Are “large-scale, commercial agricultural” expansions possible in Nigeria?

by | December 12, 2013
Category: Blog

Modeling agricultural land expansions

There appears to be an on-going movement to expand “large scale, commercial agriculture (LSA)” in Africa that is partially based on a report that asks the question of whether agricultural expansions similar to those that were accomplished in the Brazilian Cerrado would be beneficial in Africa. See one coordinated response to the “Awakening Africa’s Sleeping Giant Report” here.

Ecologist colleague & collaborator, Stephen Porder from Brown Uni., articulates my own very conflicted position on this topic well in a recent article in the NY Times here. Also, quite telling comments with regard to the controversies around this topic can be found in the cross-section of reader responses to Stephen’s article.

While the technologies certainly exist to expand LSA on flat areas with circa >1000 mm/yr rainfall, circa <100 people/km2 population density and deep soils of suitable texture & mineralogy in Africa, there are two fundamental questions that relate directly to both the feasibility and desirability of such interventions, which currently cannot be answered with any degree of reliability:

  1.     Where are LSA expansions biophysically possible/viable in Nigeria … Africa, and
  2.     If such systems were to be expanded, would this lead to equitable and generally desirable (economic, social, health & environmental/ecological) outcomes on average?

Answering these two questions initially requires knowledge about the current extent of LSA land area and subsequently about the associated expansion/contraction dynamics (Question 1), which are what I will focus on here.

Answering Question 2 is much trickier, because it is really difficult to forecast in all of its dimensions for Nigeria and elsewhere in Africa, given the currently sparse supporting evidence from other continents.

The basic dynamics of a given LSA area of interest (e.g. the “Northern Guinea Savanna of Nigeria”) can be described concisely with the meta-population model of Levins (1970) as:


where P is the current proportion of LSA land per total land area (which is spatially uncertain at this point in time), e is an expansion rate variable (which is spatially & temporally uncertain) and c is a contraction rate variable (which is also spatially & temporally uncertain).

At a local equilibrium, the proportion of LSA land, P *< 1, is given by:


The P, e & c model parameters at present are/were determined by the ecosystem state factors of ecosystem formation … climate, organisms, terrain, soils and management locally; that is, at every individual site (pixel) under consideration. LSA would be unsustainable (i.e., fail over time) at a given location where c is greater than e, on average.

This simple model can be parameterized with data, and there are already efforts under way to do this, using volunteered, crowd-sourcing approaches (see the examples and publications at Geo-Wiki).

In AfSIS, we have recently generated a cultivated area map for Ethiopia (see Fig. 1) using the data generated by the IIASA-Geo-Wiki project and a geostatistical parameterization of the Levins model (for P), which uses MODIS and SRTM remote sensing covariate data, in addition to the crowd-sourced point data from Geo-Wiki as well as ground observations collected by EthioSIS. I will post a more detailed report, including R-code and data in the near future.

Figure 1. Predictions of the proportion of cultivated areas (pCultivated) of Ethiopia on a 1 km2 grid, based on IIASA GeoWiki data (available at: and AfSIS MODIS and SRTM remote sensing data compiled by AfSIS (available at: The larger map on the left represents our current predictions based on the most recently available data. The smaller map in the upper right represents the standard error (SE, or uncertainty) of those predictions. Note, that the Ethiopian government has not officially sanctioned this map.

Figure 1. Predictions of the proportion of cultivated areas (pCultivated) of Ethiopia on a 1 km2 grid, based on IIASA GeoWiki data (available at: and AfSIS MODIS and SRTM remote sensing data compiled by AfSIS (available at: The larger map on the left represents our current predictions based on the most recently available data. The smaller map in the upper right represents the standard error (SE, or uncertainty) of those predictions. Note, that the Ethiopian government has not officially sanctioned this map.

We are also preparing similar exercises for the other AfSIS priority countries including Nigeria, and we will attempt to map the proportion of LSA land (or P, in the model above) in this context. We hope to have a similar exercise completed for Nigeria by Jul. 2014.

Nigeria grids

Using data provided by NASA and WorldClim, AfSIS has constructed the data sets at resolutions from 100 to 1000 meters at various temporal ranges from the 1950’s to 2013. These processed data are being used to construct maps showing predictions and associated uncertainties of key soil properties across large regions where ground samples are not available. Where applicable, each data set is updated with new measurements every three months.  The workflows of download, calculation, mosaicking, geographic reprojection, resampling, and rescaling the original grids to derive the Africa continent-wide data sets have been automated using Unix and Geographic Resources Analysis Support System (GRASS) GIS.

The following section provides a preliminary overview the prevailing environmental conditions of Nigeria. I have split this into sections of indicators related to the state factors of ecosystem formation: Terrain, Climate, Primary Productivity, Land Surface Temperature and Soils that would be relevant for deciding where LSA might be biophysically possible in Nigeria. An additional set of variables is based on existing infrastructure of Nigeria (e.g. distance to populated places and roads, using GRUMP data at: All of the data are available at:, or as a processed GRASS GIS stack at <Dropbox>.

