## 1. Introduction

Human activity is having unprecedented influence within the global land system. Over 80% of ice-free land has been altered by human impact (Sanderson et al., 2002), changing atmospheric compositionChapin et al., 2000Millenium Ecosystem Service, 2005Foley et al., 2005). This ‘human domination of the earth's ecosystems’ (Vitousek et al., 1997) has led to reconceptualisation of humans as integral components and engineers of the global biosphere (Ellis and Ramankutty, 2008) and the recognition of a need for novel integrated approaches, breaking down the historic barriers between natural and social science, to better understand the drivers and consequences of land use change (Kates et al., 2001Turner et al., 2007Hackman et al., 2014).
Human Appropriation of Net Primary Productivity (HANPP), the proportion of annual plantbiomass production co-opted by human activity, was first suggested as a measure of land use intensity by Vitousek et al. (1986). Land use intensity is a complex and multidimensional concept (Erb et al., 2013), and the advantage of HANPP compared to simpler metrics such as fertiliser inputs (Potter et al., 2010) or crop output (Monfreda et al., 2008) is that it is intrinsically socioecological, encompassing the interactions between human livelihoods and an ecological energy flux. Early studies quantified the annual extraction of NPP embodied in resources such as crops, livestock feed and timber as between 20 and 40% of annual global NPP (Vitousek et al., 1986Rojstaczer et al., 2001Imhoff et al., 2004), before Haberl et al. (2007) developed the HANPP concept further by expressing HANPP as a proportion of the potential NPP in a system undisturbed by human influence, thereby including resource extraction but also losses or gains in NPP caused by human land use change (such as deforestation or intensive agriculture). Using the latter approach, HANPP was calculated as 23.8% of potential global terrestrial NPP in the year 2000 (Haberl et al., 2007).
Land use intensification is a subject of particular research interest in sub-Saharan Africa for several reasons. Firstly, HANPP has increased more steeply in Africa over the last century than on any other continent (Krausmann et al., 2013), but yields of staple crops remain far below potential levels (Licker et al., 2010Sanchez, 2010). Secondly, Africa has been identified as a hotspot of potential new agricultural land (Ramankutty et al., 2002Deininger et al., 2011Lambin and Meyfroidt, 2011; although see Young, 1999Chamberlin et al., 2014), but agricultural expansion involves numerous conflicts – both social, such as poor recognition of land rights resulting in displacement of rural communities (Cotula et al., 2009), and ecological, such as the situation of much potential agricultural land in highly biodiverse regions (Gibbs et al., 2010). Thirdly, and at a more local scale, rural subsistence livelihoods in Africa are often centred on crop production, livestock rearing, and collection of wild-sourced resources such as firewood and wild foods (Angelsen et al., 2014) and as such are directly linked and highly sensitive to changes in ecosystem properties. Whether approached from ‘bottom-up’ livelihoods and development perspectives or from ‘top-down’ global change and conservation perspectives, understanding the processes and impacts of land use intensification in sub-Saharan Africa is therefore a research priority.
To date there has been little use of HANPP as a land use intensity measure in sub-Saharan Africa. HANPP has been quantified at continental (Fetzel et al., 2016) and regional scales (Abdi et al., 2014), but most national case studies have been carried out in Europe (e.g. Schwarzlmuller, 2009Musel, 2009Kolheb and Kraussman, 2009) and Asia (Prasad and Badarinth, 2004Kastner, 2009Chen et al., 2015), with the only national HANPP case study in Africa focused on South Africa (Niedertscheider et al., 2012). One recent study (Bartels et al., 2017) adapted the HANPP framework to the village scale in Tanzania and calculated village-level HANPP to be between 34 and 38% of annual potential NPP, but without further case studies it is impossible to determine whether this is a representative example. The lack of fine scale HANPP research in Africa may be a consequence of data availability, with commonly used data sets such as FAOSTAT (FAO, 2015a) and the Global Forest Resources Assessment (FAO, 2015b) having well-recognised weaknesses in the African context (Mather, 2005Fetzel et al., 2016). Previous studies may also have underestimated HANPP in rural Africa due to lack of recognition of the importance of wild-sourced resources such as firewood and construction material in rural livelihoods – a recent global analysis found that around 30% of household income in rural Africa is derived from such wild-sourced resources (Angelsen et al., 2014), but only a minority of studies have attempted to include domestic fuelwood consumption (Niedertscheider et al., 2012Fetzel et al., 2016Bartels et al., 2017), and resources such as construction poles have been largely omitted from existing analyses. Additionally, the low resolution and simplifying assumptions of published studies have potentially obscured important heterogeneity in parameters such as forest structure and population distribution. Drawing linkages between HANPP and ecosystem goods and services such as biodiversity demands a finer resolution of analysis (an approach adopted by Haberl et al., 2004).
Analysis of NPP appropriation patterns at the household scale also has the potential to improve understanding of the social and environmental consequences of changing rural livelihoods. Many studies have documented the income portfolios of rural African households (e.g. Cavendish, 2000Mamo et al., 2007Kamanga et al., 2009), but fewer have considered how livelihood strategies and socio-economic characteristics influence household-scale environmental or NPP footprints. Further, past studies indicate that wealthier households have higher absolute environmental income (Cavendish, 2000Mamo et al., 2007), partly driven by ability to obtain a higher share of the most lucrative environmental resources (Ambrose-Oji, 2003De Merode et al., 2004), but no research has assessed whether this pattern of ‘elite capture’ of environmental goods is replicated in NPP appropriation, or whether the greater capability of wealthy households to access resources or displace NPP demand during periods of scarcity results in exacerbation of rural NPP appropriation inequalities during land use intensification.
Reflecting the research gaps described above, the first objective of this study is to develop a novel method of quantifying HANPP at the community scale and to calculate HANPP in six villages in central Zimbabwe. Avoiding the inaccuracies associated with the use of national statistics, we instead base our analysis on detailed field data describing woodland structure and rural livelihoods. Our second objective is to adapt this method to the household scale, in order to assess the extent to which household wealth and income may be associated with household NPP appropriation and to discuss the potential ramifications of land use change for inequalities in NPP appropriation.

