A new forest cover map of continental southeast Asia derived from SPOT‐VEGETATION satellite imagery

Question: Can recent satellite imagery of coarse spatial resolution support forest cover assessment and mapping at the regional level? Location: Continental southeast Asia. Methods: Forest cover mapping was based on digital classification of SPOT4‐VEGETATION satellite images of 1 km spatial resolution from the dry seasons 1998/1999 and 1999/2000. Following a geographical stratification, the spectral clusters were visually assigned to land cover classes. The forest classes were validated by an independent set of maps, derived from interpretation of satellite imagery of high spatial resolution (Landsat TM, 30 m). Forest area estimates from the regional forest cover map were compared to the forest figures of the FAO database. Results: The regional forest cover map displays 12 forest and land cover classes. The mapping of the region's deciduous and fragmented forest cover remained challenging. A high correlation was found between forest area estimates obtained from this map and from the Landsat TM derived maps. The regional and sub‐regional forest area estimates were close to those reported by FAO. Conclusion: SPOT4‐VEGETATION satellite imagery can be used for mapping consistently and uniformly the extent and distribution of the broad forest cover types at the regional scale. The new map can be considered as an update and improvement on existing regional forest cover maps.     

Identifying and quantifying uncertainty and spatial disagreement in the comparison of Global Land Cover for different applications

This paper provides a methodology for comparing global land cover maps that allows for differences in legend definitions between products to be taken into account. The legends of the two maps are first reconciled by creating a legend lookup table that shows how the legends map onto one another. Where there is overlap, the specific definitions for each legend class are used to calculate the degree of overlap between legend classes. In this way, one‐to‐many mappings are accounted for unlike in most methods where the legend definitions are often forced into place. Another advantage over previous map comparison methods is that application‐specific requirements are captured using expert input, whereby the user rates the importance of disagreement between different legend classes based on the needs of the application. This user‐defined matrix in conjunction with the degree of overlap between legend classes is applied on a pixel‐by‐pixel basis to create maps of spatial disagreement and uncertainty. The user can then highlight the areas of highest thematic uncertainty and disagreement between the different land cover maps allowing for areas that require further detailed examination to be readily identified. It would also be possible for several users to input their knowledge into the process, leading to a potentially more robust comparison of land cover products. The methodology of map comparison is illustrated using different land cover products including Global Land Cover 2000 (GLC‐2000) and the MODIS land cover data set. Two diverse applications are provided including the estimation of global forest cover and monitoring of agricultural land. In the case of global forest cover, an example was provided for Columbia, which showed that the MODIS land cover map overestimates forest cover in comparison with the GLC‐2000. The agricultural example, on the other hand, served to illustrate that for Sudan, MODIS tends to underestimate crop areas while GLC‐2000 overestimates them.     

A new land-cover map of Africa for the year 2000

Aim In the framework of the Global Land Cover 2000 (GLC 2000), a land‐cover map of Africa has been produced at a spatial resolution of 1 km using data from four sensors on‐board four different Earth observing satellites. Location The map documents the location and distribution of major vegetation types and non‐vegetated land surface formations for the entire African continent plus Madagascar and the other surrounding islands. Methods The bulk of these data were acquired on a daily basis throughout the year 2000 by the VEGETATION sensor on‐board the SPOT‐4 satellite. The map of vegetation cover has been produced based upon the spectral response and the temporal profile of the vegetation cover. Digital image processing and geographical information systems techniques were employed, together with local knowledge, high resolution imagery and expert consultation, to compile a cartographic map product. Radar data and thermal sensors were also used for specific land‐cover classes. Results A total of 27 land cover categories are documented, which has more thematic classes than previously published land cover maps of Africa contain. Systematic comparison with existing land cover data and 30‐m resolution imagery from Landsat are presented, and the map is also compared with other pan‐continental land cover maps. The map and digital data base are freely available for non‐commercial uses from http://www.gvm.jrc.it/tem/africa/products.htm Main conclusions The map improves our state of knowledge of the land‐cover of Africa and presents the most spatially detailed view yet published at this scale. This first version of the map should provide an important input for regional stratification and planning purposes for natural resources, biodiversity and climate studies.     

