, but showed improved accuracy in agricultural areas and increased discrimination of small forest patches. Against lidar measurements, the Landsat-based estimates exhibited accuracy slightly less than that of the MODIS VCF (RMSE 016.8% for MODIS-based vs. 17.4% for Landsat-based estimates), but RMSE of Landsat estimates was 3.3 percentage points lower than that of the MODIS data in an agricultural region. The Landsat data retained the saturation artifact of the MODIS VCF at greater than or equal to 80% tree cover but showed greater potential for removal of errors through calibration to lidar, with post-calibration RMSE of 9.4% compared to 13.5% in MODIS estimates. Provided for free download at the Global Land Cover Facility (GLCF) website (www.landcover. org), the 30-m resolution GLCF tree cover dataset is the highest-resolution multitemporal depiction of Earth's tree cover available to the Earth science community.

The science and management of terrestrial ecosystems require accurate, high-resolution mapping of surface water. We produced a global, 30-m-resolution inland surface water dataset with an automated algorithm using Landsat-based surface reflectance estimates, multispectral water and vegetation indices, terrain metrics, and prior coarse-resolution water masks. The dataset identified 3,650,723 km 2 of inland water globally -nearly three quarters of which was located in North America (40.65%) and Asia (32.77%), followed by Europe (9.64%), Africa (8.47%), South America (6.91%), and Oceania (1.57%). Boreal forests contained the largest portion of terrestrial surface water (25.03% of the global total), followed by the nominal 'inland water' biome (16.36%), tundra (15.67%), and temperate broadleaf and mixed forests (13.91%). Agreement with respect to the Moderate-resolution Imaging Spectroradiometer water mask and Landsat-based national land-cover datasets was very high, with commission errors <4% and omission errors <14% relative to each. Most of these were accounted for in the seasonality of water cover, snow and ice, and clouds -effects which were compounded by differences in image acquisition date relative to reference datasets. The Global Land Cover Facility (GLCF) inland surface water dataset is available for open access at the GLCF website (http://www.landcover.org).

Deforestation is a major driver of climate change 1 and the major driver of biodiversity loss 1,2 . Yet the essential baseline for monitoring forest cover-the global area of forests-remains uncertain despite rapid technological advances and international consensus on conserving target extents of ecosystems 3 . Previous satellite-based estimates 4,5 of global forest area range from 32.1 × 10 6 km 2 to 41.4 × 10 6 km 2 . Here, we show that the major reason underlying this discrepancy is ambiguity in the term 'forest'. Each of the >800 o cial definitions 6 that are capable of satellite measurement relies on a criterion of percentage tree cover. This criterion may range from >10% to >30% cover under the United Nations Framework Convention on Climate Change 7 . Applying the range to the first global, high-resolution map of percentage tree cover 8 reveals a discrepancy of 19.3 × 10 6 km 2 , some 13% of Earth's land area. The discrepancy within the tropics alone involves a di erence of 45.2 Gt C of biomass, valued at US$1 trillion. To more e ectively link science and policy to ecosystems, we must now refine forest monitoring, reporting and verification to focus on ecological measurements that are more directly relevant to ecosystem function, to biomass and carbon, and to climate and biodiversity.Forests are the focus of efforts to mitigate harmful ecological and social impacts of land use, including agreements to reduce carbon dioxide emissions from deforestation and forest degradation (REDD+; refs 9-11). The goals are both scientific-to balance regional and global carbon budgets-as well as political, to reduce carbon emissions and stop species extinctions by defining national baselines and managing future anthropogenic change 12 .The Forest Resources Assessments (FRAs) of the United Nations Food and Agriculture Organization (FAO)-the authority for national and global accounting-recorded 40.8 × 10 6 km 2 of forest in 2000, equalling 31% of Earth's land area 13 . The FRAs rely on self-reporting by participating countries, raising concerns about subjectivity and consistency 14-16 . Although estimates from satellite images should provide a more objective base 9 , even these disagree significantly over the amount and distribution of forests worldwide. Figure 1 maps the consensus among eight global satellite data sets over the class 'forest' in or near the year 2000 (Methods). The densely canopied biomes of the tropical, temperate and boreal zones, and the treeless deserts, prairies and tundra show nearperfect agreement across all sources on the presence or absence of forests. Yet the data disagree over the planet's semi-arid savannahs, shrublands and woodlands, and over the northern limits of the boreal forest. Although 102.2 × 10 6 km 2 show perfect consensus among the eight data sets on either the presence or absence of forests, 9.4 × 10 6 km 2 were identified as forest by four out of the eight sources. These sparsely forested regions are the areas of greatest remaining uncertainty.There are two reasons for the uncertaint...