## 2. Methods

Our methods consist of field data collection in rural Zimbabwe and the development of a new method quantifying HANPP at village and household levels. We begin by describing the study site and field data collection methods. We then detail our village-level HANPP calculations, and finish by describing how we investigated the relationship between household wealth and appropriated NPP.

### 2.1. Study Site

The study was conducted in the Communal Area on and around Wedza Mountain, in the Mashonaland East province of Zimbabwe (Fig. 1). Due to traditional sacred values and past legal restrictions (Gumbo, 1988), Wedza Mountain has maintained high biomass woodland while the surrounding lowlands have become mainly deforested (Hansen et al., 2013). Woodland in the study area is of the dry miombo type, characterised by Brachystegia spiciformisBboehmii and Jglobiflora. Wedza Mountain is in agroecological region 2b of Zimbabwe, with estimated rainfall of 750–1000 mm yr− 1 (Vincent and Thomas, 1960).

Three pairs of villages were selected for inclusion in the study: Makumbe and Pfende in the mainly deforested lowlands with households 3 and 5 km west of the mountain woodland; Mapfanya and Betera on the western fringes of the mountain woodland; and Charambira and Mbizi on the more remote eastern side of the mountain. Livelihoods in the area are centred on subsistence agriculture, livestock husbandry (primarily cattle, goats and chickens), and casual day labour.

### 2.2. Field Data Collection

#### 2.2.1. Land Cover Maps

Two participatory mapping group discussions and four transect walks were carried out in May 2014 in each of the six villages, and these resulted in six locally understood land cover categories (Table 1). These locally derived mapping data were combined with Google Earth satellite imagery in QGIS (QGIS, 2016) to create village land cover maps (Supporting information Fig. 1; SI Table 1). The extent of cover by high biomass mountain woodland was cross-checked using maps developed by Hansen et al. (2013).
Table 1
Land cover categoryDefinition
Mountain woodlandComparatively undisturbed miombo woodland found on Wedza Mountain
Lowland woodlandMore disturbed lowland woodlands, found in village grazing areas and on riverbanks/kopjes. Also long-term abandoned fields with significant tree regrowth.
Wet grasslandsSeasonally dry (locally termed bani) and non-seasonal (dofonya) wetlands
CroplandsActive and recently fallowed fields
GardensSmall fields, usually found in wetlands or along riverbanks, used for growing vegetables.
Residual areaAreas without vegetation cover, such as roads, household yards and borrow pits left following road construction.
Ecological survey plots were established in the three land cover categories with greatest spatial extent. Five plot locations were randomly generated in QGIS in each of three land cover categories (mountain woodlands, lowland woodland and croplands) in each of the six villages, giving a total of 30 plots each for lowland woodlands and croplands and 20 plots for mountain woodland (the villages of Makumbe and Pfende have no mountain woodland area). Plots were inventoried between February and April 2015, using circular plots of 20 m radius. In each plot, diameter at breast height (DBH: measured at 1.3 m) and local vernacular name were recorded for all stems with DBH ≥ 3 cm. Where possible, names in the local Shona language were translated to scientific names in the field using Mullin (2006) or Hyde et al. (2016), and identification checked using Coates Palgrave (2002). Specimens of species unknown to research assistants and Shona ethnospecies without a previously recorded scientific translation were taken for identification at the National Herbarium of Zimbabwe in Harare.