Spatial distribution of young forests and carbon fluxes within recent disturbances in Russia

Forest stand age plays a major role in regulating carbon fluxes in boreal and temperate ecosystems. Young boreal forests represent a relatively small but persistent source of carbon to the atmosphere over 30 years after disturbance, while temperate forests switch from a substantial source over the first 10 years to a notable sink until they reach maturity. Russian forests are the largest contiguous forest belt in the world that accounts for 17% of the global forest cover; however, despite its critical role in controlling global carbon cycle, little is known about spatial patterns of young forest distribution across Russia as a whole, particularly before the year 2000. Here, we present a map of young (0–27 years of age) forests, where 12‐ to 27‐year‐old forests were modeled from the single‐date 500 m satellite record and augmented with the 0‐ to 11‐year‐old forest map aggregated from the 30 m resolution contemporary record between 2001 and 2012. The map captures the distribution of forests with the overall accuracy exceeding 85% within three largest bioclimatic vegetation zones (northern, middle, and southern taiga), although mapping accuracy for disturbed classes was generally low (the highest of 31% for user's and producer's accuracy for the 12–27 age class and the maximum of 74% for user's and 32% for producer's accuracy for the 0–11 age class). The results show that 75.5 ± 17.6 Mha (roughly 9%) of Russian forests were younger than 30 years of age at the end of 2012. The majority of these 47 ± 4.7 Mha (62%) were distributed across the middle taiga bioclimatic zone. Based on the published estimates of net ecosystem production (NEP) and the produced map of young forests, this study estimates that young Russian forests represent a total sink of carbon at the rate of 1.26 Tg C yr^?1.     

Mapping global cropland and field size

A new 1 km global IIASA‐IFPRI cropland percentage map for the baseline year 2005 has been developed which integrates a number of individual cropland maps at global to regional to national scales. The individual map products include existing global land cover maps such as GlobCover 2005 and MODIS v.5, regional maps such as AFRICOVER and national maps from mapping agencies and other organizations. The different products are ranked at the national level using crowdsourced data from Geo‐Wiki to create a map that reflects the likelihood of cropland. Calibration with national and subnational crop statistics was then undertaken to distribute the cropland within each country and subnational unit. The new IIASA‐IFPRI cropland product has been validated using very high‐resolution satellite imagery via Geo‐Wiki and has an overall accuracy of 82.4%. It has also been compared with the EarthStat cropland product and shows a lower root mean square error on an independent data set collected from Geo‐Wiki. The first ever global field size map was produced at the same resolution as the IIASA‐IFPRI cropland map based on interpolation of field size data collected via a Geo‐Wiki crowdsourcing campaign. A validation exercise of the global field size map revealed satisfactory agreement with control data, particularly given the relatively modest size of the field size data set used to create the map. Both are critical inputs to global agricultural monitoring in the frame of GEOGLAM and will serve the global land modelling and integrated assessment community, in particular for improving land use models that require baseline cropland information. These products are freely available for downloading from the http://cropland.geo-wiki.org website.     

Continental estimates of forest cover and forest cover changes in the dry ecosystems of Africa between 1990 and 2000

Aim

This study provides regional estimates of forest cover in dry African ecoregions and the changes in forest cover that occurred there between 1990 and 2000, using a systematic sample of medium‐resolution satellite imagery which was processed consistently across the continent.

Location

The study area corresponds to the dry forests and woodlands of Africa between the humid forests and the semi‐arid regions. This area covers the Sudanian and Zambezian ecoregions.

Methods

A systematic sample of 1600 Landsat satellite imagery subsets, each 20 km × 20 km in size, were analysed for two reference years: 1990 and 2000. At each sample site and for both years, dense tree cover, open tree cover, other wooded land and other vegetation cover were identified from the analysis of satellite imagery, which comprised multidate segmentation and automatic classification steps followed by visual control by national forestry experts.