The compilation of global Landsat data-sets and the ever-lowering costs of computing now make it feasible to monitor the Earth's land cover at Landsat resolutions of 30 m. In this article, we describe the methods to create global products of forest cover and cover change at Landsat resolutions. Nevertheless, there are many challenges in ensuring the creation of high-quality products. And we propose various ways in which the challenges can be overcome. Among the challenges are the need for atmospheric correction, incorrect calibration coefficients in some of the data-sets, the different phenologies between compilations, the need for terrain correction, the lack of consistent reference data for training and accuracy assessment, and the need for highly automated characterization and change detection. We propose and evaluate the creation and use of surface reflectance products, improved selection of scenes to reduce phenological differences, terrain illumination correction, automated training selection, and the use of information extraction procedures robust to errors in training data along with several other issues. At several stages we use Moderate Resolution Spectroradiometer data and products to assist our analysis. A global working prototype product of forest cover and forest cover change is included.

Abstract:The water index (WI) is designed to highlight inland water bodies in remotely sensed imagery. The application of WI for water body mapping is mainly based on the thresholding method. However, there are three primary difficulties with this method: (1) inefficient identification of mixed water pixels; (2) confusion of water bodies with background noise; and (3) variation in the threshold values according to the location and time of image acquisitions. Considering that mixed water pixels usually appear in narrow rivers or shallow water at the edge of lakes or wide rivers, an automated method is proposed for extracting rivers and lakes by combining the WI with digital image processing techniques to address the above issues. The data sources are the Landsat TM (Thematic Mapper) and ETM+ (Enhanced Thematic Mapper Plus) images for three representative areas in China. The results were compared with those from existing thresholding methods. The robustness of the new method in combination with different WIs is also assessed. Several metrics, which include the Kappa coefficient, omission and OPEN ACCESSRemote Sens. 2014, 6 5068 commission errors, edge position accuracy and completeness, were calculated to assess the method's performance. The new method generally outperformed the thresholding methods, although the degree of improvement varied among WIs. The advantages and limitations of the proposed method are also discussed.

No abstract

Recent flood events in the Prairie Pothole Region of North America have stimulated interest in modeling water storage capacities of wetlands and their surrounding catchments to facilitate flood mitigation efforts. Accurate estimates of basin storage capacities have been hampered by a lack of high-resolution elevation data. In this paper, we developed a 0.5 m bare-earth model from Light Detection And Ranging (LiDAR) data and, in combination with National Wetlands Inventory data, delineated wetland catchments and their spilling points within a 196 km 2 study area. We then calculated the maximum water storage capacity of individual basins and modeled the connectivity among these basins. When compared to field survey results, catchment and spilling point delineations from the LiDAR bare-earth model captured subtle landscape features very well. Of the 11 modeled spilling points, 10 matched field survey spilling points. The comparison between observed and modeled maximum water storage had an R 2 of 0.87 with mean absolute error of 5564 m 3. Since maximum water storage capacity of basins does not translate into floodwater regulation capability, we further developed a Basin Floodwater Regulation Index. Based upon this index, the absolute and relative water that could be held by wetlands over a landscape could be modeled. This conceptual model of floodwater downstream contribution was demonstrated with water level data from 17 May 2008.

a b s t r a c tLand-cover change detection using satellite remote sensing is largely confined to the era of Landsat satellites, from 1972 to present. However, the Corona, Argon, and Lanyard intelligence satellites operated by the U.S. government between 1960 and 1972 have the potential to provide an important extension of the long-term record of Earth's land surface. Recently declassified, the archive of images recorded by these satellites contains hundreds of thousands of photographs, many of which have very high ground resolution-6-9 ft (1.8-2.7 m) even by today's standards. This paper demonstrates methods for extending the span of forest-cover change analysis from the Landsat-5 and -7 era (1984 to present) to the previous era covered by the Corona archive in two study areas: one area covered predominantly by urban and sub-urban land uses in the eastern US and another area by tropical forest in central Brazil. We describe co-registration of Corona and Landsat images, extraction of texture features from Corona images, classification of Corona and Landsat images, and post-classification change detection based on the resulting thematic dataset. Second-order polynomial transformation of Corona images yielded geometric accuracy relative to Landsat-7 of 18.24 m for the urban area and 29.35 m for the tropical forest study area, generally deemed adequate for pixel-based change detection at Landsat resolution. Classification accuracies were approximately 95% and 96% for forest/non-forest discrimination for the temperate urban and tropical forest study areas, respectively. Texture within 7 Â 7-to 9 Â 9-pixel ($13.0-16.5 m) neighborhoods and within 11 Â 11-pixel ($30 m) neighborhoods were the most informative metrics for forest classification in Corona images in the temperate and tropical study areas, respectively. The trajectory of change from the 1960s to 2000s differed between the two study areas: the average annual forest loss rate in the urban area doubled from 0.68% to 1.9% from the 1960s to the mid-1980s and then decreased during the following decade. In contrast, deforestation in the Brazilian study area continued at a slightly increased pace between the 1960s and 1990s at annual loss rate of 0.62-0.79% and quickly slowed down afterward. This study demonstrates the strong potential of declassified Corona images for detecting historical forest changes in these study regions and suggests increased utility for retrieving a wide range of land cover histories around the world.