The supply and demand of net primary production in the Sahel

Net primary production (NPP) is the principal source of energy for ecosystems and, by extension, human populations that depend on them. The relationship between the supply and demand of NPP is important for the assessment of socio-ecological vulnerability. We present an analysis of the supply and demand of NPP in the Sahel using NPP estimates from the MODIS sensor and agri-environmental data from FAOSTAT. This synergistic approach allows for a spatially explicit estimation of human impact on ecosystems. We estimated the annual amount of NPP required to derive food, fuel and feed between 2000 and 2010 for 22 countries in sub-Saharan Africa. When comparing annual estimates of supply and demand of NPP, we found that demand increased from 0.44 PgC to 1.13 PgC, representing 19% and 41%, respectively, of available supply due to a 31% increase in the human population between 2000 and 2010. The demand for NPP has been increasing at an annual rate of 2.2% but NPP supply was near-constant with an inter-annual variability of approximately 1.7%. Overall, there were statistically significant (p < 0.05) increases in the NPP of cropland (+6.0%), woodland (+6.1%) and grassland/savanna (+9.4%), and a decrease in the NPP of forests (−0.7%). On the demand side, the largest increase was for food (20.4%) followed by feed (16.7%) and fuel (5.5%). The supply-demand balance of NPP is a potentially important tool from the standpoint of sustainable development, and as an indicator of stresses on the environment stemming from increased consumption of biomass.


Introduction
Net primary production (NPP) represents the amount of atmospheric carbon that is fixed by vegetation during photosynthesis and accumulates as biomass. It represents the availability of carbon in the form of plant material for consumption as food, fuel and feed. NPP has been linked to crop yield (Tao et al 2005), biomass energy potential (Field et al 2008), forest production (Baisden 2006), and grazing resources (Leriche et al 2001). In sub-Saharan Africa NPP resulting from extensive agriculture and dry woodlands plays an important role in maintaining food security and providing household energy. However, these resources are under pressure from high population growth, which increases demand for NPP, and climate change, which impacts supply (IPCC 2014). The 'supply' of NPP is defined here as the annual amount of carbon stored in the ecosystem as plant tissue.
Semi-arid regions are particularly vulnerable to fluctuations in the supply of NPP due to its natural scarcity. In these regions, increased human activities are often driven by the demand for food, fuel and feed, which brings about changes that alter the spatial patterns of NPP such as cropland expansion, pasture creation, and fuelwood extraction. For example, large increases in demand for NPP lead to increased Environmental Research Letters Environ. Res. Lett. 9 (2014) 094003 (11pp) doi:10.1088/1748-9326/9/9/094003 Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. pasture stocking rates (Vargas et al 2009) resulting in soil compaction and vegetation defoliation (Hiernaux et al 1999). Additionally, crop yields, particularly in parts of sub-Saharan Africa, have been declining (Aune et al 2005, Chianu et al 2006 due to, among other things, reduction in soil fertility caused by continuous cropping to meet demand (Samaké et al 2005).
We use the terms 'demand' and 'consumption' to denote two different, but closely related, things. Demand for NPP is defined as the annual amount of carbon required to derive the food, fuel and feed necessary for human survival, and which also includes carbon lost during harvest and handling. The term 'consumption' refers to the actual amount of NPP that is extracted from the ecosystem as reported in the statistical database of the Food and Agriculture Organization of the United Nations (FAOSTAT).
Field estimation of NPP involves harvesting vegetation and calculating the annual growth of wood, the mass of foliage at peak leaf display, and the difference in the mass of tissue harvested at the beginning and end of the growing season (Schlesinger 1997). NPP varies spatially due to both environmental and anthropogenic factors and, because fieldwork is both labor and cost intensive, it would be exorbitantly expensive to conduct measurements over large extents. In this regard, data from satellite remote sensing platforms provide useful tools for quantifying the availability of NPP at regional scales.
We chose the Sahel region of Africa as an example of a complex, semi-arid, human-environment system where the supply-demand balance of NPP is in a constant flux. The Sahel has been subject to numerous droughts in the late 20th and early 21st centuries. Severe, region-wide, droughtinduced famines occurred in 1972-1973and 1984-1985(Sen 1983, Ibrahim 1988, and there were localized food shortages in 1990, 2002 and 2004. Recently, food production deficits were reported by several Sahelian countries in 2011/ 2012 and it is presently estimated that over 11 million people across the region are food insecure (United Nations 2013). A combination of unfavorable climatic conditions (Gonzalez et al 2012) and anthropogenic pressure (Sop andOldeland 2011, Hassan et al 2009) have caused declines in crop yield (Thornton et al 2008) and the diversity of Sahelian woody plant species (Wezel 2005).
The objectives of this paper are to quantify for the period 2000-2010: (1) the amount of NPP required to derive food, fuel and feed, (2) the fluctuations in the supply NPP, and (3) the percent of NPP demand relative to supply across the region. The chosen study period coincides with the availability of two open-access datasets (1) a continuous, highresolution, satellite-derived NPP dataset, and (2) a high resolution population distribution dataset covering three time slices: 2000, 2005, and 2010.

