¹ Linda See, ¹ Steffen Fritz, ² Philip Thornton, ³,⁴ Liangzhi You, ⁵ Inbal Becker-Reshef, ⁵ Chris Justice, ⁶ Olivier Leo ⁷ Mario Herrero

¹ IIASA, Schlossplatz 1, A-2361 Laxenburg, Austria

² CGIAR Program on Climate Change, Agriculture and Food Security (CCAFS), ILRI, PO Box 30709, Nairobi 00100, Kenya

³ International Food Policy Research Institute, 2033 K Street, NW, Washington DC, 20006, USA

⁴ School of Nature Conservation, Beijing Forestry University, 35 Qinghua East Road, Beijing, China

⁵ Department of Geography, University of Maryland, 2181 LeFrak Hall, College Park, Maryland, 20742, USA

⁶ JRC, Via Fermi 2749, TP 266, Ispra, Italy

⁷ ILRI, PO Box 30709, Nairobi 00100, Kenya

Abstract

This paper describes the start of a data sharing process to develop a consolidated community cropland map, which was initiated through a recent workshop on characterizing and validating global agricultural land cover. Participants from different organizations around the world were asked to contribute their various cropland maps prior to the workshop. Other data such as geo-tagged photos, in-situ data, classified satellite images and videos also were provided as part of this process. The data are now available online at agriculture.geo-wiki.org. This data sharing exercise, which has culminated in a new Sub-task on Agricultural Mapping as part of the GEO Agricultural Monitoring Task, will continue as an ongoing process and represents an effective model for how data sharing could be facilitated across the GEO community.

The Need for a Consolidated Community Cropland Product

Global land cover products provide important baseline information for resource assessments as well as inputs to a variety of land use models. Accurate estimates of cropland are crucial for determining land availability and for food security purposes, yet global land cover products do not provide consensus on the spatial distribution or total amount of cropland in production currently. For example, the global area under cropland is estimated to be between 1.22 to 1.71 billion hectares, at a 90 percent confidence level [1], which indicates a high uncertainty with a 40 percent difference between the upper and lower estimates. One source of information on croplands is the different medium- to coarse-resolution satellite-derived land cover datasets that are available, including the GLC-2000 [2], the MODIS v.5 land cover products [3] and GlobCover 2005/2009 [4] [5], which have classes for cultivated areas and mosaics of cropland and natural vegetation. Although, with the potential for being up-to-date, they were developed using different data and classification algorithms and do not have particularly high accuracy for estimation of cropland or crop types. Other global cropland products exist, some of which have been calibrated using cropland statistics, such as the M3-Cropland layer of agricultural lands for 2000 [1], the cropland probability layer from MODIS [6] and a global map of rain-fed cropland areas [7]. However, a product is needed with a minimum spatial resolution, and must be of sufficient quality to meet the needs of the food security and land use modeling communities, with an accuracy of at least 80 to 85 percent. It is thus a challenge for individuals and organizations working with these datasets to find a reliable picture of cultivation in one dataset. Africa often is the focus region for agricultural development due to consistent hunger and poverty issues, and weak local capacity results in huge data gaps in national statistics. A consistent and accurate measure of agricultural land derived from satellites could help fill these gaps partly, and provide valuable basic data for designing development programs as well as in monitoring and evaluating food security in the continent.

There are some encouraging developments on the horizon, such as the planned production of 30-m global land cover products by China and the United States, to be derived using Landsat data, which are scheduled for release in December 2013 [8], and high spatial and temporal resolution data that will be provided by ESA’s Sentinel satellites in the future [9]. However, the food security and land use modeling communities require a solution that delivers more accurate cropland information in the short-term. An improved cropland product at a 1-kilometer resolution has recently been created for Africa through combining a number of existing data sources, with calibration using national and sub-national crop statistics [10]. If recent national and regional products were to be made freely available, an even more accurate cropland extent map could be developed with minimal cost through a harmonization of land use categories. The challenge has been to raise awareness and mobilize the community to share these products.

