Figure 1. The location of the 13,066 sample sites of the RSS. Sites are located at 1 degree intersections of latitude and longitude, except in Canada and the Russian Federation. Canadian sites are located to correspond to the Canada National Forest Inventory. Sites in the Russian Federation represent a stratified sample taken from the systematic grid based on MODIS derived forest area and change. Sites are not available for most of Siberia due to missing Landsat acquisitions in 1990.
Lindquist, E. J.1, McInerney, D.2, D’Annunzio, R.1, Barredo, J.2
1Food and Agriculture Organization of the United Nations, Rome, Italy.
2Joint Research Centre of the European Commission, Institute for Environment and Sustainability, Ispra, Italy.
1.0 Introduction
Forests cover about four billion hectares, or 30 percent, of the Earth’s land area (FAO, 2010). At all spatial scales, from global to local, trees and forests play a critical role in human livelihoods, as well as ecosystem function and health. Monitoring the forest resource requires accurate and timely information on both extent and area change over time.
The Food and Agriculture Organization of the United Nations (FAO) evaluates and compiles data on the state of the world’s forests in the Global Forest Resources Assessment (FRA). The FRA, published every five to 10 years since 1948, summarizes forestry statistics reported by all 233 countries around the world, and provides a critical baseline for global efforts to improve the sustainable management of forests.
In 2010, the FRA included a global remote sensing survey (RSS) of the world’s forests derived from a systematic sample of Landsat satellite imagery from 1990, 2000 and 2005. The goal of the RSS was to obtain independent, globally consistent information on the distribution and changes in tree cover and forest land-use from 1990 to 2005 at the regional, climatic domain and global levels.
Table 1. FRA variables collected and available through web-based services for analysis.
This paper describes the creation and provision of forestry statistics and metadata compiled from the FRA and RSS data layers through web-based services. FRA data can be queried at the national and multi-national level. RSS-derived information can be summarized over large geographic areas irrespective of country boundaries. Both sources of forest information can be integrated into assessments as background information, used for statistical comparisons, to increase the temporal span of assessments and to support the formulation of forest policies. The goal of this work is to improve the access and usability of FRA and RSS data, enable easier analysis of key forestry statistics through time, and provide baseline national and regional information on forest area and change through web-based services.
2.0 The FRA and RSS data
The FRA global data tables cover a wide range of forestry information from forest area and change, to policy and legal framework issues, all linked to a specific country or territory (Table 1) (FAO, 2010). The FRA tabular attributes can therefore be joined with country political boundary spatial data layers to create spatial datasets to be used within Geographic Information Systems (GIS).
Figure 2. The satellite imagery (left), the land use classification (center) and the changes between time periods (right) for a RSS sample site in Australia. At each RSS sample site worldwide, land cover and land use are classified and changes between time periods are recorded. The same, single polygon is outlined (in red) as an example to help track the spectral and associated land use label changes.
The RSS is a collaborative project between the FAO and the Joint Research Centre of the European Commission (JRC), which integrates the work of the JRC TREES, FOREST and MONDE research groups. The RSS produced land cover and land-use classifications, in vector format, are for a systematic sample of Landsat imagery located at 1-degree latitude and longitude intersections from 1990, 2000 and 2005. The total number of sites analyzed after accounting for individual national cases, missing data and statistical outliers was 13,066 (Figure 1) (FAO and JRC, in press). The RSS database contains nearly 7 million polygons of classified land cover, land use and change (Figure 2) with a minimum mapping unit of 5 hectares (Bodart et al., 2011). The land cover and land use legend used in the RSS classification are located in Table 2.
3.0 FRA web-based services
Both FRA and RSS data naturally lend themselves to geographic analysis, however, neither the FRA data tables nor the RSS polygons had previously been available through Open Geospatial Consortium (OGC) Web Services, either Web Map Services (WMS) or Web Feature Services (WFS). These technologies allow the geographic data produced by the FAO to be visualized and analyzed from any GIS client via the Internet.
Table 2. The RSS land use and land cover classes
The FRA and RSS data are being made available through web-based services within the context of the larger EuroGEOSS project (Pearlman et al., 2011). FAO has initiated an instance of GeoNetwork to host the metadata records. FAO metadata have been created for each of the FRA forestry statistics datasets conforming to the ISO19139 standard. FAO has also initiated an instance of GeoServer for the storage and publication of the geographic data corresponding to the metadata records contained in the GeoNetwork instance. Geographic data are stored in a PostgreSQL/PostGIS database and the layers are now available as OGC WMS and WFS layers. The FRA data are accessible in a simple, dedicated web-based viewer and the services are federated within the EuroGEOSS Discovery broker (Nativi et al., 2011), which in turn are now part of the GEOSS Data Core.