Study area
The study area covers the portion of the African continent between 5°and 25°north latitude with a focus on the Sahel, which is bounded by the 100 mm and 600 mm rainfall isohyets (figure 1). Northern parts of the area that border the Sahara Desert have a mean annual rainfall of less than 100 mm. To south of the Sahel are the humid savannas of the Sudano-Guinean zone that receive ample rainfall, between 600-1000 mm, enabling high vegetation productivity. Most of the region has one growing season, the start and length of which varies with latitude. The growing season in northern parts of the of the study area begins in July and ends in October; in the southern parts (<10°N) the season starts around April and lasts until November (Vrieling et al 2013).

NPP supply
We obtained the latest Collection 5.1 annual MODIS NPP data from 2000 to 2010 from the NASA Earth Observation System repository at the University of Montana (www.ntsg. umt.edu/). The MOD17 light use efficiency model (Heinsch et al 2003) calculates NPP as the difference between gross primary production (parameters within parentheses in equation (1)) and autotrophic respiration including growth and maintenance components: where NPP Supply is the NPP available in the ecosystem regardless of land use type; PAR is incoming photosynthetically active radiation; FPAR is the fraction of incident PAR absorbed by the vegetation canopy; ε max represents maximum light use efficiency under hypothetical biomespecific ideal conditions according to Monteith (1972); VPD and T min are environmental scalars that constrain light use efficiency and stand for vapor pressure deficit and minimum temperature, respectively; Ra (autotrophic respiration) represents maintenance and growth respiration of leaves, fine roots and woody tissue. Further descriptions of the MOD17 model are detailed in Running et al (2004), Zhao et al (2005) and Zhao and Running (2010). In order to delineate the land cover classes that source NPP for food, fuel, and feed, we obtained the latest Collection 5.1 of the MODIS Land Cover Type Product (MCD12C1) (Friedl et al 2010) from NASA's Land Processes Distributed Active Archive (LPDAAC) (https://lpdaac.usgs.gov). We used the International Geosphere Biosphere Program (IGBP) classification system and aggregated the 14 land cover classes within the study area into four land cover groupings (table 1S, supplementary information) based on ecophysiological similarities and their shared NPP sourcing capacity. For example, grasslands (<10% tree cover) and savannas (10-30% tree cover) were grouped together because both categories are predominantly composed of herbaceous cover that serve as grazing habitat for the region's livestock population. Similarly croplands and cropland/natural vegetation mosaics were grouped together into one category. We calculated NPP Supply trends for the region based on regression slopes resulting from ordinary least squares versus time. We then isolated areas that displayed significance (p < 0.05) and intersected them with reclassified grids of the land cover groupings. This procedure resulted in grids of statistically significant NPP Supply change for each land cover grouping. We also calculated annual NPP Supply anomalies for the 2000-2010 time period by subtracting the decadal average from each yearly total.