Results

Land cover and land‐cover changes were estimated at continental and ecoregion scales and compared with existing pan‐continental, regional and local studies. The overall accuracy of our land‐cover maps was estimated at 87%. Between 1990 and 2000, 3.3 million hectares (Mha) of dense tree cover, 5.8 Mha of open tree cover and 8.9 Mha of other wooded land were lost, with a further 3.9 Mha degraded from dense to open tree cover. These results are substantially lower than the 34 Mha of forest loss reported in the FAO's 2010 Global Forest Resources Assessment for the same period and area.

Main conclusions

Our method generates the first consistent and robust estimates of forest cover and change in dry Africa with known statistical precision at continental and ecoregion scales. These results reduce the uncertainty regarding vegetation cover and its dynamics in these previously poorly studied ecosystems and provide crucial information for both science and environmental policies.     

Mapping land cover and forest density in Zagros forests of Khuzestan province in Iran: A study based on Sentinel-2, Google Earth and field data

Zagros forests in western Iran have widely been destroyed because of various reasons. This study was performed to provide the land cover and forest density maps in Zagros forests of Khuzestan province using Sentinel-2, Google Earth and field data. The forest boundary in Khuzestan province was digitized in Google Earth. Sentinel-2 satellite images were provided for the study area. One 1:25000 index sheet of Iranian Mapping Organization (IMO) was selected as pilot area in the province. Sentinel-2 image of the pilot area was classified using different supervised classification algorithms to select the best algorithm for land cover mapping in Khuzestan province. In addition, to evaluate the accuracy of Google Earth data, field sampling was performed using random plots in different land covers. Field data of forest plots were applied to investigate tree canopy cover percent (forest density), as well. Classification of Sentinel-2 image in Zagros area of Khuzestan province was done using the best algorithm and the land cover was obtained. The forest density map was also obtained using a linear regression model between tree canopy cover percent (obtained from field plots) and normalized difference vegetation index (NDVI) (obtained from NDVI map). Finally, the accuracy of land cover map was assessed by some square plots on Google Earth. Results demonstrated that support vector machine (SVM) algorithm had the highest accuracy for land cover mapping. Results also showed that Google Earth images had a good accuracy in the Zagros forests of Khuzestan province. Results demonstrated that NDVI has been a good predicator to estimate tree canopy cover in the study area. Based on results, an area of 443,091.22 ha is covered by Zagros forests in Khuzestan province. Results of accuracy assessment of the land cover map showed the good accuracy of this map in Khuzestan province (overall accuracy: 91% and kappa index: 0.83). For optimum management of Zagros forests, it is suggested that the land cover and forest density mapping will be performed using SVM algorithm, NDVI, and Sentinel-2 satellite images in Zagros forests of Khuzestan province in the certain periods.     

Remote sensing of upland vegetation: the potential of high spatial resolution satellite sensors

Aim Traditional methodologies of mapping vegetation, as carried out by ecologists, consist primarily of field surveying or mapping from aerial photography. Previous applications of satellite imagery for this task (e.g. Landsat TM and SPOT HRV) have been unsuccessful, as such imagery proved to have insufficient spatial resolution for mapping vegetation. This paper reports on a study to assess the capabilities of the recently launched remote sensing satellite sensor Ikonos, with improved capabilities, for mapping and monitoring upland vegetation using traditional image classification methods. Location The location is Northumberland National Park, UK. Methods Traditional remote sensing classification methodologies were applied to the Ikonos data and the outputs compared to ground data sets. This enabled an assessment of the value of the improved spatial resolution of satellite imagery for mapping upland vegetation. Post‐classification methods were applied to remove noise and misclassified pixels and to create maps that were more in keeping with the information requirements of the NNPA for current management processes. Results The approach adopted herein for quick and inexpensive land cover mapping was found to be capable of higher accuracy than achieved with previous approaches, highlighting the benefits of remote sensing for providing land cover maps. Main conclusions Ikonos imagery proved to be a useful tool for mapping upland vegetation across large areas and at fine spatial resolution, providing accuracies comparable to traditional mapping methods of ground surveys and aerial photography.     