#### 2.2.2. Household Survey

To understand household livelihood strategies, a household survey was undertaken using stratified random sampling. Household lists were generated during participatory mapping exercises, and in each village the households were selected proportional to sample size, selection stratified into three categories of household size (1–2 residents, 3–5 residents and 6 + residents) and three categories of household head (male-headed, widow-headed, and de facto female-headed with husbands working away; categories follow Cavendish, 2000) to give a total sample size of 104 households. High population mobility in Zimbabwe due to economic instability resulted in high survey attrition compared to similar studies, resulting in a final sample size of 91 households. Village size ranged from 10 to 53 permanently inhabited households (mean of 33 households) and sampling intensity ranged from 37 to 80% (mean of 52%).
The questionnaire was adapted from the CIFOR-PEN prototype questionnaire (CIFOR-PEN, 2008) and used to collect detailed data on use of wild-sourced resources (such as firewood and wild fruits) in the month preceding the questionnaire. Data were also collected on household assets and all other income sources, including crops, livestock, informal labour and remittances. The questionnaire was used three times to capture seasonal variation in livelihood strategies (June/July 2014, February/March 2015 and October/November 2015).

### 2.3. Quantifying HANPP

Following Haberl et al. (2007) and Haberl et al. (2014), we define HANPP as:
$\text{HANPP}={\text{HANPP}}_{\mathrm{luc}}+{\text{HANPP}}_{\text{harv}}$
where HANPPluc is the loss of potential NPP due to land use change and HANPPharv is NPP harvested by humans. HANPPharv is further subdivided into used extraction (consumed by humans) and unused NPP, meaning NPP influenced by human activity but not extracted from the ecosystem such as unrecovered crop residues.
HANPPluc is calculated as:
${\text{HANPP}}_{\mathrm{luc}}={\mathrm{NPP}}_{\mathrm{pot}}-{\mathrm{NPP}}_{\mathrm{act}}$

where NPPpot is the potential NPP in a hypothetical undisturbed system, and NPPact is the actual NPP of the prevailing human-altered system. Reflecting the greater uncertainty associated with calculating below-ground NPP, we follow a number of previous studies by focusing solely on aboveground HANPP (Prasad and Badarinth, 2004Kolheb and Kraussman, 2009Vačkář and Orlitova, 2010), hereafter aHANPP. All results are presented in tonnes of dry matter.

