracy. Annual thematic maps were created using a random forest classification of MODIS data over the Jakarta metropolitan areas for the period of 2001–2013. A logistic geographically weighted model was used to estimate annual spatial measures of user, producer and overall accuracies. A principal component analysis was then used to extract summaries of the multi-temporal accuracy. The results showed how the spatial distribution of user and producer accuracy varied over space and time, and overall spatial variance was confirmed by the principal component analysis. The results indicated that areas of homogeneous land cover were mapped with relatively high accuracy and low variability, and areas of mixed land cover with the opposite characteristics. A multi-temporal spatial approach to accuracy is shown to provide more informative measures of accuracy, allowing map producers and users to evaluate time series thematic maps more comprehensively than a standard confusion matrix approach. The need to identify suitable properties for a temporal kernel are discussed.     

Accuracy assessment of the global forest watch tree cover 2000 in China

Inaccurate information on forest resources could hamper forest conservation, reforestation and sustainable management. Remote-sensing products have emerged as key tools in forest cover monitoring. The Global Forest Watch (GFW) dataset as an interactive remote sensing product, is now applied by more than 2 million users including researchers, conservationists and local communities for analyzing forest cover changes. The quality of this product varies spatially, and local validations are recommended before using the data for inventory and management tasks. Our study evaluated the accuracy and suitability of the GFW dataset for analyzing China’s forest cover. We conducted a validation based on a streamlined visual interpretation procedure using high-resolution optical imagery on Google Earth to map the uncertainties and inaccuracies of GFW Tree Cover 2000 in China. We then estimated China’s forest area after considering the data uncertainty, made a comparison with the data reported by the National Forest Inventory of China (CNFI) to understand where and how the land-based inventory differs from the presence/absence-based remote sensing data. The results showed that the overall accuracy of the GFW Tree Cover 2000 data reached 94.5 %. The user’s and producer’s accuracy of forest classification was 89.26 % and 82.13 %. The sample-based area estimation using GFW showed a larger forest area than the figure reported by CNFI in mainland China, while data discrepancy varied at provincial levels. The study provides a detailed performance assessment of GFW in terms of accuracy of defining forest, and we advise the consideration of data uncertainty in forest cover estimates for future forest management.     

Imputing forest carbon stock estimates from inventory plots to a nationally continuous coverage

The U.S. has been providing national-scale estimates of forest carbon (C) stocks and stock change to meet United Nations Framework Convention on Climate Change (UNFCCC) reporting requirements for years. Although these currently are provided as national estimates by pool and year to meet greenhouse gas monitoring requirements, there is growing need to disaggregate these estimates to finer scales to enable strategic forest management and monitoring activities focused on various ecosystem services such as C storage enhancement. Through application of a nearest-neighbor imputation approach, spatially extant estimates of forest C density were developed for the conterminous U.S. using the U.S.’s annual forest inventory. Results suggest that an existing forest inventory plot imputation approach can be readily modified to provide raster maps of C density across a range of pools (e.g., live tree to soil organic carbon) and spatial scales (e.g., sub-county to biome). Comparisons among imputed maps indicate strong regional differences across C pools. The C density of pools closely related to detrital input (e.g., dead wood) is often highest in forests suffering from recent mortality events such as those in the northern Rocky Mountains (e.g., beetle infestations). In contrast, live tree carbon density is often highest on the highest quality forest sites such as those found in the Pacific Northwest. Validation results suggest strong agreement between the estimates produced from the forest inventory plots and those from the imputed maps, particularly when the C pool is closely associated with the imputation model (e.g., aboveground live biomass and live tree basal area), with weaker agreement for detrital pools (e.g., standing dead trees). Forest inventory imputed plot maps provide an efficient and flexible approach to monitoring diverse C pools at national (e.g., UNFCCC) and regional scales (e.g., Reducing Emissions from Deforestation and Forest Degradation projects) while allowing timely incorporation of empirical data (e.g., annual forest inventory).     

