SEARCH

SEARCH BY CITATION

Keywords:

  • Land cover;
  • remote sensing;
  • satellite data;
  • operational monitoring;
  • global vegetation;
  • tropical deforestation

Summary

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES

Satellite data provide the basis for geographically referenced global land cover characterization that is internally consistent, repeatable over time, and potentially more reliable than ground-based sources. During the last 20 years considerable research efforts have been devoted to the extraction of land cover information from these data. Only during the last few years have these methods begun to be applied in operational contexts. Such applications have thus far primarily addressed key global change issues such as the global carbon balance. Examples of the successful quasi-operational implementation of remote sensing include NASA's Humid Tropical Landsat Pathfinder project, where high resolution data are being used at subcontinental scales to measure change in the areal extent of tropical rain forests throughout the world, and the Tropical Ecosystem Environment observation by Satellite (TREES) project to assess forest cover in the tropics. At coarser resolutions, a number of land cover products suitable for incorporation in global and regional models have been developed. Alternatives to traditional land cover classifications have also been developed to describe gradients and mosaics in the vegetation more realistically. These land cover products offer the possibility for applications in ecological and human dimensions research at regional and global scales, as well as for implementation of international agreements that require land cover information. Recently launched and future satellites will carry sensors that provide data with greatly improved capabilities for land cover characterization and advancements in computing environments make it feasible to take advantage of these new data. However, several challenges must be overcome in making a transition from research to operational land cover monitoring, including automation of methods to analyse the satellite data, more effective techniques for validation, and assurance of long-term continuity in the availability of satellite measurements.


INTRODUCTION

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES

There is hardly an interaction between the atmosphere, biosphere, and humans that is not affected by the type of land cover present on the Earth's surface. The Earth's vegetation is a major component of the global carbon cycle. For example, land cover change is a source of carbon dioxide to the atmosphere in the case of deforestation and a sink in the case of regenerating forest. Currently, land cover changes from human disturbance are a net source of carbon emissions to the atmosphere (Houghton, 1995; Houghton, 1999). Land cover affects exchanges of energy, water, and momentum between the biosphere and atmosphere, as has been demonstrated by experiments which quantify the effects of changes in land cover on climate (Shukla et al., 1990; Chase, 1995; Bonan, 1997). Apart from the role of land cover in the Earth's climate, land cover controls the flow of water through the terrestrial portion of the hydrologic cycle, maintains soil productivity, and provides food and fibre for human populations.

Despite the importance of accurate depiction of land cover for understanding the Earth system, knowledge of the geographical distribution of land cover types over the Earth's surface is far from complete. A comparison of land cover data sets from ground-based sources (DeFries & Townshend, 1994a) revealed substantial disagreements among them (Fig. 1). These disagreements result from varying definitions of cover types, inconsistent interpretations of land cover definitions, confusion between natural or potential vegetation and existing vegetation modified by human activity, and actual disagreements about the geographical coverage of land cover types.

image

Figure 1. Estimates of global land areas for fifteen major land cover classes derived from ground-based information (Matthews (1983), Olson et al. (1983) and Wilson & Henderson-Sellers (1985)). The land area where the three data sets agree that the same land cover class is presnt is shown by the dark bar. Figure adapted from DeFries & Townshend (1994a).

Download figure to PowerPoint

Because of these disagreements, remote sensing has increasingly been used as a source of information for characterizing land cover and land cover change at regional and global scales (Skole & Tucker, 1993; Loveland & Belward, 1997; DeFries et al., 1998; Mayaux et al., 1998). Satellite data provide global coverage, internal consistency, and repeatability that is only possible with this synoptic view of the earth. These advances suggest the possibility of applying satellite data to routine monitoring of land cover. This paper discusses both the challenges and the opportunities for moving from research to operational monitoring of global land cover with satellite data.

PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES

Over the past 20 years, both the use of satellite data and the sophistication of the methods for inferring land cover characteristics from satellite data have increased many fold. Satellite data are now a primary source of information for both static depictions of land cover and identification of land cover change.

