There are some notable problems associated with using soil legacy data for soil mapping and surveillance. For the most part, soil legacy data were not purposely sampled to cover large areas using statistical sampling criteria and randomization procedures, and are thus not representative of the overall condition of soils in SSA. Most traditional soil surveys emphasized management invariant sub-soil properties that may not reflect changes in soil health and degradation. Over the past 25 years very few soil surveys have been conducted, and so many of the data are not adequately geo-referenced (e.g. they cannot be linked geographically to a particular pixel in a Landsat or SRTM image). Moreover, chemical and physical data from different countries or different survey campaigns are based on a wide variety of laboratory tests and analytical procedures, which are often difficult to harmonize from a diagnostic perspective. In addition, interpretation of soil tests is mostly based on expert opinion or response data from very few locations so that there are major uncertainties over the accuracy of the tests themselves.

We have undertaken a statistically designed baseline effort that uses standardized sampling and analytical procedures to facilitate a comprehensive assessment of soil health and degradation prevalence for an area that is representative of ~18.1M km2 of SSA.

Based on the Landsat World Reference System (WRS2) grid, we have drawn a randomized sample of 60 sentinel sites (i.e. 100 km2 areas, +60 alternate sites in case primary sites should prove to be inaccessible and/or insecure). Location of AfSIS Sentinel Landscapes

The sample was stratified (in proportion to size) based on the distribution of the major Köppen-Geiger climate zones: A  - the Equatorial zone, ~57% of the land area sampled with 34 sites (i.e. Landsat scenes, incl., zones Af, Am, As & Aw), B – the Semi-Arid zone, at ~21% of the land area (incl. zones BSh & BSk) sampled with 13 sites, and C – the Continental climate zone, at ~22% of the land area (incl. zones Cfa, Cfb, Cwa & Cwb), also sampled with 13 sites. Based on preliminary analyses the ISRIC-WISE (v.4) database of soils this sample is statistically representative of the range of soil variation that is likely to be observed in SSA.

The sample is also representative of the occurrence of different agricultural and cropping systems in SSA (notably, the AGRA cropping system impact zone maps are based on similar criteria). Using the CIESIN population databases, we estimate that the total year 2000 human population in Köppen zone A within the project area was ~340M people, Zone B had ~96M people and Zone C had ~176M people. Based on UnPop (2005) projections, the population in Köppen zone A is likely to increase to 756M by 2050, in zone B to ~173 and in zone C to ~405M people.

We have established 3 subregional field offices, in Arusha (Tanzania, at the Agricultural Research Institute, ARI), in Lilongwe (Malawi, at the Agricultural Research Service, ARS), and in Bamako (Mali, at the Institut d’Economie Rural, IER) to facilitate vehicle access to the selected sentinel sites, to provide Internet and other essential infrastructure services, and to establish regional soil laboratory infrastructure. Field office staff in Arusha are responsible for surveying sentinel and/or alternate sites located in Eastern Africa, the team based in Lilongwe is covering sites in Southern Africa, and the team based in Bamako is covering sites located in Western Africa.

Each regional team is covering ~5 sentinel sites per year (estimated at +/-2 months per site, depending on road access, terrain conditions, etc.).