Population density
Gridded population data for the years 2000, 2005 and 2010 were downloaded from the WorldPop Project (http:// worldpop.org.uk/). The dataset is derived by using high resolution census data as a dependent variable in a Random Forest model along with a suite of independent variables that include land use and land cover, digital elevation, night-time lights, mean annual precipitation and mean annual temperature. Detailed descriptions of the population model are presented in Tatem

NPP demand
The Sahelian growing season coincides with the summer rains; crops are generally planted between May and July and harvested from October to November (USDA 1994). The NPP of croplands during the growing season represents the annual provision crops that are produced in the region. Sorghum, millet and maize, which together constitute 50% of Sahelian crop distribution (Leff et al 2004) are important both for human consumption and as fodder for livestock. Sahelian pastoralists depend on grassland and savanna productivity as forage for their livestock herds. Presently, 80% of the crop calories produced in the region are used for food, 10% for animal feed, and the remaining 10% for other uses (Cassidy et al 2013). In most parts of the Sahel, energy demand is met by woodfuels, and NPP provides biomass for energy in the form of fuelwood and charcoal.
We calculate demand for NPP (NPP Demand ) as the annual amount (in grams of carbon) of terrestrial carbon necessary to derive food, fuel and feed requirement in each country and account for carbon lost due to burning and harvest losses. Approximately 90% of the food produced in the region is consumed domestically (Gollin 2009, Barrett 2013, leaving 10% for other use. Therefore we constrained NPP Demand within each country in the region by using domestic production data from FAOSTAT (http://faostat3.fao.org/) (FAOSTAT 2013).
where NPP food is the NPP required to produce domestically consumed food items which includes meats sourced from six types of domestic animals (cattle, goats, sheep, pigs, camels, chickens) and two non-meat animal products (eggs and milk) (table 2S, supplementary information). Also included were 27 types of regionally important primary crops (table 3S, supplementary information). These crops represent 95% of all those that are domestically consumed by most of the countries in the study area (FAOSTAT 2013). A conceptual flowchart of the demand module is depicted in figure 1S in the supplementary information. NPP fuel represents fuelwood and charcoal extracted from the region's dry woodlands. Woodfuel requirement ranges from 55% in Senegal (Pires 2003) to over 90% in Chad (van der Plas and Abdel-Hamid 2005), but on average it is around 80% across the study area. NPP feed is the total amount of feed required to sustain the region's livestock population. This was calculated by converting the total number of heads of the six Sahelian domestic animal species to their equivalent Tropical Livestock Units (TLU), then multiplying by the annual feed requirement for each TLU (table 4S, supplementary information). Domestic human-driven NPP loss resulting from biomass burning of forest resources for land clearing is represented by NPP burned . Agricultural byproducts are represented by NPP residues and were calculated by applying harvest factors provided in literature (table 5S, supplementary  information).
Urban and rural consumption ratios vary due to differences in diet, lifestyle and wealth (Popkin 1999). We separated urban and rural areas by masking the urban extent of the WorldPop grids using data from the Global Rural-Urban Mapping Project (CIESIN et al 2011). We derived per capita NPP Demand by dividing total consumption with each country's population, and used the WorldPop data as a spatial surrogate to apply the per capita NPP Demand to grid cells. Consumption patterns of NPP in sub-Saharan Africa can differ considerably and depend on several factors such as income and product availability. We applied ratio factors for urban and rural consumption to national sums of each component of NPP Demand based on figures from the literature (Reardon 1993, Teklu 1996, Marufu et al 1997 figure 2(A)). The years 2002 and 2005 had the lowest total regional NPP Supply values with anomalies of −0.23 and −0.11 PgC, respectively ( figure 4(A)). The trend in NPP Supply between 2000 and 2010 for each of the 22 countries is given in table 1. There were significant (p < 0.05) region-wide increases in the NPP Supply of cropland (+6.0%), woodland (+6.1%) and grassland/ savanna (+9.4%) and a decrease in forest NPP Supply (−0.7%). NPP Supply in individual countries varied considerably; some such as Niger demonstrated large NPP Supply decreases in cropland (−6.7%), woodland (−20.8%) and grassland/savanna (−6.9%), while others such as Senegal had relatively substantial increases in the NPP Supply of cropland (+16.3%), woodland (+17.8%) and grassland/savanna (+14.4%) ( figure 2(B)). In 2002, the total region-wide NPP Supply dropped −9.7% below the decadal average due to droughts in Ethiopia, Eritrea, parts of Sudan, Senegal, Mali and Mauritania.