Assessing large area forest cover products derived from the same imaging source across Victoria,Australia

Forest cover products are an essential tool for land managers and policy makers. They are used at a variety of spatial scales to inform decision‐making and policy across a range of ecosystem drivers and services. This article compares three forest cover products (FCP), all of which were created using Landsat satellite imagery, but using different methodologies and covering different spatial extents that range from global to state. It also explores their use and utility across the state of Victoria, Australia. It asks the question, how interchangeable are the forest cover maps? FCP are also validated against a very high‐resolution reference data set. Overall accuracy was around 89% for the state and national FCP, and 84% for the global FCP. The global map produced the lowest estimate of total forest cover, while estimates obtained by the national and state FCP were similar across the study area. Spatially, differences, however, were apparent. The national forest cover map obtained higher estimates across most of Victoria except in the most arid region which is dominated by low open woodland. While the national and global scale forest cover maps were found to have good diagnostic ability for large area assessment and reporting, their use for land management is not optimal and can lead to gross error.     

Woody vegetation dynamics in the tropical and subtropical Andes from 2001 to 2014: Satellite image interpretation and expert validation

The interactions between climate and land‐use change are dictating the distribution of flora and fauna and reshuffling biotic community composition around the world. Tropical mountains are particularly sensitive because they often have a high human population density, a long history of agriculture, range‐restricted species, and high‐beta diversity due to a steep elevation gradient. Here we evaluated the change in distribution of woody vegetation in the tropical Andes of South America for the period 2001–2014. For the analyses we created annual land‐cover/land‐use maps using MODIS satellite data at 250 m pixel resolution, calculated the cover of woody vegetation (trees and shrubs) in 9,274 hexagons of 115.47 km², and then determined if there was a statistically significant (p < 0.05) 14 year linear trend (positive—forest gain, negative—forest loss) within each hexagon. Of the 1,308 hexagons with significant trends, 36.6% (n = 479) lost forests and 63.4% (n = 829) gained forests. We estimated an overall net gain of ~500,000 ha in woody vegetation. Forest loss dominated the 1,000–1,499 m elevation zone and forest gain dominated above 1,500 m. The most important transitions were forest loss at lower elevations for pastures and croplands, forest gain in abandoned pastures and cropland in mid‐elevation areas, and shrub encroachment into highland grasslands. Expert validation confirmed the observed trends, but some areas of apparent forest gain were associated with new shade coffee, pine, or eucalypt plantations. In addition, after controlling for elevation and country, forest gain was associated with a decline in the rural population. Although we document an overall gain in forest cover, the recent reversal of forest gains in Colombia demonstrates that these coupled natural‐human systems are highly dynamic and there is an urgent need of a regional real‐time land‐use, biodiversity, and ecosystem services monitoring network.     

Accelerating forest loss in Southeast Asian Massif in the 21st century: A case study in Nan Province,Thailand

Farmers are carving a new agricultural frontier from the forests in the Southeast Asian Massif (SAM) in the 21st century, triggering significant environment degradation at the local scale; however, this frontier has been missed by existing global land use and forest loss analyses. In this paper, we chose Thailand's Nan Province, which is located in the geometric center of SAM, as a case study, and combined high resolution forest cover change product with a fine‐scale land cover map to investigate land use dynamics with respect to topography in this region. Our results show that total forest loss in Nan Province during 2001–2016 was 66,072 ha (9.1% of the forest cover in 2000), and that the majority of this lost forest (92%) had been converted into crop (mainly corn) fields by 2017. Annual forest loss is significantly correlated with global corn price (p < 0.01), re‐confirming agricultural expansion as a key driver of forest loss in Nan Province. Along with the increasing global corn price, forest loss in Nan Province has accelerated at a rate of 2,616 ± 730 ha per decade (p < 0.01). Global corn price peaked in 2012, in which year annual forest loss also reached its peak (7,523 ha); since then, the location of forest loss has moved to steeper land at higher elevations. Spatially, forest loss driven by this smallholder agricultural expansion emerges as many small patches that are not recognizable even at a moderate spatial resolution (e.g. 300 m). It explains how existing global land use/cover change products have missed the widespread and rapid forest loss in SAM. It also highlights the importance of high‐resolution observations to evaluate the environmental impacts of agricultural expansion and forest loss in SAM, including, but not limited to, the impacts on the global carbon cycle, regional hydrology, and local environmental degradation.     