Static depictions of global and regional land cover types

Static depictions of land cover are required as input to Earth system models and have a wide range of applications for land management. Satellite data have long been used to characterize land cover, following the launch of the first Landsat in 1972 (e.g. Swain & Davis, 1978). The use of satellite data for depicting land cover over large areas began with regional studies in Africa (Tucker et al., 1985) and South America (Townshend et al., 1987). These early works laid the basis for characterizing land cover type by its phenology, as represented in the annual time series of the Normalized Difference Vegetation Index (NDVI). (NDVI is the difference of the near infrared and the red reflectances divided by the sum of the two. It is positively related to photosynthetic activity (Sellers, 1985) due to the high reflectance of healthy green vegetation in the infrared portion of the electromagnetic spectrum relative to unvegetated surfaces). With this approach, change in the NDVI profile over an annual cycle is used as the characteristic signature for each land cover type. Each pixel is classified as a cover type based on its NDVI profile and the signature it most closely matches.

These early works laid the foundation for depiction of global land cover from the annual time series of NDVI at a coarse one-by-one degree resolution (DeFries & Townshend, 1994b), based on data acquired by the Advanced Very High Resolution Radiometer (AVHRR) on board the NOAA series of meteorological satellites (Los et al., 1994). The coarse resolution land cover map was intended primarily for application in global biosphere-atmosphere models. As global AVHRR data became available at finer resolutions (Eidenshink & Faudeen, 1994; James & Kalluri, 1994), it became possible to derive global land cover characterizations at 8 km spatial resolution (DeFries et al., 1998) and 1 km resolution (Loveland & Belward, 1997; Hansen et al., in press). These finer resolution land cover data sets are applicable to a wider range of regional models and are at a spatial scale fine enough to be used for addressing land management issues and studying global change processes (Fig. 2). Along with the development of global land cover data sets at finer spatial resolutions, methodologies have been applied to take advantage of individual reflectance and temperature bands (DeFries et al., 1995a; Lambin & Ehlirch, 1996), the fusion of optical and radar data (Mayaux et al., 1998) and more sophisticated classification algorithms (Gopal et al., 1996; Hansen et al., 1996; Friedl & Brodley, 1997).

image

Figure 2. (opposite) Global land cover maps derived at one-by-one spatial resolution (≈10,000 km2) from DeFries & Townshend (1994b) (Fig. 2a), at 8 km resolution from DeFries et al. (1998) (Fig. 2b), and at 1 km resolution from Hansen et al. (in press) (Fig. 2c). Cover type codes are: 1=needleleaf evergreen forest, 2=evergrean broadleaf forst, 3=deciduous needleleaf forest, 4=deciduous broadleaf forest, 5=mixed forest, 6=woodland, 7=wooded grassland, 8=closed shrubland, 9=open shrubland, 10=grassland, 11=crops, 12=bare, 13=tundra.

Download figure to PowerPoint

Regionally, satellite data have increasingly been applied to map land cover, for example in Europe (Mucher et al., in press), the humid tropics (Townshend et al., 1995; Mayaux et al., 1998), and Canada (Cihlar & Beaubien, 1997). Backscatter from active radar sensors has also been used to obtain information on the structure and density of vegetation, particularly in the tropics where wavelengths in the portion of the electromagnetic spectrum measured by optical sensors cannot penetrate clouds and hazy atmospheric conditions (Saatchi et al., 1997).

Sub-pixel land cover characterization

The most common approach for describing the global distribution of vegetation is to categorize the land surface into a discrete number of vegetation types (DeFries & Townshend, 1994b; Loveland & Belward, 1997; DeFries et al., 1998). Currently, global climate and biogeochemical models estimate variables used in the models on the basis of discrete land cover types. This approach, by definition, cannot capture mosaics and gradients that occur over much of the landscape and often produces artificial boundaries between cover types (DeFries et al., 1995b). For this reason, several methods have been explored for characterizing sub-pixel proportions of land cover characteristics from satellite data.

(i) End-member linear unmixing. Proportions of vegetation types are estimated based on the assumption that the spectral signature is a linear combination of reflectances from the components within the pixel. Implementation of this method requires knowledge of reflectances of ‘pure pixels’, called end-members, from spectral libraries, field measurements, or high resolution data. This approach has been applied at regional and global scales (Pech et al., 1986; Bierwirth, 1990; Quarmby et al., 1992; Settle & Drake, 1993; Adams et al., 1995; DeFries, Townshend & Hansen 1999). Recently, methods have been applied to incorporate nonlinear mixtures (Foody et al., 1997) and multiple end-members to include varying end-member values in different portions of the scene (Roberts et al., 1998).

(ii) Spectral regressions. This approach is based on empirical relationships derived from co-registration between fine resolution (e.g. Landsat Thematic Mapper) and coarse resolution (e.g. AVHRR) data. Regression equations are derived to describe the relationships between percentage forest cover and spectral measures. The regression equations are then used to extrapolate over larger regions to estimate percentage forest density (Iverson et al., 1989; Zhu & Evans, 1992; Iverson et al., 1994; Zhu & Evans, 1994; DeFries et al., 1997).