Population density
Between 2000 and 2010 the population of the region grew from 367 million to 471 million at an average rate of 2.8% per year, and is projected to increase to nearly one billion by 2050 (United Nations 2011). The increase was highest in Eritrea (43.2%) and lowest in Central African Republic (18.9%).
Most of the growth took place around and between existing settlements, creating networks that link different urban centers (figure 3).

NPP demand
The total regional NPP Demand between 2000 and 2010 ranged from 0.44 PgC to 1.13 PgC, representing between 19% and 41%, respectively, of NPP Supply ( figure 4(A)). Overall, NPP Demand increased 22% between 2000 and 2010 with the highest increase being for food (20.4%) followed by feed (16.7%) and fuel (5.5%) (table 2). We found large decreases in demand for woodfuel in the Sahelian countries of Eritrea (−23.1%) and Niger (−44%), which could be the result of initiatives promoting alternative fuels such as liquefied petroleum gas (LPG), provision of efficient cookstoves, and rural fuelwood markets aimed at sustainable forest management (Foley 1997

Discussion
At the present time around 41% of Sahelian NPP Supply is consumed by humans. The low regional inter-annual variability of NPP, at 1.7%, is representative of the global trend over the past 35 years (Running 2012) suggesting that the supply of NPP is at a near constant level. On the other hand, both population and NPP Demand are increasing at similar annual rates of 2.8% and 2.2%, respectively. This supply-demand relationship is sensitive to systemic shocks such as droughts or pest invasions that might lower the regional NPP Supply . For example, there were two years, 2002 and 2005, that exhibited large negative NPP Supply anomalies. The year 2002 was marked with a severe drought in Ethiopia and parts of Sudan that caused food and water deficit to over 12 million people (Balogun et al 2013) as well as localized food shortages in West Africa. In 2004, the largest desert locust (Schistocerca gregaria) invasion in 17 years, resulting from heavy rains the preceding year that created ideal breeding conditions, reduced the regional NPP Supply until the 2005 growing season (Ceccato et al 2006). These incidents illustrate the vulnerability of the regional NPP Supply to fluctuations in climate that can lead to declines in the availability of NPP for human consumption.
The spatial pattern of significant trends in the NPP Supply of the four land cover groupings that source food, fuel and feed to the region exhibited non-uniform patterns (figure 2(B)): cropland exhibited modest increases just south of the 600 mm rainfall isohyet; woodland exhibited increases that were mostly confined to the Central African Republic (CAR); grassland/savanna displayed no change in the Sahelian zone but increases south of the 600 mm rainfall isohyet, particularly in CAR, southern Chad and Sudan; forest displayed decreases everywhere except in CAR.
The overall increase in the demand for NPP fuel (table 2) generally triggers more fuelwood extraction unless there is a Table 2. Variability in the demand for food, fuel, feed and total demand between 2000 and 2010. Food includes crop-and meat-derived products; fuel is the household energy requirement in the form of woodfuel and charcoal; feed is the amount of NPP required to sustain the region's livestock population. Total demand includes the above as well as crop residues and burnt biomass. Total population refers to the percent change in a country's aggregated population (i.e. total persons) between 2000 and 2010 according to the WorldPop data.  (Bode 2005, Habtetsion andTsighe 2007), though statistics about the implementation of those plans are not available. Additionally, the Eritrean government had invested heavily in promotion of LPG as an alternative household energy source to the extent that LPG became cheaper than woodfuels for the purposes of cooking (Habtetsion and Tsighe 2007). This paper revealed that the Sahelian NPP supply and demand nexus is driven by the geographic distribution of population density and climatic variability, and projections for the future are alarming. A combination of increasing temperatures and changes in the precipitation regime will likely decrease the NPP Supply of important crops in the region as well as increase the prevalence of pest infestations (IPCC 2014). Forecasts of Sahelian rainfall for the next four decades are inconsistent (Ben Mohamed 2011), and the bestcase scenarios predict that the region will receive more rainfall, which should cause an increase in NPP Supply . However, the impact of higher temperatures may counteract this effect, producing a net reduction in NPP Supply across the region (Delire et al 2008). Similarly, forecasts of reduced crop yields of up to −41% due to increased temperature have been consistent across various studies despite projected increases in rainfall (Jones and Thornton 2003, Wolfram and David 2010, Sultan et al 2013. These factors point to a complex system in a delicate balance between human requirement on the one hand, and the ecosystem's capacity to satisfy it on the other. Only areas with a statistical significance of p < 0.05 are shown here. Large areas, most notably in West Africa and Ethiopia, experience NPP Demand that is much larger than available supply.