#### 2.3.1. Potential Aboveground NPP (aNPPpot)

It is assumed that areas currently covered by woodland (both mountain and lowland), fields and residual area would, in the absence of human activity, be covered by undisturbed miombo woodlands. aNPP of undisturbed savanna woodland is the sum of annual woody growth, leaf production and understorey grass production. The area cover of all land cover types in each village is provided in the supporting information (SI Table 1).
Ten mountain woodland plots with minimal evidence of disturbance were used as proxies for undisturbed woodlands. Current plot stem biomass in undisturbed plots was determined using stem biomass allometric equations derived from similar dry miombo ecosystems (Grundy, 1995Chidumayo, 1997Ryan et al., 2011). Three different sets of annual woody increment estimates (Chidumayo, 1988Frost, 1996Flack, 2013) were used to project DBH one year in the future, and annual woody production determined by calculating plot biomass from the projected DBH values and deducting current standing biomass. Leaf production estimates were calculated using two leaf production equations (Chidumayo, 1997Frost, 1996). The non-contiguous canopy in savanna woodlands also allows persistence of a grass understorey (Parr et al., 2014). As this was not measured directly in the field, expected annual grass production per hectare in relation to tree basal area was identified using Frost (1996; p26). Full details of all allometric equations, woody growth increments and annual grass production estimates can be found in the supporting information (SI Table 2).
There is comparatively little data on annual increments and leaf production in dry miombo systems, so all six possible combinations of increment and leaf production equation were calculated to give upper and lower aNPPpot estimates. Mean woody and leaf production in tonnes of dry matter ha− 1 yr− 1 were calculated for the ten reference plots to give the aNPPpot of undisturbed woodlands, and this was multiplied by the number of hectares of woodland, fields and residual area in each village (identified from village land cover maps) to give aNPPpot of undisturbed woodland per village.
All study villages also contain areas of wet grassland which would have low tree cover even in an undisturbed state, and total village aNPPpot is therefore the sum of woodland aNPPpot and grassland aNPPpot. aNPPpot of wet grassland was assumed to be equal to peak annual grass biomass. Mean peak grass biomass was estimated from five studies in Zimbabwean and Zambian wet grassland (Scoones, 1991Hoffa et al., 1999; Jeanes & Baars, 1991 inScholes et al., 1996Shea et al., 1996Nyamadzawo et al., 2014; see SI Table 3) and gave a mean value of 4.2 ± 0.6SE t DM ha− 1 yr− 1. This aNPPpot was multiplied by the number of hectares of wet grassland and gardens in each study village, and added to the aNPPpot of the woodland area to give total village aNPPpot.

#### 2.3.2. Actual Aboveground NPP (aNPPact)

The aNPPact for woodland area was calculated by applying the method described above to the ten plots in each land cover type in each pair of adjacent study villages, giving separate annual NPP estimates for mountain woodland, lowland woodland, and trees on agricultural land in each village pair. These woodland aNPPact values were multiplied by the area cover in hectares of the relevant land cover type in each village. Reflecting poor data availability on NPP in disturbed vs. undisturbed wet grasslands, wet grassland NPPact was assumed to be the same as it would be in the altered landscape, so the same aNPP value of 4.2 ± 0.6 t DM ha− 1 was applied to wetland areas.
NPPact of crop production was determined by calculating village production of eight key crops (maize, milletsorghum, rice, sugar beans, cowpeas, sunflower and leafy green vegetables). Groundnuts and sweet potatoes are also grown in the study area, but were excluded from the analysis as the majority of biomass production is below-ground. Mean per capita production of each of the crops over the last three harvest seasons was calculated for study households in each village and used to scale crop production to the village level using village household composition lists. As survey households within each village represent a broad range of socioeconomic and agroecological conditions, we believe these village level production estimates to be robust. Locally reported units were converted to kilograms using USDA (1992) and adjusted to dry matter using appropriate moisture content estimates (Gebhardt and Thomas, 2002OMAFRA, 2016). Associated crop residues and pre-harvest crop losses were calculated from crop yield data using harvest factors from Haberl et al. (2007) (SI Table 4).
Grass production in agricultural land is focused on contour ridges; raised boundaries between fields intended to prevent soil erosion which are also an important source of livestock feed (Scoones, 1995). However, reflecting the small area coverage of contour ridges (< 2 ha per village) and the lack of data on contour ridge NPP, contour ridge grass production was not included in the analysis.