Combining land cover mapping of coastal dunes with vegetation analysis

Question: Coastal dune systems are characterized by a natural mosaic that promotes species diversity. This heterogeneity often represents a severe problem for traditional mapping or ground survey techniques. The work presented here proposes to apply a very detailed CORINE land cover map as baseline information for plant community sampling and analysis in a coastal dune landscape. Location: Molise coast, Central Italy. Method: We analysed through an error matrix the coherence between land cover classes and vegetation types identified through a field survey. The CORINE land cover map (scale 1: 5000) of the Molise coast was used with the CORINE legend expanded to a fourth level of detail for natural and semi‐natural areas. Vegetation data were collected following a random stratified sampling design using the CORINE land cover classes as strata. An error matrix was used to compare, on a category‐by‐category basis, the relationship between vegetation types (obtained by cluster analyses of sampling plots) and land cover classes of the same area. Results: The coincidence between both classification approaches is quite good. Only one land cover class shows a very weak agreement with its corresponding vegetation type; this result was interpreted as being related to human disturbance. Conclusions: Since it is based on a standard land cover classification, the proposal has a potential for application to most European coastal systems. This method could represent a first step in the environmental planning of coastal systems.     

Spatial patterns of the Chaco vegetation of central Argentina: Integration of remote sensing and phytosociology

Abstract. In this study we report the first application of Landsat TM imagery to Chaco vegetation studies at a regional scale in Argentina. We produced a map showing 13 clearly differentiated land‐cover types, and described the composition and structure of the plant communities, in an area of almost 42002 km² in central Argentina. The land‐cover map obtained shows that the Chaco vegetation in central Argentina is highly disturbed. In the lowland part of the area the dominant land‐cover types are largely cultural landscapes and substitute shrublands, which have displaced the original Chaco forests, leaving only small isolated remnants generally confined to sites with some kind of constrain for agriculture. The use of TM images and the multivariate analysis of phytosociological data showed a qualified, high accuracy mapping capability for land‐cover types in the Chaco region (ca. 85% overall accuracy). Our results highlight the utility of TM and field data in a subtropical to warm‐temperate region, which is promising where other ancillary data are not available and a rapid acquisition of reliable vegetation data is required, so constituting a starting point for an imperative and more extensive classification and mapping of the endangered Chaco region.     

A large‐area,spatially continuous assessment of land cover map error and its impact on downstream analyses

Land cover maps increasingly underlie research into socioeconomic and environmental patterns and processes, including global change. It is known that map errors impact our understanding of these phenomena, but quantifying these impacts is difficult because many areas lack adequate reference data. We used a highly accurate, high‐resolution map of South African cropland to assess (1) the magnitude of error in several current generation land cover maps, and (2) how these errors propagate in downstream studies. We first quantified pixel‐wise errors in the cropland classes of four widely used land cover maps at resolutions ranging from 1 to 100 km, and then calculated errors in several representative “downstream” (map‐based) analyses, including assessments of vegetative carbon stocks, evapotranspiration, crop production, and household food security. We also evaluated maps’ spatial accuracy based on how precisely they could be used to locate specific landscape features. We found that cropland maps can have substantial biases and poor accuracy at all resolutions (e.g., at 1 km resolution, up to ～45% underestimates of cropland (bias) and nearly 50% mean absolute error (MAE, describing accuracy); at 100 km, up to 15% underestimates and nearly 20% MAE). National‐scale maps derived from higher‐resolution imagery were most accurate, followed by multi‐map fusion products. Constraining mapped values to match survey statistics may be effective at minimizing bias (provided the statistics are accurate). Errors in downstream analyses could be substantially amplified or muted, depending on the values ascribed to cropland‐adjacent covers (e.g., with forest as adjacent cover, carbon map error was 200%–500% greater than in input cropland maps, but ～40% less for sparse cover types). The average locational error was 6 km (600%). These findings provide deeper insight into the causes and potential consequences of land cover map error, and suggest several recommendations for land cover map users.     

