I N N O V A T I V E A C T I V I T Y P R O F I L E 5 . 5
This profile was prepared by K. D. Shepherd, T.-G. Vồgen, and T. Gumbricht, World Agroforesty Centre (ICRAF), Nairobi, Kenya, and M. G. Walsh, Earth Institute, Columbia University, New York.
■ What caused the degradation in places where it exists, and how can further degradation be prevented?
■ Can land degradation be reversed, and if so, what are the costs to individuals and to society?
■ Are there cost-effective and socially acceptable means for treating degraded lands to increase their productivity, while avoiding harmful side effects to the environment, such as the pollution of surface waters and accelerated greenhouse gas emissions?
PROJECT OBJECTIVES AND DESCRIPTION Surprisingly, the world does not have clear answers to these questions at present. The basic premise of this project is that a problem cannot be managed unless progress can be measured from a baseline toward a well-defined target. Thus, a land health surveillance system must accomplish the following:
■ Provide high spatial resolution and practical, timely, and cost-effective information about where specific land degradation processes occur in a given region or country and how those processes are changing over time.
■ Identify areas at risk of degradation and the commensu- rate preventive measures in a spatially explicit way.
■ Provide a framework for rigorous scientific testing and implementation of locally relevant rehabilitative soil management interventions, addressing what works, what does not, where, how, and at what cost to individuals and society.
■ Anticipate and respond to external requests from a wide audience (that is, farmers, conservationists, scientists, and policy makers).
PRESENTATION OF DIAGNOSTIC SURVEILLANCE AND OPERATIONAL FRAMEWORK
Human health surveillance techniques are a normal part of public health. Health surveillance is based on case definitions that define prevalence (percentage of people affected) and incidence (new cases). This project proposes an analogous land health surveillance system that provides the scientific and factual database essential to informed decision making and appropriate policy action (Shepherd and Walsh 2007).
Soil health diagnostic surveillance aims are as follows:
■ Provide diagnostic information on land degradation problems to guide resource allocation and management decisions.
■ Identify cause-and-effect relationships needed for pri- mary prevention, early detection, and rehabilitation of degraded land at different spatial scales.
■ Provide a scientifically rigorous platform for testing and monitoring land management interventions.
■ Provide a conceptual and logical framework for under- standing coupled social-ecological systems.
A diagnostic surveillance framework (box 5.1) can pro- vide a basis for a quantitative, evidence-based approach to land management. After a problem has been identified, a critical step is to describe a case definition through which the problem can be quantified. Problems such as disease in populations generally exist as a continuum of severity; how- ever, for practical reasons, dichotomizing the diagnostic continuum into “cases” and “noncases” or “affected” and
“nonaffected” is often helpful. The lack of rigorous stipula- tion of diagnostic criteria for key land degradation prob- lems is a major impediment in formulating a sound devel- opment policy. Adequate definitions of degraded landand nondegraded landare a prerequisite to assessing the extent of land degradation.
After case definitions are stipulated, a screening test is required to measure the problem in individuals or sample units and classify them as “case” or “noncase.” The availabil- ity of rapid, reliable (that is, highly repeatable and repro- ducible), and cost-effective screening tests (for example, equivalent to blood tests used in medicine) is key to using the surveillance framework to conduct prevalence surveys involving measurement of a large number of sample units.
In clinical medicine, large investments are made in develop- ment of screening tests, and even the case definition may be defined in relation to the screening test. For example, for some disorders, an operational case definition is used that assigns an arbitrary cut-off value of the screening test as a decision threshold for treatment.
The surveillance approach is put into effect using a com- bination of cutting-edge tools (figure 5.3), including satel- lite remote sensing at multiple scales; georeferenced ground-sampling schemes based on sentinel sites; infrared spectroscopy for rapid, reliable soil and plant tissue analysis;
and mixed-effects statistical models to provide population- based estimates from hierarchical data. The methods pro- vide accurate information on the areas where land degrada- tion is taking place, on the different manifestations of land degradation and soil constraints, on the extent of the prob- lems, and on the sort of intervention strategies that are required to prevent or reverse degradation. The methods have been designed to be simple and cost-effective so that
142 CHAPTER 5: RAINFED DRY AND COLD FARMING SYSTEMS
they can be implemented in isolated areas and in countries with limited resources.