#### 2.3.3. Harvested NPP (aHANPPharv)

aHANPPharv was calculated as the sum of crop production and recovered residues, material grazed by livestock, and consumption of three wild-sourced resources accounting for the highest proportion of extracted biomass (firewood, construction poles and thatching grass).
For crops, aHANPPharv was assumed to be equal to aNPPact. The proportion of crop residues recovered for use was calculated using conversion factors in Haberl et al. (2007).
Size of the village livestock herd was calculated by estimating ownership of the main livestock species (cattle, goats, chickens, turkeys and guinea fowl) in the village from the household survey. The proportion of livestock feed derived from the environment (the feed gap) was calculated by estimating annual feed demand using daily food intake estimates from Haberl et al. (2007) and deducting dry matter mass of feed crops, purchased concentrate, and the proportion of crop residues used as livestock feed (estimated as 41% in sub-Saharan Africa by Haberl et al., 2007).
Annual firewood consumption was estimated from the three months incorporated in the household questionnaire, including both firewood used domestically and for commercial purposes such as beer brewing and tobacco curing. Firewood consumption was recorded in local units, and following a review of the literature (see SI Table 5) and conversion to dry matter weight using moisture content estimates from Abbott and Lowore (1999) (SI Table 6), headloads were assigned a weight of 11 kg DM, wheelbarrows 20 kg DM, and cartloads 158 kg DM.
Volume of wood required annually for construction and maintenance of household structures (wooden huts, fences and cattle kraals) was calculated using volumes reported in Grundy et al. (1993) and converted to dry matter weight using published wood density values for the most prevalent local construction species (Goldsmith and Carter, 1981Malimbwi et al., 1994Abbott and Lowore, 1999Williams et al., 2008Chave et al., 2009). Firewood used for brick burning to construct household structures was not included, as the lifespan of brick buildings means that firewood demand is very low when expressed on an annual basis.
Annual thatching grass consumption was estimated from the household survey. Following Grundy et al. (2000), we assume thatching grass bundles to weigh 5 kg (fresh weight). Thatching grass is mainly collected in the early dry season, when grass moisture content is estimated to be 55% (Woollen et al., 2016). Consumption of firewood, construction materials and thatching grass was calculated on a mean per capita basis for study households in each village and then scaled to the non-surveyed households in the village.
A key consideration in calculating aHANPP is that estimates of aNPPpot, aNPPact and aHANPPharv all apply to the same spatially bounded area, in this case the village. Village boundaries were determined during participatory mapping groups and confirmed during four GPS-tracked transect walks with key informants in each village. Use of village land cover maps and village household survey data meant that aNPPpot, aNPPact and aHANPPharv embodied in crops could all be reliably calculated within village boundaries, while data collected during the household survey on the source location of wild-sourced resources meant that estimates of aHANPPharv of firewood and construction materials could be restricted to reflect only aNPP appropriated within the village area. There is however some uncertainty over the proportion of cattle graze sourced within village boundaries. Cattle are herded during the farming season from October to May and stay primarily within the village area, while following the harvest from May to July field crop residues are a main food source, giving cattle little motivation to roam. The only season when cattle roam further is the later dry season from July to September when food becomes scarcer; however, Scoones (1995) found that cattle roaming distance is related to distance to permanent water source. As all study villages have permanent water sources, and as roaming distance is limited by the return of all cattle to the homestead each night due to fears of theft or predation by hyenas, we have assumed for this analysis that all environmental livestock feed is derived from within the village area. The limitations of this assumption are considered in the discussion.
Unrecovered crop residues and pre-harvest crop losses were included in aHANPPharv but reported as unused extraction. Although human-caused veld fires were common in the study area, there are no accurate data available at sufficiently high resolution; we therefore follow Niedertscheider et al. (2012) in omitting biomass changes caused by human-induced fires from the analysis.