Accurate mapping of forest types using dense seasonal Landsat time-series

An accurate map of forest types is important for proper usage and management of forestry resources. Medium resolution satellite images (e.g., Landsat) have been widely used for forest type mapping because they are able to cover large areas more efficiently than the traditional forest inventory. However, the results of a detailed forest type classification based on these images are still not satisfactory. To improve forest mapping accuracy, this study proposed an operational method to get detailed forest types from dense Landsat time-series incorporating with or without topographic information provided by DEM. This method integrated a feature selection and a training-sample-adding procedure into a hierarchical classification framework. The proposed method has been tested in Vinton County of southeastern Ohio. The detailed forest types include pine forest, oak forest, and mixed-mesophytic forest. The proposed method was trained and validated using ground samples from field plots. The three forest types were classified with an overall accuracy of 90.52% using dense Landsat time-series, while topographic information can only slightly improve the accuracy to 92.63%. Moreover, the comparison between results of using Landsat time-series and a single image reveals that time-series data can largely improve the accuracy of forest type mapping, indicating the importance of phenological information contained in multi-seasonal images for discriminating different forest types. Thanks to zero cost of all input remotely sensed datasets and ease of implementation, this approach has the potential to be applied to map forest types at regional or global scales.     

Comparison and relative quality assessment of the GLC2000, GLOBCOVER,MODIS and ECOCLIMAP land cover data sets at the African continental scale

Land cover dynamics at the African continental scale is of great importance for global change studies. Actually, four satellite-derived land cover maps of Africa now available, e.g. ECOCLIMAP, GLC2000, MODIS and GLOBCOVER, are based on images acquired in the 2000s. This study aims at stressing the compliances and the discrepancies between these four land cover classifications systems. Each of them used different mapping initiatives and relies on different mapping standards, which supports the present investigation. In order to do a relative comparison of the four maps, a preamble was to reconcile their thematic legends into more aggregated categories after a projection into the same spatial resolution. Results show that the agreement between the four land cover products is between 56 and 69%. While all these land cover datasets show a reasonable agreement in terms of surface types and spatial distribution patterns, mapping of heterogeneous landscapes in the four products is not very successful. Land cover products based on remote sensing imagery can indeed significantly be improved by using smarter algorithms, better timing of image acquisition, improved class definitions. Either will help to improve the accuracy of future land cover maps at the African continental scale. Data producers may use the areas of spatial agreement for training area selection while users might need to verify the information in the areas of disagreement using additional data sources.     

Using NDVI Differencing and Temporal Logic to Enhance ISODATA Classification in Urban Environments

This paper presents a new approach to improving land use/cover mapping accuracy in an urban environment. Bi-temporal Landsat TM images (1987 and 1997) were initially classified using the ISODATA method. An NDVI difference image was derived and classified, with each class indicating certain land use/cover changes. Temporal logical reasoning was then performed on the classified NDVI difference map and the initial land use/cover maps. The procedure successfully resolved the confusion between forest clear-cuts/fallow cropland and urban, as well as between forest clear-cuts and cropland. The kappa analysis test led to a Z value of 1.837 with the p-value of 0.026 for the year 1987, and a Z value of 1.924 with the p-value of 0.014 for 1997, indicating significant enhancement at the 95% confidence level.     

Assessing the ecological vulnerability of forest landscape to agricultural frontier expansion in the Central Highlands of Vietnam

Forest conservation in human-dominated tropical landscapes ensures provision of major ecosystem services. However, conservation goals are threatened by growing demands for agricultural products. As the expansion of agricultural frontiers continues to exert increasing pressure on forest cover, it is crucial to provide indicators on forest vulnerability to improve our understanding of forest dynamics and prioritize management actions by local decision-makers. The purpose of this study is to develop a rigorous methodological framework to assess forest ecological vulnerability. We aim at evaluating the potential of remote sensing to characterize forest landscape dynamics in spatial and temporal dimensions. We present an innovative method that spatially integrates current landscape mosaic mapping with 45 years of landscape trajectories using Sentinel-2 and Landsat imagery. We derive indicators of exposure to cropland expansion, sensitivity linked with forest degradation and fragmentation, and forest capacity to respond based on forest landscape composition in Di Linh district in the Central Highlands of Vietnam. We map current forest-agricultural mosaics with high accuracy to assess landscape intensification (kappa index = 0.78). We also map the expansion of the agricultural frontier and highlighted heterogeneous agricultural encroachment on forested areas (kappa index = 0.72-0.93). Finally, we identify degradation and fragmentation trajectories that affect forest cover at different rates and intensity. Combined, these indicators pinpoint hotspots of forest vulnerability. This study provides tailored management responses and levers for action by local decision makers. The accessibility of multi-dimensional remote sensing data and the developed landscape approach open promising perspectives for continuously monitoring agricultural frontiers.     