A Land Cover Workshop to Facilitate Data Sharing

To initiate this process of sharing data, with the aim of developing an enhanced cropland product, the Characterizing and Validating Global Agricultural Landcover workshop was held at the International Institute for Applied Systems Analysis (IIASA), from June 13-15. Data sharing was not limited to cropland maps but also included geo-tagged photos and other related in-situ data. The land cover workshop was hosted by IIASA and the CGIAR Consortium for Spatial Information, in close collaboration with the Group on Earth Observation (GEO), GOFC-GOLD and the Joint Research Centre of the European Commission (JRC). More than 70 international experts on remote sensing, land cover, land use, cropland and rangeland mapping, crop type mapping, area estimation and crowd-sourcing attended the workshop representing universities, national mapping agencies, research institutes and several international organizations including the Food and Agriculture Organization of the United Nations (FAO), the International Food Policy Research Institute (IFPRI), the International Livestock Research Institute (ILRI) and the International Crops Research Institute for the Semi-Arid-Tropics (ICRISAT). The emphasis was on improving African cropland maps, but the workshop discussions were broadened to a global scope and included participants from all major continents. Funding was provided by the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) and the JRC to allow participants from African countries and other developing nations to contribute their data and expertise during the workshop.

The requirement for the sharing of cropland maps and other geo-tagged data prior to the workshop was a change from past workshops on similar topics. This resulted in the active engagement of many participants before the workshop and contributed toward its success. These contributed data were then consolidated into a restricted version of geo-wiki that was used during the workshop to highlight the wealth of data contributed through this process. This model of data sharing prior to the workshop required participants to invest time in preparing their data and providing metadata. In addition to an African cropland map provided by the JRC, national cropland maps for Africa were provided for Burkina Faso, Gambia, the Horn of Africa and Senegal, Mali; Nigeria, South Africa, Southern Sudan, and Zimbabwe. Outside of Africa, crop masks for sugar cane and summer crops were provided for Brazil, and crop percentage maps for were provided for China, Kazakhstan, India, and the U.S. For Australia and Europe, it was possible to download detailed land cover and land use maps from ACLUMP and the European Environment Agency. Other data contributed include field survey data in Argentina and Russia; classified Landsat images; and geo-tagged photos, in-situ points and videos. The full list of datasets and contributors can be found in the final workshop report [11].

A number of key issues were raised during the two-day workshop from the plenary presentations and the breakout groups. Specific issues discussed include methods for cropland and crop-type mapping; crop area estimation; rangeland mapping, the value of crowd-sourcing for training, calibration and validation of land cover; integration of remote sensing and socio-economic data; and the availability of cropland data at the national and regional level. A key action from the workshop was to establish a Sub-Task on Agricultural Mapping under the GEO Agriculture Monitoring Task to be led by IIASA. A follow-up workshop on rangeland mapping and monitoring was also recommended in the framework of a new Sub-Task on Rangeland Productivity.

Building a Living, Community-based Consolidated Cropland Map

One of the first aims of the new Sub-Task on Agricultural Mapping will be to build a living, community-based consolidated cropland map. This initial product will provide the agricultural monitoring, food security and land use change communities with a better cropland product than currently exists, and will be freely available to researchers and the general public. All workshop participants agreed for their data contributions to be used in the development of an integrated cropland product. This new cropland extent map will integrate all the products contributed by workshop participants using a methodology similar to that reported in [10] at a 1-kilometer resolution. The map will also be calibrated with national and sub-national crop statistics as available. Validation will involve the wider community using crowd-sourced data and Google Earth. The first version of the map was published at the end of 2011 and is downloadable from the same site (agriculture.geo-wiki.org). Crop experts with knowledge of crop locations can use the new tools that have been developed to undertake qualitative validation with drawing tools and commenting facilities. These tools will be trialed as part of outreach activities in upcoming workshops.

The community-based consolidated cropland product will be updated when more crop information becomes available at the national and regional level. In this way, the product will become a living map, which will continue to improve with more contributions. The process of data sharing, which began with the workshop, should be seen as the start of an ongoing process that will continue through the new Sub-Task on Agricultural Mapping. The broader agricultural mapping community is encouraged to take part by providing more national and regional data on croplands, to help validate the product, and to improve our current knowledge of how much cropland there is, its location, and if data quality improves sufficiently or changes over time.

Lessons Learned from the Workshop

The workshop was used as a vehicle to kick-start the data sharing process. A number of lessons have been learned that may be of value to those wanting to compile global datasets based on national and regional data products such as in socio-economic, geological, ecological and earth systems science in general:

‰Û¢ Target the right individuals: It is important to invite people who have data to contribute. Names were initially provided through the steering committee and through evolving contacts with potential participants but not all areas of Africa were covered. The benefits of sharing and contributing actively to the workshop and the community as a whole were used as arguments to persuade other organizations to contribute. This proved to be a very effective approach that gained momentum as the date of the workshop approached;

‰Û¢ Provide incentives: Two main types of incentives were provided to the participants. Payment or partial payment of travel expenses to attend the workshop from developing countries was offered on the conditions that the data were shared before the workshop. A second incentive was co-authorship on a scientific paper to all participants;

‰Û¢ Follow up after the workshop: Data contributions that were promised during the workshop were actively followed up by email along with new leads for sources of data.