At a regional or national scale, land degradation risk domains are first established using low-resolution time- series satellite information on vegetation cover. These
domains are further sampled using sentinel sites, consisting of 10-by-10-kilometer blocks. Within sentinel sites, high- resolution imagery and ground sampling are used to gather data on vegetation and soil condition at randomized points.
Infrared spectroscopy is used for rapid, reliable, and low-cost
INNOVATIVE ACTIVITY PROFILE 5.5: LAND DEGRADATION SURVEILLANCE 143
The diagnostic surveillance framework involves the following steps:
1. Identify the specific land degradation problem or groups of problems.
2. Develop a rigorous case definition of affected and nonaffectedstates.
3. Develop a screening test (or set of tests) so that sub- jects can be assigned rapidly to affected or nonaf- fected states. Infrared spectroscopy can play a key role as a screening tool for identification of cases.
4. Apply the screening test to subjects in randomized sampling schemes designed to provide unbiased prevalence data on the specified problem.
5. Conduct measurements. Simultaneous measure- ment of environmental and socioeconomic corre- lates permits problem risk factors to be identified.
Controllable risk factors point to the main manage- ment levers for controlling the problem.
6. Confirm risk factors through follow-up surveys that measure changes in the problem over time (inci- dence) and assess intervention outcomes. Assess- ment of outcomes may lead to a new or refined problem definition.
The accompanying figure shows the relationship of these steps.
Box 5.1 Steps in the Diagnostic Surveillance Framework
Source:International Centre for Research in Agroforestry.
Diagnostic Surveillance Framework
infrared spectroscopy
identify problem
develop case definition
measure prevalence (number of cases/area)
measure environmental
correlates
measure incidence (number of cases/area/time) develop
screening tests
differentiate risk factors
confirm risk factors
Source:Authors.
soil analysis and development of soil condition indexes.
Degradation indexes are related to risk factors such as vege- tation type and cover and are then mapped out through cal- ibration to the satellite imagery using statistical inference.
This information is used to spatially target land management strategies for systematic testing. The sentinel sites provide not only a framework for change detection through follow- up surveys (for example, after five years) but also a spatial platform for testing recommended land management options. For example, spatially distributing tree planting tri- als in each sub-block ensures that species are tested over a wide range of land conditions; consequently, growth perfor- mance can be correlated with site indexes, which can be used to predict tree performance at new sites. The steps used in the framework are described in more detail in box 5.2
The land degradation surveillance framework is being used in a UNEP capacity-building project to guide strategies for land restoration in five West African dryland countries
(see http://www.worldagroforestry.org/wadrylands/index .html) and in a World Bank GEF project in Kenya, led by the Kenya Agricultural Research Institute, which is designed to tackle land degradation problems in the Lake Victoria basin.
Soil health surveillance has been recommended as part of a NEPAD-endorsed strategy for saving Africa’s soils (Swift and Shepherd 2007) and is proposed for Sub-Saharan Africa as a component of the Global Digital Soil Map of the World project (see http://www.globalsoilmap.net/). Further infor- mation on infrared spectroscopy for sensing soil quality is available at http://www.worldagroforestrycentre.org/sens ingsoil/.
BENEFITS AND RESULTS OF THE ACTIVITY The activity provides a scientifically rigorous framework for evidenced-based management of land resources, modeled on well-tested scientific approaches used in epidemiology.
144 CHAPTER 5: RAINFED DRY AND COLD FARMING SYSTEMS
Figure 5.3 Successive Samples of Land Degradation Problem Domains at a Hierarchy of Scales Using Satellite Imagery, Ground Sampling, and Laboratory Analysis of Soils by Infrared Spectroscopy
r2 = 0 .9 4 0
5 0 1 0 0 1 5 0
0 5 0 1 0 0 1 5 0
actual value
national or regional
district or watershed sentinel sites
predicted value soil organic carbon calibration
block or sentinel site Lake Baringo, Kenya
soil spectral signatures trend in net primary
productivity for Kenya 1981–2000
2001
Conversions
Source:ICRAF.