Capturing ephemeral forest dynamics with hybrid time-series and composite mapping in the Republic of Panama

Measuring and progressing toward international goals of curbing deforestation and improving livelihoods of people who depend on forests requires nuanced understanding of forests and the processes surrounding deforestation and degradation. Despite rapid improvements in Earth Observation technology, monitoring of tropical forests remains hindered by persistent cloud cover, heterogeneous landscapes, long wet seasons, and small and ephemeral clearings masked by rapid growth. A hybrid method is presented that combines elements of both time-series and compositing approaches to best overcome these obstacles to map forest cover and change in the Republic of Panama based on Landsat imagery. The resulting Panama Vegetation-Cover Time-Series (PVCTS) maps depict forest cover in Panama from 1990 to 2016 at 30 m resolution. Acknowledging the fuzzy boundary between forest and non-forest classes, these maps employ a hierarchical classification scheme that reflects the natural process of regeneration and can accommodate different definitions of forest and deforestation. Classification accuracy is 97–98 % between forest/non-forest categories and 76–81 % for deforestation events. The maps show a slight greening of Panama from 1990 to 2016 caused by expansion of young secondary growth. The annual rate of deforestation in mature forest has remained around -0.6 %/yr, although young forests have matured at a similar rate such that there is no net loss of forest. While estimates of total forest cover are similar to official national estimates depending on forest definition, there is little agreement in location of deforestation events.     

Use of local and global maps of forest canopy height and aboveground biomass to enhance local estimates of biomass in miombo woodlands in Tanzania

Field surveys are often a primary source of aboveground biomass (AGB) data, but plot-based estimates of parameters related to AGB are often not sufficiently precise, particularly not in tropical countries. Remotely sensed data may complement field data and thus help to increase the precision of estimates and circumvent some of the problems with missing sample observations in inaccessible areas. Here, we report the results of a study conducted in a 15,867 km² area in the dry miombo woodlands of Tanzania, to quantify the contribution of existing canopy height and biomass maps to improving the precision of canopy height and AGB estimates locally. A local and a global height map and three global biomass maps, and a probability sample of 513 inventory plots were subject to analysis. Model-assisted sampling estimators were used to estimate mean height and AGB across the study area using the original maps and then with the maps calibrated with local inventory plots. Large systematic map errors – positive or negative – were found for all the maps, with systematic errors as great as 60–70 %. After being calibrated locally, the maps contributed substantially to increasing the precision of both mean height and mean AGB estimates, with relative efficiencies (variance of the field-based estimates relative to the variance of the map-assisted estimates) of 1.3–2.7 for the overall estimates. The study, although focused on a relatively small area of dry tropical forests, illustrates the potential strengths and weaknesses of existing global forest height and biomass maps based on remotely sensed data and universal prediction models. Our results suggest that the use of regional or local inventory data for calibration can substantially increase the precision of map-based estimates and their applications in assessing forest carbon stocks for emission reduction programs and policy and financial decisions.     

Monitoring intra and inter annual dynamics of forest degradation from charcoal production in Southern Africa with Sentinel – 2 imagery

The contribution of forest degradation to changes in forest carbon stocks remains poorly quantified and constitutes a main source of uncertainty in the forest carbon budget. Charcoal production is a major source of forest degradation in sub-Saharan Africa. We used multitemporal Sentinel-2 imagery to monitor and quantify forest degradation extent in the main supplying area of a major urban center of southern Africa over a 4-year period. We implemented an indirect approach combining Sentinel-2 imagery to map kiln and field measurements to estimate AGB removals and carbon losses from charcoal production. This work generated 10 m resolution maps of forest degradation extent from charcoal production in the study area at quarterly intervals from 2016–2019. These maps reveal an intense and rapid forest degradation process and expose the spatial and temporal patterns of forest degradation from charcoal production with high detail. The total area under charcoal production over the study period reached 26,647 ha (SD = 320.8) and the forest degradation front advanced 10.5 km in a 4-year period, with an average of 19.4 ha of woodlands degraded daily. By the end of 2019, charcoal production disturbed most mopane stands in the study area and woodland fragmentation increased in 70.4 % of the mopane woodlands. We estimated that charcoal production was responsible for 2,568,761 Mg (SD = 42,130) of aboveground biomass extracted from the forest and 1,284,381 Mg (SD = 21,075) of carbon loss. The magnitude of these figures underlines the relevance of charcoal production as a main cause of forest cover change and remarks the existing uncertainties in the quantification of forest degradation processes. These results illustrate the potential of multitemporal medium resolution imagery to quantify forest degradation in sub-Saharan Africa and improve REDD + Monitoring, Reporting, and Verification systems in compliance with international reporting commitments.     