Uncertainties and limitations
Uncertainties in the estimates of NPP Supply stem from MOD17's implementation of water stress control through atmospheric VPD. Mu et al (2007) demonstrated that VPD underestimates water stress and overestimates NPP Supply in Sahelian conditions. Pan et al (2006) have shown that correcting for soil water content improves MOD17 NPP estimates. Another source of uncertainty in NPP Supply stems from the static biome-specific values of the maximum light use efficiency (ε max , see section 2.2) that MODIS implements, which have been found to be considerably lower than field estimated values (Sjöström et al 2013).
We acknowledge the limitations presented by using data from FAOSTAT, which is sometimes challenged either due to shortage of resources to perform meticulous surveys or over-/ under-reporting of national statistics by certain countries (Krausmann et al 2008). Despite these shortcomings, FAO-STAT represents the only available, widely used, resource that compiles, cross-checks and standardizes global agricultural data (Goudriaan et al 2001, Ciais et al 2005, Ramankutty et al 2008, Ma et al 2012. We, therefore, consider it acceptable for the purposes of this study. A drawback to employing annual land cover groupings to represent exploitable resources is the complete dependence on the MODIS Land Cover Type Product. Although the latest version of the product (Collection 5.1) has undergone several refinements and possesses an overall accuracy of 75% (Friedl et al 2010), the amount of training data from the Sahel that the algorithm employs is considerably small (n = 58) and for one class (woody savanna) it is as low as a single site.

Conclusion
In this paper, we quantified the spatiotemporal variation of supply of NPP in relation to anthropogenic demand for a portion of sub-Saharan Africa using a combination of satellite remote sensing and socioeconomic data. We identified region-wide spatial patterns of annual NPP extraction to derive food, fuel and feed from the available supply.
This research adds to a series of studies on Sahelian vegetation dynamics with a focus on the relationship between anthropogenic exploitation and supply of NPP. The combination of rapidly increasing demand for food, fuel and feed driven by population growth makes the region vulnerable to climatic changes that may alter the per capita availability of NPP. Additionally, the region's location in a geographic transition zone between the arid Sahara desert and the high NPP Sudano-Guinean zone increases its sensitivity to fluctuations in rainfall. Investigating inter-annual variations of NPP supply and demand could help explain patterns of ecosystem change by identifying areas under anthropogenic pressure. Considering the current 2.2% annual increase in NPP Demand , there is a risk that ecosystems may not be able to provide food, fuel and feed for the region's humans and livestock without a corresponding increase in NPP Supply . If future droughts occur at similar climatic magnitudes as the ones that took place in the 1970s and 1980s the Sahel will be at risk of mega famines.
The integrated analysis of the supply and demand of NPP allows the identification of areas where demand for carbon exceeds available supply, either presently or in future scenarios. Hence, the methodology presented in this paper can be applied to other human-environment systems where the supply-demand balance of NPP approaches critical levels.