(iii) Calibration of areal estimates from spatial aggregation of classifications derived from coarse resolution data. Effects of spatial aggregation can lead to inaccurate areal estimates of forest cover from coarse resolution classifications. To compensate, the aerial estimates determined from coarse resolution data can be adjusted according to the spatial arrangement of land cover determined by fine resolution data (Mayaux & Lambin, 1997; Moody, 1998).

The characterization of sub-pixel proportions provides a flexible alternative to land cover classification schemes with discrete numbers of land cover classes. Users of this type of information can either aggregate the layers to produce a map of land cover classes based on their own definitions of canopy density or use the proportional estimates directly in modelling or in other applications.

Methods for deriving sub-pixel land cover proportions have been successfully applied at the global (DeFries et al., 1999) and regional scales (Zhu & Evans, 1994). For example, Landsat imagery and multitemporal AVHRR data were used in a linear mixture model to obtain estimates of proportional cover of woody vegetation, herbaceous vegetation, and bare ground (Fig. 3). Validation of the results is, however, particularly challenging and a comparison of methods would be required for operational implementation.

image

Figure 3. Sub-pixel proportions of (a) woody vegetation, (b) herbaceous vegetation, and (c) bare ground derived from AVHRR 1 km data for 1992-3. Methodology for deriving the maps is described in DeFries et al. (1999). Values range from 0 to 100%, indicating proportional aerial coverage of woody vegetation, herbaceous vegetation, or bare ground.

Download figure to PowerPoint

Identification of land cover change from satellite data

Accurate knowledge of rates and locations of land cover change is crucial for a wide range of purposes, which include understanding the carbon budget, studying the socio-economic drivers of land use change, and determining prudent land management practices. Because repeated coverage is often not feasible with ground measurements over large spatial areas, satellite data usually provide the only practical means to identify routinely where land cover change is occurring.

Land cover can change over time as a result of interannual variability in climate, disturbances such as fire, or anthropogenic land use change such as deforestation and urbanization. Data requirements and methods for identifying land cover change vary with the type of change. In the case of anthropogenic land cover change, human activities generally create patchy landscapes which are only captured at fine spatial resolutions, typically less than 250 m (Townshend & Justice, 1988). In the case of broader-scale land cover changes due to climatic variability, changes in the temporal NDVI profile of coarser resolution data can be used to characterize the land cover change (Lambin & Strahler, 1994).

A prototype for operational monitoring of land cover using satellite data is the Landsat Pathfinder Project on Deforestation in the Humid Tropics (Skole & Tucker, 1993; Townshend et al., 1995), which has mapped forest cover throughout the humid tropics for three epochs (1970s, 80s and 90s) using complete spatial coverage of Landsat data. As an example, Fig. 4 shows a window from a series of Landsat scenes acquired in 1975, 1984, 1992 and 1996 around Santa Cruz, Bolivia. Analysis done by the Landsat Pathfinder Project illustrates the rapid rate of land cover conversion from human activities. Similar analyses were carried out on a mosaic of Landsat scenes with complete spatial coverage to obtain country-wide estimates of deforestation and regrowth (Fig. 4). To obtain reliable estimates of deforestation rates, experience with the Landsat Pathfinder Project indicates that the nonrandom patterns resulting from human disturbance of the landscape make it necessary to analyse a mosaic of Landsat scenes comprising complete spatial coverage rather than sampling a subset. Other lessons learnt from this project are: (1) the need for acquisition strategies of satellite missions to ensure adequate spatial and temporal coverage; (2) the inadequacy of current search and ordering mechanisms when dealing with very large numbers of images; and (3) the fact that a successful methodology for analysing a large number of Landsat scenes combines automated classification approaches and visual inspection of the images (Townshend et al., 1997).

image

Figure 4. Time series showing deforestation in southern Bolivia derived from Landsat data. The Landsat data was classified with a maximum likelihood classifier in combination with visual interpretation following procedures of the NASA Lansat Pathfinder Project on Humid Tropical Deforestation (http://www.inform.umd.edu.tropical/main/html). The centre of the window shown is ≈17°21′0″S, 62°10′48″W.