INNOVATIVE ACTIVITY PROFILE 5.5: LAND DEGRADATION SURVEILLANCE 145
The land degradation surveillance framework involves the following steps:
1. At a regional or national scale, establish land degra- dation risk domains using low-resolution time- series satellite information on vegetation cover in combination with long-term rainfall records. The risk domains indicate areas where land may have been degraded or recovered over the past 25 years and are used as a sampling frame for more detailed studies. Alternatively, stratification and sampling of the Landsat World Reference System grid can be used as a sampling frame. Ancillary data on popula- tion, infrastructure, climatic zones, and the like are integrated to build quantitative scenario analyses.
2. Sample contrasting areas using moderate-resolution (for example, Landsat, ASTER, SPOT) satellite imagery, which provides data on major land-cover conversions. Processing of full-coverage imagery at this scale provides data on prevalence of woody cover and bare soil areas. Variation within these areas is fur- ther sampled through sentinel sites to provide more detailed information on land condition. Sentinel sites consist of 10-by-10-kilometer blocks, which are logis- tically convenient for field sampling while being large enough to encompass major landscape variability.
3. For the sentinel sites, obtain high-resolution (0.6 to 2.4 meter) satellite imagery, which allows individual fields, trees, and erosion features to be observed.
Within the sites, a standardized, georeferenced ground survey is used to provide direct measurement of land condition. The 10-by-10-km blocks are spa- tially stratified into 2.5-kilometer sub-blocks. Within each sub-block, an area of 1 square kilometer is sam- pled using a cluster of 10 randomized 1,000-square- meter observation plots. Direct observations are made in four 100-square-meter subplots. Socioeco- nomic surveys also use the cluster design (for exam- ple, sampling of households or villages nearest to clus- ter centroids).
4. Within plots, observe landform, topography, visible signs of soil erosion, land use, vegetation type and cover, and vegetation density and distribution, and take soil samples. Vegetation type is classified using the Food and Agriculture Organization Land Cover Classification System, supplemented with woody bio- mass estimates. Single-ring infiltration measurements are made on a selection of plots (three in each cluster).
A field crew of four people can complete a block in
about 14 to 16 field days. The number of plots can be adjusted, if desired, to meet different objectives.
5. Characterize soil samples using infrared spectroscopy.
This technique is widely used in industry for rapid and routine characterization of materials and has been adapted for rapid, reliable, and low-cost soil analysis. This no-chemical method is attractive for laboratories in developing countries because it mini- mizes sample preparation and requires only a source of electricity. Furthermore, many agricultural inputs and products can be analyzed using the same instru- ment. Subsets of samples are sent to specialized labo- ratories for conventional soil analysis and isotope analysis. These expensive analyses, conducted on rela- tively few samples, are calibrated to the infrared spec- tral data and predicted for all samples. Also, spectral indicators of soil condition are derived that success- fully screen soils into intact or degraded categories.
6. Compile standard data-entry sheets that can be enabled for Web-based data entry. The data are compiled in a central database. Individual users are provided with password access to their own data.
7. Use specialized statistical analyses for handling hier- archical data to derive population-based estimates for indicators of land condition and to analyze the effect of environmental covariates (for example, veg- etation cover and soil spectral indicators) at different spatial scales. Robust statistical inference mecha- nisms with spatial models, pedotransfer functions, and expert systems are under development.
8. Use the georeferenced sampling scheme to allow ground observations (for example, soil condition index) to be calibrated directly to satellite imagery and to be spatially interpolated and mapped.
9. Produce electronic atlases showing areas that are already degraded, areas at risk, and intact areas, with matched recommendations on intervention strategies.
10.Propose spatially explicit land management strategies for systematic testing (for example, enrichment plant- ing of trees to meet specific tree-density targets).
11.Through the sentinel sites, provide not only a frame- work for change detection through follow-up sur- veys (for example, after five years) but also a spatial platform for testing land management interventions.
For example, spatially distributing tree planting tri- als in the blocks ensures that species are tested over a wide range of land conditions, so that growth per- formance can be related back to site indexes, which can be used to predict tree performance at new sites.
Box 5.2 Steps in the Land Degradation Surveillance Framework
Source: International Centre for Research in Agroforestry.