Evaluating the Performance of Multi-Class and Single-Class Classification Approaches for Mountain Agriculture Extraction Using Time-Series NDVI

Supervised multi-class classification (MCC) approach is widely being used for regional-level land use–land cover (LULC) mapping and monitoring. However, it becomes inefficient if the end user wants to map only one particular class. Therefore, an improved single-class classification (SCC) approach is required for quick and reliable map production purpose. In this regard, the current study attempts to evaluate the performance of MCC and SCC approaches for extracting mountain agriculture area using time-series normalized differential vegetation index (NDVI). At first, samples of eight LULC classes were acquired using Google Earth image, and corresponding temporal signatures (TS) were extracted from time-series NDVI to perform classification using minimum distance to mean (MDM) and spectral angle mapper (i.e., multi-class SAM—MCSAM) under MCC approach. Secondly, under SCC approach, the TS of three agriculture classes (i.e., agriculture, mixed agriculture and plantation) were utilized as a reference to extract agriculture extent using Euclidean distance (ED) and SAM (i.e., single-class SAM—SCSAM) algorithms. The area of all four maps (i.e., MDM—19.77% of total geographical area (TGA), MCSAM—21.07% of TGA, ED—15.23% of TGA, SCSAM—13.85% of TGA) was compared with reference agriculture area (14.54% of TGA) of global land cover product, and SCC-based maps were found to have close agreement. Also, the class-wise detection accuracy was evaluated using random sample point-based error matrix which reveals the better performance of ED-based map than rest three maps in terms of overall accuracy and kappa coefficient.     

Use of local and global maps of forest canopy height and aboveground biomass to enhance local estimates of biomass in miombo woodlands in Tanzania

Field surveys are often a primary source of aboveground biomass (AGB) data, but plot-based estimates of parameters related to AGB are often not sufficiently precise, particularly not in tropical countries. Remotely sensed data may complement field data and thus help to increase the precision of estimates and circumvent some of the problems with missing sample observations in inaccessible areas. Here, we report the results of a study conducted in a 15,867 km² area in the dry miombo woodlands of Tanzania, to quantify the contribution of existing canopy height and biomass maps to improving the precision of canopy height and AGB estimates locally. A local and a global height map and three global biomass maps, and a probability sample of 513 inventory plots were subject to analysis. Model-assisted sampling estimators were used to estimate mean height and AGB across the study area using the original maps and then with the maps calibrated with local inventory plots. Large systematic map errors – positive or negative – were found for all the maps, with systematic errors as great as 60–70 %. The maps contributed nothing or even negatively to the precision of mean height and mean AGB estimates. However, after being calibrated locally, the maps contributed substantially to increasing the precision of both mean height and mean AGB estimates, with relative efficiencies (variance of the field-based estimates relative to the variance of the map-assisted estimates) of 1.3–2.7 for the overall estimates. The study, although focused on a relatively small area of dry tropical forests, illustrates the potential strengths and weaknesses of existing global forest height and biomass maps based on remotely sensed data and universal prediction models. Our results suggest that the use of regional or local inventory data for calibration can substantially increase the precision of map-based estimates and their applications in assessing forest carbon stocks for emission reduction programs and policy and financial decisions.     

Mapping wetland characteristics using temporally dense Sentinel-1 and Sentinel-2 data: A case study in the St. Lucia wetlands,South Africa