The Web Processing Service (WPS) technology is based on PyWPS, a Python WPS environment. PyWPS is light-weight, but it enables processes to access underlying geo-spatial and analytical software, such as GDAL/OGR (GDAL, 2012), R Statistics (R Development Core Team, 2008) and other UNIX/Linux programs. The R statistical package routines for analyzing forest area and forest area change will be encapsulated within Python and accessible from PyWPS.
4.0 Web-based analysis scenarios
Two scenarios have been developed to highlight the utility of web-based services for accessing and analyzing FAO, FRA and RSS data. FRA tabular statistics, linked to country geographic data, allow regional or national scale analysis of key forestry statistics including forest area (Figure 3), forest type and forest uses. These variables can be summarized at the national level or combined in country groupings or regional levels, where appropriate, to produce results suitable for general assessment of forest characteristics.
Figure 3. FAO FRA GeoNetwork Metadata catalogue showing an example of forest area as percentage of total country area for year 2000 globally.
The web processing services for analysis of the RSS sample data allow a user to query and produce summary statistics, with corresponding estimates of precision, for tree and other land cover, forest land use and changes in the areal extent of each over time for a user-defined area of interest. This information can be integrated into assessments as background information on forest area and change, extend the temporal span of the analysis, and be used for statistical comparisons against other existing data. Currently, data are available only for Spain, as developed for the EuroGEOSS project. However, the ultimate goal will be to work with other countries to make the global land cover and land use data available in its entirety.
The nature of the FRA tabular data precludes sub-national analysis; therefore only countries or country groups can be selected for analysis. FRA data services can be used to produce summary results in the form of graphs, tables or chloropleth map products (via WMS/WFS) or augment other WMS/WFS forest thematic layers. Information from the RSS, however, could be summarized for any suitably sized geographic area; in this case defined by the precision of the parameter estimate desired. For example, very large areas have generally higher precision estimates than smaller areas, depending on landscape heterogeneity, based on sampling error. For all RSS-based analyses, confidence intervals will be generated with each statistic.
5.0 Conclusion
The availability of FRA and RSS data through web-based services will improve the accessibility and usefulness of the extensive FAO forestry databases. The use scenarios already developed with the information are good examples of baseline data creation and analysis. Land managers, researchers and civil society will, hopefully, access the data and expand the types and utility of analyses available.
References
Bodart, C., Eva, H., Beuchle, R., RaÁi, R., Simonetti, D., Stibig, H.-J., Brink, A., Lindquist, E., and Achard, F.. (2011). Pre-processing of a sample of multi-scene and multi-date Landsat imagery used to monitor forest cover changes over the tropics. ISPRS Journal of Photogrammetry and Remote Sensing, 66(5), 555-563. International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). doi:10.1016/j.isprsjprs.2011.03.003
EuroGEOSS (2011) D.3.2b: Report on the Design Specifications for EuroGEOSS Forestry Components and Interfaces (FAOC), available at: http://www.eurogeoss.eu/.
FAO (2010) Global forest resources assessment 2010. Main report. FAO Forestry Paper 163. Rome.
FAO and JRC (in press). Global forest land use change from 1990 to 2005. FAO Forestry Paper in press, FAO Forestry Department, Rome.
GeoNetwork (2011) http://geonetwork4.fao.org/geonetwork/srv/en/fra.home
GDAL (2012) GDAL – Geospatial Data Abstraction Library: Version 1.9.0, Open Source Geospatial Foundation, http://gdal.osgeo.org
Nativi, S., Craglia, M., Vaccari, L., & Santoro, M. (2011). Searching the New Grail: Inter-Disciplinary Interoperability. In S. Geertman, W. Reinhardt, & F. Toppen (Eds.), Proceedings of the 14th AGILE International Conference on Geographic Information Science, April 18-22, 2011 (pp. 1-9). Utrecht, Netherlands. ISBN: 978-90-816960-1-2
Pearlman, J., Craglia, M., Bertrand, F., Nativi, S., Gaigalas, G., Dubois. G., Niemeyer, S., Fritz., S. (2011) EuroGEOSS: an interdisciplinary approach to research and applications for forestry, biodiversity and drought. In: “Proceedings of the 34th International Symposium on Remote Sensing of Environment”, April 10-15, 2011, Sydney, Australia
R Development Core Team (2008). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.
Erik Lindquist is a remote sensing specialist with the Food and Agriculture Organization of the United Nations (FAO). He is coordinator of remote sensing activities within the Global Forest Resources Assessment (FRA) team based in Rome, Italy. He is completing his PhD in Geospatial Science and Engineering from South Dakota State University (USA).