A Simple Semi-Automatic Approach for Land Cover Classification from Multispectral Remote Sensing Imagery

Land cover data represent a fundamental data source for various types of scientific research. The classification of land cover based on satellite data is a challenging task, and an efficient classification method is needed. In this study, an automatic scheme is proposed for the classification of land use using multispectral remote sensing images based on change detection and a semi-supervised classifier. The satellite image can be automatically classified using only the prior land cover map and existing images; therefore human involvement is reduced to a minimum, ensuring the operability of the method. The method was tested in the Qingpu District of Shanghai, China. Using Environment Satellite 1(HJ-1) images of 2009 with 30 m spatial resolution, the areas were classified into five main types of land cover based on previous land cover data and spectral features. The results agreed on validation of land cover maps well with a Kappa value of 0.79 and statistical area biases in proportion less than 6%. This study proposed a simple semi-automatic approach for land cover classification by using prior maps with satisfied accuracy, which integrated the accuracy of visual interpretation and performance of automatic classification methods. The method can be used for land cover mapping in areas lacking ground reference information or identifying rapid variation of land cover regions (such as rapid urbanization) with convenience.     

Are CORINE land cover classes reliable proxies of plant species assemblages? A test in Mediterranean forest landscapes

In land cover mapping, the complexity of landscapes is fitted into classes that may limit the recognition of natural variability. In this study, we tested the power of land cover classes (defined on the CORINE land cover classification scheme, a standardized legend set by EU for land cover inventory) to separate different vascular plant assemblages in forest ecosystems. In order to separately identify the role of different sources of inconsistency between land cover classes and species composition, we compared three different inventory processes, based on (i) dominant tree species as observed in the field, (ii) visual interpretation of remotely sensed images and (iii) semi-automatic supervised classification of satellite images. Our results underline that classifying forest ecosystems on the basis of their canopy species produces an over-simplification of habitat variability. Consequently, land cover maps based on non-specialized classification schemes should not be regarded as good proxies for plant biodiversity. If land cover maps are intended to describe and manage landscapes and their associated biodiversity, it is necessary to improve their capacity to represent the complexity of ecosystems.     

Remote sensing imagery for resource inventories in central Africa: The importance of detailed field data

The rate of rain forest clearing throughout central Africa is of national and international interest because it affects both the region's contribution to global warming and impacts the sustainable productive capacity of its natural resource base. The size and inaccessibility of much of central Africa makes remote sensing imagery the most suitable data source for regional land cover mapping and land transformation monitoring. Present image availability is poor. Most regional studies have had to rely on coarse resolution AVHRR 1 km data that fails to detect the small-scale agricultural clearings that are the primary cause of land cover change throughout the region. This study demonstrates that higher spatial resolution Landsat MSS imagery, which comprises the most available, geographically comprehensive and longest time series dataset, is too coarse to map land cover in low population density areas typical of most of central Africa. Furthermore, this study cautions that the use of high resolution imagery without detailed collateral field data on population density and land use practices while generating superficially plausible results, will most probably produce highly inaccurate estimates of land cover and land transformation. Policies for future regional remote sensing surveys of central Africa should focus on acquisition of higher spatial, spectral, and radiometric resolution imagery and must be accompanied by detailed, systematic field data collection.     

Quantifying mortality of tropical rain forest trees using high-spatial-resolution satellite data

Assessment of forest responses to climate change is severely hampered by the limited information on tree death on short temporal and broad spatial scales, particularly in tropical forests. We used 1‐m resolution panchromatic IKONOS and 0.7‐m resolution QuickBird satellite data, acquired in 2000 and 2002, respectively, to evaluate tree death rates at the La Selva Biological Station in old‐growth Tropical Wet Forest in Costa Rica, Central America. Using a calibration factor derived from ground inspection of tree deaths predicted from the images, we calculated a landscape‐scale annual exponential death rate of 2.8%. This corresponds closely to data for all canopy‐level trees in 18 forest inventory plots, each of 0.5 ha, for a mostly‐overlapping 2‐year period (2.8% per year). This study shows that high‐spatial‐resolution satellite data can now be used to measure old‐growth tropical rain forest tree death rates, suggesting many new avenues for tropical forest ecology and global change research.