Wetlands have been determined as one of the most valuable ecosystems on Earth and are currently being lost at alarming rates. Large-scale monitoring of wetlands is of high importance, but also challenging. The Sentinel-1 and -2 satellite missions for the first time provide radar and optical data at high spatial and temporal detail, and with this a unique opportunity for more accurate wetland mapping from space arises. Recent studies already used Sentinel-1 and -2 data to map specific wetland types or characteristics, but for comprehensive wetland characterisations the potential of the data has not been researched yet. The aim of our research was to study the use of the high-resolution and temporally dense Sentinel-1 and -2 data for wetland mapping in multiple levels of characterisation. The use of the data was assessed by applying Random Forests for multiple classification levels including general wetland delineation, wetland vegetation types and surface water dynamics. The results for the St. Lucia wetlands in South Africa showed that combining Sentinel-1 and -2 led to significantly higher classification accuracies than for using the systems separately. Accuracies were relatively poor for classifications in high-vegetated wetlands, as subcanopy flooding could not be detected with Sentinel-1’s C-band sensors operating in VV/VH mode. When excluding high-vegetated areas, overall accuracies were reached of 88.5% for general wetland delineation, 90.7% for mapping wetland vegetation types and 87.1% for mapping surface water dynamics. Sentinel-2 was particularly of value for general wetland delineation, while Sentinel-1 showed more value for mapping wetland vegetation types. Overlaid maps of all classification levels obtained overall accuracies of 69.1% and 76.4% for classifying ten and seven wetland classes respectively.     

Mediterranean forest species mapping using classification of Hyperion imagery

Regional operational forest species mapping is an active research topic that aims to provide the systematic and updatable information necessary for understanding and monitoring the rapidly changing forest environment. In this study, we investigated the potential of satellite hyperspectral imagery in regional forest species mapping by employing a pixel-based and an object-based nearest neighbour classifier in two different Mediterranean study areas. The overall thematic accuracy of the produced maps was assessed using reference data collected in the field and ranged between 0.72 and 0.83. No approach was found to be superior for the study areas. The McNemar test showed no statistically significant difference at the 95% confidence level in the classification accuracies achieved by the two approaches. Both pixel- and object-based approaches provide useful maps, suggesting that regional forest species mapping from space has much potential.     

Mapping trees in high resolution imagery across large areas using locally variable thresholds guided by medium resolution tree maps

Large area tree maps, important for environmental monitoring and natural resource management, are often based on medium resolution satellite imagery. These data have difficulty in detecting trees in fragmented woodlands, and have significant omission errors in modified agricultural areas. High resolution imagery can better detect these trees, however, as most high resolution imagery is not normalised it is difficult to automate a tree classification method over large areas. The method developed here used an existing medium resolution map derived from either Landsat or SPOT5 satellite imagery to guide the classification of the high resolution imagery. It selected a spatially-variable threshold on the green band, calculated based on the spatially-variable percentage of trees in the existing map of tree cover. The green band proved more consistent at classifying trees across different images than several common band combinations. The method was tested on 0.5 m resolution imagery from airborne digital sensor (ADS) imagery across New South Wales (NSW), Australia using both Landsat and SPOT5 derived tree maps to guide the threshold selection. Accuracy was assessed across 6 large image mosaics revealing a more accurate result when the more accurate tree map from SPOT5 imagery was used. The resulting maps achieved an overall accuracy with 95% confidence intervals of 93% (90–95%), while the overall accuracy of the previous SPOT5 tree map was 87% (86–89%). The method reduced omission errors by mapping more scattered trees, although it did increase commission errors caused by dark pixels from water, building shadows, topographic shadows, and some soils and crops. The method allows trees to be automatically mapped at 5 m resolution from high resolution imagery, provided a medium resolution tree map already exists.     

Comparison of MODIS vegetation continuous field — based forest density maps with IRS-LISS III derived maps

The study compared forest cover maps derived using coarse resolution vegetation continuous fields (MODIS VCF; 500m resolution) with the maps derived from medium resolution (24m; IRS LISS-III) data. The comparison of VCF, per cent tree cover product, for the years 2000 to 2004 with LISS III forest density class maps of 2001 and 2003 was carried out for two sites representing hilly (Uttarakhand) and undulating terrains (Madhya Pradesh). Slicing VCF to corresponding forest crown cover, i.e., 0–10%, 10–40%, 40–70% and >70% produced considerable difference in forest area estimates when compared to original LISS III derived crown cover area. The corresponding value range in VCF for 0–10% of actual forest cover were 0–31% and 0–25% in 2 sites respectively, and the respective limit was consistent at 1–20% when VCF range were sliced with respect to upscaled LISS III at 500m resolution. Similarly, all other class limits were also found through iterative process. These limits were similar, within a site, across five years. Spatial Kappa match between these two data indicated higher match in 40–70% class, and also in undulating site. When compared at same resolution, similar forest area cover estimated with weighted area upscaling gave closest match. The study is useful in knowing the usability and limits of VCF product, and utility of spatial Kappa.