### 2.4. Household Wealth and aNPP Appropriation

Twelve interviews were carried out, two in each village, to identify features that indicated whether households were very poor, less poor, or wealthy by local standards. Interview respondents were purposively sampled: one high income and one low income householdidentified from the household survey in each village to obtain a broad range of perspectives, and including only long-term village residents with a good knowledge of all other households in the village.
Wealth indicator interview responses were combined to give a wealth index with seven categories (Table 2). Indicators which were locally important but which were linked directly to NPP consumption such as cattle ownership were not included in the index. Surveyed households were assigned a wealth index between 0 and 7.
Table 2
CategoryVery poor (assigned score of 0)Less poor (assigned score of 0.5)Locally wealthy (assigned score of 1)
Farming equipmentNo large farming equipmentOwn two or more of:
plough, wheelbarrow, scotch cart
Fulfil ‘less poor’ criteria, and also own one or more of:
harrow, cultivator, planter, tractor
TransportNo form of transportBicycleCar
Household structuresNo large bedroom house, only one or two roundhouse kitchensMain house with 1–3 roomsMain house with 4 or more rooms
SanitationNo toiletToilet
Household furnishingsNo expensive furnishingsOwn two or more of the following:
Fulfil ‘less poor’ criteria and also own two more of:
TV, beds (2 or more), generator, lounge suite
Domestic helpNo domestic workerDomestic worker employed at any time during study period
Water supplyUse shared water sources such as wells or boreholesPrivate well, borehole or water pump
Household level aHANPPluc was calculated by deducting household crop production and aNPP of trees in fields from the aNPPpot of the land area belonging to the household. Respondents could not provide reliable estimates of the area in hectares of household land holdings and so were instead asked for the number of fields owned. Number of fields was then multiplied by 0.4 ha, this being the mean area of 10 fields measured using QGIS and Google Earth imagery. aHANPPharv was calculated as the sum of crop and crop residue production, livestock environmental feed demand, and wild-sourced resource consumption.
Restricting aHANPPharv to only the area owned by the household would seriously underestimate household consumption of NPP embodied in firewood and livestock feed, and the method presented here therefore represents the household ‘footprint’ of aNPP appropriated within the village area. However, it should be noted that there are no suitable data available to calculate the contributions of individual households to changes in the aNPPact of common property woodlands, which may result in underestimates of household aHANPPluc.
Linear regression was used to examine the relationship between aNPP appropriation and three different wealth measures: wealth index, household cash income (calculated from the household survey), and the median of household rank by these two measures (hereafter the combined wealth rank). Households were also split into two groups, those belonging to the three villages with highest land use intensity (Makumbe, Pfende and Mapfanya) and those with the lowest land use intensity (Betera, Charambra and Mbizi). The ratio of total aNPP appropriation by households with a combined wealth rank in the top and bottom 20% of each village set was compared to assess NPP appropriation inequality. The same method was used to calculate inequality in cash income. All analyses were carried out in Excel and R (R Core Team, 2014).

## 3. Results

### 3.1. aNPPpot and aNPPact in Miombo Woodland Systems

aNPPpot calculated from undisturbed woodland reference sites ranged from 3.6 ± 0.2 t DM ha− 1 yr− 1 to 6.0 ± 0.2 t DM ha− 1 yr− 1 + dependent on the combination of leaf and increment equations used (Fig. 2), with these aNPP estimates being within the range of published studies (SI Table 7).

Annual aNPPact was highest in mountain woodland plots, with the six combinations of increment and leaf equations giving a mean annual aNPP of 4.7 ± 0.4 t DM ha− 1 yr− 1 in the Charambira/Mbizi village pair and 3.8 ± 0.3 t DM ha− 1 yr− 1 in Mapfanya/Betera (Fig. 3). Increased grass production only partially compensated for loss of tree productivity in more disturbed lowland woodlands. Tree aNPPact in croplands was almost twice as high in Charambira/Mbizi as in the other two village pairs, but even in these villages amounted to only 0.5 ± 0.05 t DM ha− 1 yr− 1.

### 3.2. aHANPP at the Village Scale

aHANPP ranged from 113% in Makumbe village to 48% in Charambira village (Table 3). aHANPPharv and aHANPPluc made an equal contribution to aHANPP in Mapfanya and Makumbe villages, whereas aHANPPluc accounted for a much higher proportion of aHANPP in the other four villages. Although different equation combinations resulted in quite high levels of variation in estimated appropriation of aNPP in tonnes of dry matter, aHANPP expressed as a percentage of aNPPpot was much less sensitive, with a maximum difference of 10% between upper and lower estimates. Although the choice of equation used altered the relative contributions of aHANPPharv and aHANPPluc, in four of the study villages there is no overlap in the ranges of the potential percentage contributions of aHANPPluc and aHANPPharv. Therefore the conclusion that aHANPPluc is the greater contribution to HANPP in these villages remains robust whichever equation combination is used. Compared to the findings of previous African HANPP studies (Table 4), aHANPP was substantially higher in each of the six villages.
Table 3
VillageVillage area
(ha)
aHANPP
(%)
aHANPP
(t DM yr− 1)
aHANPPluc
(%)
aHANPPharv
(used, %)
aHANPPharv
(unused, %)
Makumbe368113
(110–118)
1897
(1548–2294)
56
(46–65)
56
(43–68)
1.6
(1.2–1.9)
Pfende32384
(80–86)
1253
(915–1623)
59
(48–69)
25
(19–32)
0.5
(0.4–0.6)
Mapfanya44172
(68–78)
1461
(1125–1809)
35
(26–41)
36
(27–46)
0.8
(0.6–1.0)
Betera63653
(50–55)
1571
(1120–2045)
37
(30–42)
16
(12 − 20)
0.4
(0.3–0,6)
Charambira24948
(43–50)
548
(374–726)
31
(22–37)
16
(12–20)
0.2
(0.2–0.3)
Mbizi26858
(54–60)
716
(511–929)
39
(30–44)
19
(14–23)
0.4
(0.3–0.5)
Table 4
RegionHANPP estimate (%)Reference
Sub-Saharan Africa12.4aImhoff et al. (2004)
Sub-Saharan Africa18Haberl et al. (2007)
South Africa21–25Niedertscheider et al. (2012)
Africa20Krausmann et al. (2013)
Southern Africa23Fetzel et al. (2016)
Ololosokwan village, Tanzania34–38Bartels et al. (2017)
a
Imhoff et al. (2004) express HANPP as the used proportion of NPPact, while all other studies define HANPP as human appropriation of NPPpot.
Livestock grazing was the main contributor to aHANPPharv, accounting for between 42 and 66% of total aHANPPharv (Fig. 4). Collection of firewood and construction materials accounted for between 21% and 31% of aHANPPharv, while crop production accounted for a relatively small proportion, between only 9 and 20%