Download figure to PowerPoint

The Tropical Ecosystem Environment Observation by Satellite (TREES) project provides another prototype for operational monitoring of land cover from satellites. The project aims to update annually coarse resolution forest cover maps to identify ‘deforestation hot spots’ in the humid tropics. Currently, the maps have been derived from AVHRR data, although in the future it is proposed to use data from other sensors, including radar data that are not limited by the clouds and haze that is present for much of the year in the humid tropics (Achard & Estreguil, 1995; Mayaux et al., 1998).

OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES

While the data availability and methods for characterizing land cover and land cover change from satellite data have greatly advanced in the past decade, there are still many challenges in moving from research to operational implementation of routine land cover monitoring. Among these are the issues of validation procedures for land cover products, automation of methods for operational implementation, assured continuity of satellite data in the future, and consensus on how land cover should be characterized to maximize usefulness for a wide range of applications.

Validation of land cover characterizations and land cover change

A particular challenge in using remote sensing data to characterize land cover and land cover change is the validation of global and regional products. Field data are only available sparsely, and it is not feasible to collect ground-based data from a large enough portion of the land surface to be statistically representative. Validating land cover change products is particularly challenging because independent records of where changes have occurred usually do not exist.

High resolution satellite data can be used to assess coarser resolution land cover products (Belward, 1996), although the high resolution data are themselves subject to interpretation and possible errors. Collaboration between researchers in the field and those generating the land cover products from remote sensing is needed to meet this challenge.

Automated methods for operational implementation

Many of the methods used to derive land cover products and detect land cover change from remotely sensed data rely on a substantial degree of human analysis and interpretation, either to label clusters derived in an unsupervised classification approach (Loveland & Belward, 1997), to derive training data in a supervised approach (DeFries et al., 1998), or to interpret visually high resolution images (Skole & Tucker, 1993; Townshend et al., 1995). In an operational framework, methods that require a great deal of human intervention will not be suitable for obtaining real-time results. An effort is needed to automate these processes as much as possible through, for example, the application of machine learning techniques (Friedl & Brodley, 1997). While this goal is clearly not achievable with currently available sensors, the availability of data from future sensors with improved calibration, atmospheric correction, and geometric registration hold promise for more automated analyses that will make operational land cover analysis achievable. Data from a number of sensors (Table 1), including the Earth Observing System's Moderate Resolution Imaging Spectrometer (MODIS) (Justice et al., 1998), will have many of these qualities. In addition, methodological advances for fusing spectral, spatial, temporal, and ancillary data in algorithms to characterize land cover (Kimes et al., 1998) should lead to more accurate results from an automated procedure.

Table 1.  Sensors for monitoring global land cover that are currently operating or planned for the near future. This list is not intended to be comprehensive. An asterisk indicates that the platform is operational as of December 1998. Thumbnail image of
Continuity of satellite data

The main advantage of remote sensing data is that they provide a long-term record of internally consistent measurements. This potential cannot be achieved without the continuity of satellites to collect the data and long-term archives to store them. To date, the AVHRR, which was launched for meteorological and not land cover applications, is the only source of continuous global data with appropriate spectral responses and with a length of record more than a few years (James & Kalluri, 1994). Landsat Thematic Mapper (™) and Multispectral Scanner System (MSS) and SPOT High Resolution Visible (HRV) data have been acquired for many years and can be used locally and regionally to monitor land cover.

In practice, satellite-borne sensors used for land cover characterization have not provided a sufficiently reliable stream of data. For example, Landsat data acquisitions have been, for several reasons, much less spatially comprehensive in the 1990s than in the early 1970s when the series began (Justice & Townshend, 1994). The AVHRR sensor's record has also been unsatisfactory in many respects due to inadequate calibration and severe orbital drift for a number of the NOAA platforms (Privette et al., 1995).

In the future, many new sensors will be launched (see above). Long-term continuity of measurements and adequate provision for long-term archives are imperative for operational land cover monitoring.

Definitions of products appropriate for a wide range of applications

To date, most applications of remote sensing for land cover characterization have been related to global climate and biogeochemical models or, in the case of the Landsat Pathfinder project in the humid tropics, to quantifying deforestation for purposes of understanding the carbon budget. For these applications, only a relatively small number of land cover characteristics are required (Running et al., 1994). Operational use of land cover information for land management and for implementation of international agreements such as the Kyoto Protocol, which requires an accounting of sources and sinks of terrestrial carbon (IGBP, 1998), is only beginning to be developed. The required land cover characteristics are likely to vary substantially for different applications. A hierarchical approach will almost certainly be required to satisfy many diverse requirements. The beginnings of operational definitions of requirements can be found in the report of the Global Climate Observing System (GCOS, 1997).

OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES

Improved sensors and acquisition strategies

A number of new sensors have recently been launched or will be launched in the near future (Table 1). Many of these sensors have been designed specifically for the purpose of studying land cover. The sensors listed in Table 1 will provide a powerful suite of measurements which will vastly improve abilities to characterize land cover from satellite data.

The acquisition strategies for these sensors are crucial if the data are to be used for operational land cover monitoring, in particular because a monitoring system will require integration of coarse and fine resolution data from different sensors. The proposed acquisition strategy of Landsat 7 is intended to ensure as near to complete global coverage as possible, with a depiction of seasonal variability (Goward et al., 1996; Goward & Williams, 1997).

Very high resolution systems have been proposed by several commercial organizations. These will have resolutions between 1 m and 5 m and will have considerable value in the detection and monitoring of areas with variability at high spatial frequency, such as intensively farmed areas and most settlement types. These data will be of considerable help in assigning land use information to the land cover characterizations available from coarser resolution data from systems such as Landsat and SPOT. One example of a high resolution commercial system is the Space Imaging-EOSAT 1KONOS system, which will provide images at 4 m resolution in four multispectral bands and 1 m resolution in a single panchromatic band (http://www/spaceimage.com). The data can be used to create digital elevation models and image products in which spatial displacements due to topography have been removed.

High performance computing environment

The advancement of the computing environment provides the possibility for analysing the vast quantities of data that will be acquired by the array of sensors collecting land surface data. The computational and storage requirements for deriving regional and global land cover data from multi-sensor and multi-temporal data sets far exceed what is currently possible with a single workstation. Efforts to develop high performance software and processing, indexing, and organizing of satellite data are underway (e.g. Kalluri et al., in press) and hold promise for operational use of satellite data for land cover monitoring.

International attention on land cover as a key component of global change

Inclusion of land cover change in international agreements such as the Kyoto Protocol under the United Nations Framework Convention on Climate Change, makes it essential to develop an international operational program to monitor land cover effectively. Efforts such as the Global Observation of Forest Cover (GOFC) under the auspices of the Committee on Earth Observation Satellites (Janetos & Ahern, 1997) are being organized to apply the advances from research in land cover characterization from satellite data in an operational context. This is one of seven prototype activities of an integrated global observation strategy being coordinated through the Committee for Earth Observing Satellites. Current plans call for analysis of fine resolution data with complete spatial coverage to monitor the forest cover of the whole Earth every 3–5 years, supplemented by coarser resolution data acquired at least seasonally to identify the location of sites with especially rapid forest conversion (Ahern et al., 1998). GOFC, together with the World Forest Watch (Malingreau et al., 1992) and other international activities, provide a framework for meeting the needs of both the policy-making and the scientific communities for information about land cover. This international attention provides the imperative to take advantage of the opportunities and meet the challenges of moving from research to operational land cover monitoring.

CONCLUSIONS

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES

The use of satellite data to characterize land cover and land cover change at global and regional scales has progressed significantly over the last few decades. These advancements are due both to availability of improved data with finer spatial resolutions and improved methods for deriving land cover characteristics. Recent and upcoming launches of many new sensors will vastly improve capabilities to characterize land cover and land cover change. International attention on land cover change, particularly as it affects the global carbon budget, creates an imperative that the scientific community translate these advancements into operational monitoring of land cover. Several quasi-operational prototypes for land cover monitoring from satellite data have been successful and provide a foundation for developing an operational monitoring system.

Many challenges must be met before a transition from research to operational monitoring can become a reality, however. Among these is the need to make routine analysis of satellite data feasible with more automated methods than are currently used. Long-term continuity of satellite measurements and archives for storage are also key requirements for operational land cover monitoring. In addition, the development of operational methods to validate land cover products, which to date has not been carried out, presents an additional challenge that calls upon the ecological, remote sensing, and social science communities.

ACKNOWLEDGMENTS

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES

Research reported in this paper was supported by the US National Aeronautics and Space Administration (NASG56004, NAS596060, and NAG56282). Thanks to the Landsat Pathfinder Humid Tropics Project for providing graphics.