Terrain: LSA requires reasonably flat terrain to deploy tractors, combines and other farm machinery. Local differences in terrain also affect the hydrology and water balance of sites as well as soil erosion/deposition potentials. Fig. 2 presents examples of the “Shuttle Radar Topography Mission (SRTM) elevation data for Nigeria and Slope and Terrain Convergence Index (TCI, see: derivatives that can be calculated based on these data.

Figure 2. SRTM based terrain shade, slope and terrain convergence index (TCI) images for Nigeria.

Climate: Average annual precipitation and average annual air temperatures for Nigeria are shown in Fig. 3. These are based on WorldClim predictions (see: and cover the period between 1950-2000. We are in the process of updating the WorldClim rainfall predictions with NASA’s Tropical Rainfall Monitoring Mission data (see TRMM, Our TRMM derivatives can be downloaded at:

Figure 3. 1950-2000 Mean annual Precipitation (MAT) and Mean Annual Temperature (MAT) from WorldClim. Note that the thick black running through the northern portion of the MAP image on the left denotes the 1000 mm/year isohyet for these data.

Primary Productivity: Fig. 4 shows examples of long-term average terrestrial primary productivity indicators derived from MODIS time series (2000-2013). The fAPAR image on the right is being used to delineate “Primarily Vegetated Terrestrial Areas”.  The white spots on this image indicate urban areas, bare areas, and water bodies were agriculture is not likely to be possible.

Figure 4. 2000-2013 average Enhanced Vegetation Index (EVI) and fraction of Absorbed Photosynthetically Active Radiation (fAPAR) images, derived from MODIS time series.

Land surface temperature: The land surface temperature indicators shown in Fig. 5 represent the overall solar energy balance at the land surface. From the MODIS satellite’s perspective, the “surface” is whatever it sees when it looks through the atmosphere to the ground. It could be town, a pasture, a lake or a forest. Thus, land surface temperature is not the same as the air temperature in Fig. 3 above.

Figure 5. 2000-2013 average day-time land surface temperature (LST day) and diurnal (day-night) land surface temperature differences (LST dif) in degrees C, and derived from MODIS time series.

Soils: AfSIS has generated initial (and preliminary) digital soil mapping (DSM) predictions of some of the major soil properties for Nigeria, which may be of interest for national level agricultural planning (Fig. 6).

Figure 6. Topsoil (0-20 cm) soil property predictions (pH, sand content, soil organic carbon (SOC) & cation exchange capacity (CEC) for Nigeria based on soil legacy data (available at: and remote sensing grids (at:

While I expect that the prevailing soil conditions in the Northern Guinea Savanna of Nigeria to be quite different from those of the Brazilian Cerrado, I currently have no basis for comparing these two biomes, because similar maps for Brazil do not exist yet.

In general, the soils in the central latitudinal portion of Nigeria (presumably the Northern Guinea Savanna zone?) appear to be of low overall fertility and would require substantive rehabilitation with lime additions, mineral fertilizers and organic matter amendments.

Fig. 7 shows Nigeria’s road and settlement network from GRUMP at:

Figure 7. Nigeria road network from Red areas are within a 5 km distance buffer to a mapped road.

Figure 7. Nigeria road network from Red areas are within a 5 km distance buffer to a mapped road.

Some personal heuristics on LSA suitability in Nigeria (sort of an investor’s guide)

I am not currently buying or leasing land in Nigeria for large-scale, commercial agricultural production, but if I were to do that, I would initially explore the possibilities in places where the following simple heuristics apply:

  1. Land receiving more than 1000 mm of predicted rainfall per year on average, <and>
  2. Flat areas of less than 5% slope over a 1 km2 area, <and>
  3. Areas in places where fAPAR is currently > 0.25 on average, <and>
  4. Areas were sand content of soils is predicted to be less than 60%, <and>
  5. Areas were topsoil pH is predicted to be greater than 5.2 and less than 8, <and>
  6. Areas were topsoil organic carbon is predicted to be greater than 10 g/kg <and>
  7. Areas that are within a ~5 km from the major road network of the country <and>
  8. Areas currently unoccupied (or largely unsettled), potentially cultivable land.

By the same first-order logic (, I might apply a few other rules that reflect my personal priorities/preferences about criteria related to biodiversity, ecosystem service & conservation, ecological footprint and social equity considerations, but I in general think this first cut represents an initially suitable and balanced biophysical rule-set for evaluating the feasibility of LSA in African geographies. Of course, my simple heuristics should be evaluated further before committing private or public sector resource investments in this context! Also note, that some of the existing rule indicators (especially those related to soils) should really be verified with more ground-measurements, because they are quite uncertain.

With the exception of rule 8, the current data realizations that I have for Nigeria, are presented in Fig. 8. Other people might have different preferences (heuristics) that they would apply in this context that would broaden (or further narrow) the potential area of interest. My take on this is simply provided as an example the first-order logic that could be used to make such decisions … there are other possibilities (see e.g.

Figure 8. Areas that may be biophysically suitable for “large scale, commercial agriculture in Nigeria (in red).

Based on my admittedly fairly rough analysis, there are would circa 23,000 km2 that would be suitable for LSA in Nigeria. LSA might also be feasible in other portions of Africa and I shall explore those possibilities in subsequent blogs on this topic.

Submit Comment

Your email address will not be published. Required fields are marked *