Currently, no comparable system is in operation. The sys- tematic application of the approach will provide unbiased prevalence data on land degradation problems and permit quantification of land degradation risk factors, thereby enabling preventive and rehabilitative measures for SLM to be appropriately targeted. The approach provides a spatial framework for testing interventions in landscapes in a way that samples the variability in conditions, thereby increasing the ability to generalize from outcomes. The baseline that the protocol generates provides a scientifically rigorous platform for monitoring effects of intervention projects at a landscape level. The hierarchical sampling frame and statis- tical methods used allow systematic aggregation of results and population-level inferences to be made about land properties at different scales. The approach is particularly well suited to providing high-quality information at low cost in areas such as Sub-Saharan Africa, where existing data on land resources are sparse.
LESSONS LEARNED AND ISSUES FOR WIDER APPLICATION
The most difficult area for adoption is the advanced data analysis techniques used. An efficient solution to this barrier could be establishment of regional analytical centers, which would provide sampling schemes (global position system points, standardized forms, and protocols), as well as remote-sensing information and processing of field data posted by field teams on the Internet. In addition, the cen- ters would fulfill a technical and scientific capacity-building and support role.
INVESTMENT NEEDS AND PRIORITIES
Widespread application of this approach principally requires investment in capacity building of national teams in the approaches and methods. Operating costs for imple- menting a national surveillance system in the field are mod- est, and existing soil or natural resource survey departments could easily take up this role. The advanced data analysis techniques used are the most difficult area for adoption. An efficient solution to this barrier could be establishment of regional analytical centers that would provide sampling schemes (global position system points, standardized forms, and protocols); remote-sensing information; and process- ing of field data posted by field teams on the Internet. The centers would also fulfill a technical and scientific capacity-
building and support role. A government would need to take the following steps to implement a national-level sur- veillance program:
■ Provide exposure training in the approaches and meth- ods to a national team of scientists.
■ Equip a national soil laboratory with a near-infrared spectrometer (about US$75,000), provide basic training, ensure basic facilities for soil processing and storage, and provide limited conventional soil analysis.
■ Provide resources for two survey teams for about 12 months of fieldwork every five years (each team will need one surveyor and two field assistants, as well as a vehicle, a global positioning system, an auger set, and field operating funds) to establish sentinel sites (for example, 50 sites) throughout the country.
■ Train a national remote-sensing and geographic infor- mation system lab in data analytical techniques with sup- port from the regional surveillance center.
■ Orient national agronomic testing and socioeconomic research programs to work through the sentinel sites.
■ Establish additional sentinel sites for setting up baselines and monitoring outcomes for individual development projects aimed at land improvement.
REFERENCES
Shepherd, K. D., and M. G. Walsh. 2007. “Infrared Spec- troscopy—Enabling an Evidence-Based Diagnostic Sur- veillance Approach to Agricultural and Environmental Management in Developing Countries.” Journal of Near Infrared Spectroscopy15: 1–19.
Swift, M. J., and K. D. Shepherd, eds. 2007. Saving Africa’s Soils: Science and Technology for Improved Soil Manage- ment in Africa. Kenya, Nairobi: World Agroforestry Centre.
SELECTED READING
Infrared Diagnostics for Agriculture and the Environment.
2008. “Sensing Soil Condition: Infrared Diagnostics for Agriculture and the Environment.” World Agroforestry Centre, Nairobi. http://www.worldagroforestrycentre .org/sensingsoil/.
Vồgen, T-G., K. D. Shepherd, and M. G. Walsh. 2006. “Sens- ing Landscape Level Change in Soil Quality Following Deforestation and Conversion in the Highlands of Madagascar Using Vis-NIR Spectroscopy.” Geoderma 133: 281–94.
146 CHAPTER 5: RAINFED DRY AND COLD FARMING SYSTEMS
WEB RESOURCES
Global Digital Soil Map of the World. Global Digital Soil Map of the World project seeks to make a new digital soil map of the world using state-of-the-art and emerging tech- nologies for soil mapping and predicting soil properties at fine resolution. The map will be supplemented by interpretation and functionality options that aim to assist better decision-making in various global issues, such as food production and hunger eradication, climate
change, and environmental degradation: http://www .globalsoilmap.net/.
West Africa Drylands Project. The West Africa Drylands project emphasizes the application of science-based tools to help accelerate learning on sustainable dryland management and increase adaptive capacity at all scales, from local com- munities to regional and international policy bodies. Learn more about the project on its web site: http://www.world agroforestry.org/wadrylands/index.html.
INNOVATIVE ACTIVITY PROFILE 5.5: LAND DEGRADATION SURVEILLANCE 147