REFERENCES

  1. Top of page
  2. Summary
  3. INTRODUCTION
  4. PROGRESS IN THE APPLICATION OF SATELLITE DATA TO CHARACTERIZE GLOBAL LAND COVER
  5. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE CHALLENGES
  6. OPERATIONAL LAND COVER CHARACTERIZATION FROM SATELLITE DATA: THE OPPORTUNITIES
  7. CONCLUSIONS
  8. ACKNOWLEDGMENTS
  9. REFERENCES
  • Achard, F & Estreguil, C. (1995) Forest classification of Southeast Asia using NOAA AVHRR data. Remote Sensing Environ. 54, 198-208.
  • Adams, J.B, Sabol, D.E, Kapos, V, Filho, R.A, Roberts, D.A, Smith, M.O, Gillespie, A.R. (1995) Classification of multispectral images based on fraction endmembers: application to land-cover change in the Brazilian Amazon. Remote Sensing Environ. 52, 137-154.
  • Ahern, F, Janetos, A, Langham, E. (1998) Global observations of forest cover: one component of CEOS’ integrated global observing system strategy. Proceedings of the 27th International Symposium on Remote Sensing of Environment, Tromso, Norway.
  • Belward, A.S. (1996) The IGBP-DIS Global 1km Land Cover Data Set: Proposal and Implementation Plans. WP no. 13. IGBP-DIS, Toulouse.
  • Bierwirth, P.N. (1990) Mineral mapping and vegetation removal via data-calibrated pixel unmixing, using multispectral images. Int J. Remote Sens. 11, 1999-2017.
  • Bonan, G.B. (1997) Effects of land use on the climate of the United States. Clim. Change, 37, 449-486.
  • Chase, T. (1995) The sensitivity of a General Circulation Model to large scale vegetation changes. Atmospheric Science Paper no. 581, Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado.
  • Cihlar, J & Beaubien, J. (1997) Land cover of Canada 1995, Version 1. Canada Centre for Remote Sensing, Ottawa.
  • DeFries, R.S, Field, C.B, Fung, I, Justice, C.O, Matson, P.A, Matthews, M, Mooney, H.A, Potter, C.S, Prentice, K, Sellers, P.J, Townshend, J, Tucker, C.J, Ustin, S.L, Vitousek, P.M. (1995b) Mapping the land surface for global atmosphere-biosphere models: toward continuous distributions of vegetations functional properties. J. Geophys Res.100, 20,867-20,882.
  • DeFries, R, Hansen, M, Steininger, M, Dubayah, R, Sohlberg, R, Townshend, J. (1997) Subpixel forest cover in Central Africa from multisensor, multitemporal data. Remote Sensing Environ. 60, 228-246.
  • DeFries, R, Hansen, M, Townshend, J. (1995a) Global discrimination of land cover types from metrics derived from AVHRR Pathfinder data. Remote Sensing Environ. 54, 209-222.
  • DeFries, R, Hansen, M, Townshend, J.R.G, Sohlberg, R. (1998) Global land cover classifications at 8 km spatial resolution: The use of training data derived from Landsat imagery in decision tree classifiers. Int. J. Remote Sens., 19, 3141-3168.
  • DeFries, R.S & Townshend, J.R.G. (1994a) Global land cover: comparison of ground-based data sets to classifications with AVHRR data. Environmental remote sensing from regional to global scales. (ed. by G. Foody and P. Curran). John Wiley and Sons, Chichester.
  • DeFries, R.S & Townshend, J.R.G. (1994b) NDVI-derived land cover classification at global scales. Int J. Remote Sens. 15, 3567-3586.
  • DeFries, R.S, Townshend, J.R.G, Hansen, M. (1999) Continuous fields of vegetation characteristics at the global scale. J. Geophys. Res., 104, 16911-16925.
  • Eidenshink, J.C & Faudeen, J.L. (1994) The 1 km AVHRR global land data set: First stages in implementation. Int J. Remote Sens. 15, 3443-3462.
  • Foody, G.M, Lucas, R.M, Curran, P.J, Honzak, M. (1997) Non-linear mixture modelling without end-members using an artificial neural network. Int J. Remote Sens. 18, 937-953.
  • Friedl, M.A & Brodley, C.E. (1997) Decision tree classification of land cover from remotely sensed data. Remote Sensing Environ. 61, 399-409.
  • GCOS. (1997) GCOS/GTOS Plan for Terrestrial Climate-related Observations, Version 2.0. WMO/TD-no. 796, GCOS-28. Geneva.