The workshop was an intensive process that required more effort than a conventional workshop due to the exchange, reformatting, display and documentation of the datasets. However, the success of the workshop in terms of data sharing, networking and the initiation of an ongoing agricultural mapping process fully justifies the efforts.

If you have a cropland product you want to contribute to the consolidated community global cropland map, please contact us. The map will be registered in the GEOSS portal and become a citable product that will include your authorship details.

References

[1] N. Ramankutty, A. T. Evan, C. Monfreda, and J. A. Foley, ‰ÛÏFarming the planet: 1. Geographic distribution of global agricultural lands in the year 2000‰Û, Global Biogeochemical Cycles, vol. 22, GB1003, doi:10.1029/2007GB002952, 2008.

[2] E. BartholomÌ© and A. S. Belward, ‰ÛÏGLC2000: A new approach to global land cover mapping from earth observation data‰Û, International Journal of Remote Sensing, vol. 26(9), pp. 1959-1977, 2005.

[3] M. A. Friedl, D. Sulla-Menashe, B. Tan, A. Schneider, N. Ramankutty, A. Sibley and X. Huang, ‰ÛÏMODIS Collection 5 global land cover: Algorithm refinements and characterization of new datasets‰Û, Remote Sensing of Environment, vol. 114(1), pp. 168-182, 2010.

[4] P. Bicheron, P. Defourny, C. Brockmann, L. Schouten, C. Vancutsem, M. Huc, S. Bontemps, M. Leroy, F. Achard, M. Herold, F. Ranera and O. Arino (2008). GlobCover: Products Description and Validation Report, 18, Toulouse, France. [Online]. Available: http://ionia1.esrin.esa.int/docs/GLOBCOVER_Products_Description_Validation_Report_I2.1.pdf

[5] S. Bontemps, P. Defourney, E. Van Bogaert, O. Arino, V. Kalogirou and J. R. Perez (2011) GLOBCOVER 2009: Products Description and Validation Report. [Online]. Available: http://ionia1.esrin.esa.int/docs/GLOBCOVER2009_Validation_Report_2.2.pdf.

[6] K. Pittman, M. C. Hansen, I. Becker-Reshef, P. V. Potapov, and C. O. Justice, ‰ÛÏEstimating global cropland extent with multi-year MODIS data‰Û, Remote Sensing, vol. 2, pp. 1844-1863, 2010.

[7] C.M. Biradar, P.S. Thenkabail, P. Noojipady, Y. Li, V. Dheeravath, H. Turral, M. Velpuri, M. K. Gumma, O. R. P. Gangalakunta, X. L. Cai, X. Xiao, M.A. Schull, R. D. Alankara, S. Gunasinghe and S. Mohideen, ‰ÛÏA global map of rainfed cropland areas (GMRCA) at the end of the last millennium using remote sensing‰Û, International Journal of Applied Earth Observation and Geoinformation, vol. 11, pp. 114-129, 2009.

[8] US Department of the Interior (2010). United States Launches New Global Initiative to Track Changes in Land Cover and Use. [Online]. Available: http://www.doi.gov/news/pressreleases/United-States-Launches-New-Global-Initiative-to-Track-Changes-in-Land-Cover-and-Use-Data-Sharing-Will-Assist-Land-Managers-Worldwide.cfm.

[9] European Space Agency (ESA) (2011). GMES Sentinels. [Online]. Available: http://www.esa.int/esaLP/SEM097EH1TF_LPgmes_0.html.

[10] S. Fritz, L. You, A. Bun, L. See, I. McCallum, C. Schill, C. Perger, J. Liu, M. Hansen and M. Obersteiner, ‰ÛÏCropland for sub-Saharan Africa: A synergistic approach using five land cover data sets‰Û, Geophysical Research Letters, 38, L04404, doi:10.1029/2010GL046213, 2010.

[11] IIASA, ‰ÛÏCharacterizing and Validating Global Land Cover Workshop. IIASA 13-15 June 2011. Workshop Report‰Û, http://www.iiasa.ac.at/Research/FOR/lc/IIASAWorkshopReportJun2011.pdf, 2011.