  • Gopal, S, Woodcock, C, Strahler, A.H. (1996) Fuzzy AR™AP classification of global land cover fom AVHRR data set, Proceedings of the 1996 International Geoscience and Remote Sensing Symposium , Lincoln, Nebraska, 27-31 May, pp. 538-540.
  • Goward, S, Dubuyah, R, Haskett, J. (1996) Landsat-7 long term acquisition plan: Seasonal acquisitions, Proc. Pecora-13, August, 20-22, 1996 , Sioux Falls, South Dakota.
  • Goward, S & Williams, D. (1997) Landsat and earth system science: Development of terrestrial monitoring. Photogramm. Engng Remote Sensing, 63, 887-900.
  • Hansen, M, DeFries, R, Townshend, J.R.G, Sohlberg, R. (in press) Global land cover classification at 1km resolution using a decision tree classifier. J Remote Sens .
  • Hansen, M, Dubayah, R, DeFries, R. (1996) Classification trees: an alternative to traditional land cover classifiers. Int J. Remote Sens. 17, 1075-1081.
  • Houghton, R.A. (1995) Land-use change and the carbon cycle. Global Change Biol. 1, 275-287.
  • Houghton, R.A. (1999) The annual net flux of carbon to the atmosphere from changes in land use 1850-1990. Tellus, 51B, 298-313.
  • IGBP, Terrestrial Carbon Working Group. (1998) The terrestrial carbon cycle: Implications for the Kyoto Protocol. Science, 280, 1393-1394.
  • Iverson, L.R, Cook, E.A, Graham, R.L. (1989) A technique for extrapolating and validating forest cover across large regions. Int J. Remote Sens. 10, 1805-1812.
  • Iverson, L.R, Cook, E.A, Graham, R.L. (1994) Regional forest cover estimation via remote sensing: the calibration center concept. Landscape Ecol. 9, 159-174.
  • James, M.E & Kalluri, S.N.V. (1994) The Pathfinder AVHRR land data set: an improved coarse resolution data set for terrestrial monitoring. Int J. Remote Sens. 15, 3347-3364.
  • Janetos, A.C & Ahern, F. (1997) CEOS Pilot Project: Global Observations of Forest Cover (GOFC), Report from Meeting. Ottawa, Ontario, Canada.
  • Justice, C.O. et al . (1998) The Moderate Resolution Imaging Spectroradiometer (MODIS): Land remote sensing for global change research. IEEE Trans. Geosci. Remote Sensing, 36, 1228-1249.
  • Justice, C.O & Townshend, J.R.G. (1994) Data sets for global remote sensing: lessons learnt. Int J. Remote Sens. 15, 3621-3639.
  • Kalluri, S.N.V, JaJa, J, Bader, D, Zhang, Z, Townshend, J.R.G, Fallah-adl, H. (in press) High performance computing algorithms for land cover dynamics using remote sensing data. Int J Remote Sens. (in press) .
  • Kimes, D.S, Nelson, R.F, Manry, M.T, Fung, A.K. (1998) Attributes of neural networks for extracting continuous vegetation variables from optical and radar measurements. Int J. Remote Sens. 19, 2639-2663.
  • Lambin, E.F & Ehlirch, D. (1996) The surface temperature-vegetation index space for land cover and land-cover change analysis. Int J. Remote Sens. 17, 463-487.
  • Lambin, E & Strahler, A. (1994) Change-vector analysis in multitemporal space: a tool to detect and categorize land-cover change processes using high temporal-resolution satellite data. Remote Sensing Environ. 48, 231-244.
  • Los, S.O, Justice, C.O, Tucker, C.J. (1994) A global 1×1 degree NDVI data set for climate studies derived from the GIMMS continental NDVI data. Int J. Remote Sens. 15, 3493-3518.
  • Loveland, T.R & Belward, A.S. (1997) The IGBP-DIS global 1 km land cover data set, DISCover: first results. Int J. Remote Sens. 18, 3289-3295.
  • Malingreau, J.P, Da Cunha, R, Justice, C. (1992) Proceedings of the World Forest Watch Conference, Say Jose Dos Campos, European Commission .
  • Matthews, E. (1983) Global vegetation and land use: new high resolution data bases for climate studies. J. Clim. appl. Meteor. 22, 474-487.
  • Mayaux, P, Achard, F, Malingreau, J.P. (1998) Global tropical forest area measurements derived from coarse resolution satellite imagery: a comparison with other approaches. Environ. Conserv. 25, 37-52.
  • Mayaux, P & Lambin, E.F. (1997) Tropical forest area measured from global land cover classifications: inverse calibration models based on spatial textures. Remote Sensing Environ. 59, 29-43.
  • Moody, A. (1998) Using landscape spatial relationships to improve estimates of land-cover area from coarse resolution remote sensing. Remote Sensing Environ. 64, 202-220.
  • Mucher, C.A, Steinnocher, K.T, Kressler, F.P, Heunks, C. (in press) Land cover characterization and change detection for environmental monitoring of pan-Europe. Int J. Remote Sens. in press .
  • Olson, J.S, Watts, J, Allison, L. (1983) Carbon in live vegetation of major world ecosystems. W-7405-ENG-26. U.S. Department of Energy, Oak Ridge National Laboratory.
  • Pech, R.P, Graetz, R.D, Davis, A.W. (1986) Reflectance modelling and the derivation of vegetation indices for an Australian semi-arid shrubland. Int J. Remote Sens. 7, 389-403.
  • Privette, J.L, Fowler, C, Wick, G.A, Baldwin, D, Emery, W.J. (1995) Effects of orbital drift on Advanced Very High Resolution Radiometer products: Normalized Difference Vegetation Index and sea surface temperature. Remote Sensing Environ. 53, 164-171.
  • Quarmby, N.A, Townshend, J.R.G, Settle, J.J, Milnes, M, Hindle, T.L, Silleos, N. (1992) Linear mixture modelling applied to AVHRR data for crop area estimation. Int J. Remote Sens. 13, 415-425.
  • Roberts, D.A, Gardner, M, Church, R, Ustin, S, Scheer, G, Green, R.O. (1998) Mapping chaparral in the Santa Monica Mountains using multiple endmember spectral mixture models. Remote Sensing Environ. 65, 267-279.
  • Running, S.W, Loveland, T.R, Pierce, L.L. (1994) A vegetation classification logic based on remote sensing for use in global biogeochemical models. Ambio, 23, 77-81.
  • Saatchi, S.S, Soares, J.V, Alves, D.S. (1997) Mapping deforestation and land use in Amazon rainforest by SIR-C imagery. Remote Sensing Environ. 59, 191-202.
  • Sellers, P.J. (1985) Canopy reflectance, photosynthesis and transpiration. Int J. Remote Sens. 6, 1335-1372.
  • Settle, J & Drake, N.A. (1993) Linear mixing and the estimation of ground cover proportions. Int J. Remote Sens. 14, 1159-1177.
  • Shukla, J, Nobre, C, Sellers, P. (1990) Amazon deforestation and climate change. Science, 247, 1322-1325.
  • Skole, D & Tucker, C. (1993) Tropical deforestation and habitat fragmentation in the Amazon: satellite data from 1978 to 1988. Science, 260, 1905-1910.
  • Swain, P.H & Davis, S.M. (eds) (1978) Remote sensing. The quantitative approach. McGraw-Hill Book Company, New York.
  • Townshend, J.R.G, Bell, V, Desch, A, Havlicek, C, Justice, C, Lawrence, W.L, Skole, D, Chomentowski, W, Moore, B.I, Salas, W, Tucker, C.J. (1995) The NASA Landsat Pathfinder Humid Tropical Deforestation Project. Land satellite information in the next decade, ASPRS conference , Vienna, VA. 25-28 Sept. 1995, IV-76 - IV-87.
  • Townshend, J, DeFries, R, Dubayah, R, Goward, S, Kearney, M, Tucker, C.J, Vermote, E. (1997) Land Cover Characterization at Regional and Global Scales: Lessons Learnt and Prospects, Proceedings of the 23rd Annual Conference of the Remote Sensing Society on Observations and Interactions , Reading, UK . The Remote Sensing Society, Nottingham.
  • Townshend, J.R.G & Justice, C.O. (1988) Selecting the spatial resolution of satellite sensors required for global monitoring of land transformations. Int J. Remote Sens. 9, 187-236.
  • Townshend, J.R.G, Justice, C.O, Kalb, V.T. (1987) Characterization and classification of South American land cover types using satellite data. Int J. Remote Sens. 8, 1189-1207.
  • Tucker, C.J, Townshend, J.R.G, Goff, T.E. (1985) African land-cover classification using satellite data, Scence. 227, 369-375.
  • Wilson, M.F & Henderson-Sellers, A. (1985) A global archive of land cover and soils data for use in general circulation models. J. Climatol. 5, 119-143.
  • Zhu, Z & Evans, D.L. (1992) Mapping midsouth forest distributions. J. For. 90, 27-30.
  • Zhu, Z & Evans, D.L. (1994) U.S. forest types and predicted percent forest cover from AVHRR data. Photogramm Eng Remote Sensing, 60, 525-531.