### 3.3. aNPP Appropriation and Household Wealth

There were significant relationships between aNPP appropriation and household wealth index score and combined household ranking, but no correlation with log cash income per capita (Table 5a). There was no apparent relationship between household aHANPPluc and any wealth indicator, whereas there was a significant relationship between aHANPPharvand all wealth indicators. However, the low adjusted R2 in all cases (Table 5) indicates that there are numerous factors other than wealth influencing variation in aNPP appropriation.
Table 5
Wealth index scoreLog cash income per capitaCombined ranking
(a)
aHANPP
(t DM)
0.21⁎⁎0.020.12⁎⁎
aHANPPluc
(t DM)
0.00.010.0
aHANPPharv
(used fraction, t DM)
0.34⁎⁎0.040.20⁎⁎

(b)
aHANPPharv
(livestock feed)
0.32⁎⁎0.040.21⁎⁎
aHANPPharv
(crops)
0.30⁎⁎0.030.15⁎⁎
aHANPPharv
(firewood/construction)
0.00.00.0

(c)
Agricultural efficiency
(crop production t DM field− 1)
0.0060.010.04
Household field holdings0.020.00.01
Significance levels:

p < 0.05.
⁎⁎
p < 0.001.
When aHANPPharv was disaggregated by source category there was a significant positive correlation between wealth indicators and appropriation of aNPP embodied in livestock feed and crops, but no apparent relationship between any wealth indicators and appropriation of aNPP embodied in environmental resources (Table 5b). Increasing household wealth was associated with increased ownership of livestock, particularly cattle and poultry, but was not associated with increased field holdings (Table 5c).
In the higher land use intensity villages (Makumbe, Pfende and Mapfanya) the top 20% of households by combined wealth ranking appropriated on average 30.8 ± 4.1 t DM hh− 1 yr− 1, while the poorest 20% of households appropriated on average 18.4 ± 2.9 t DM hh− 1 yr− 1. In the lower land use intensity villages (Betera, Charambira and Mbizi), households in the wealthiest 20% appropriated on average 27.2 ± 4.2 t DM hh− 1 yr− 1 while households in the poorest 20% appropriated 13.0 ± 2.5 t DM hh− 1 yr− 1. The ratio of total combined NPP appropriation by the richest 20% of households to that appropriated by the poorest 20% of households was 1.7 in the higher land use intensity villages and 2.1 in lower land use intensity villages, indicating slightly higher inequality in HANPP appropriation in the lower land use intensity villages.
Inequality in HANPP consumption was much lower than inequality in cash income. In the higher land use intensity villages, the ratio of cash income in the wealthiest 20% of households by combined wealth ranking compared to the poorest 20% was 23.3, with the wealthiest 20% earning US$1912.91 ± 570 per capita yr− 1 compared to US$82.09 ± 13 per capita yr− 1. Even excluding the wealthiest household, which had cash income per capita six times that of the next wealthiest household, the ratio of cash income in the wealthiest and poorest 20% was still 15.8. In the three lower land use intensity villages the equivalent ratio was only 8.0, with the wealthiest 20% of households earning US$1079.14 ± 249 per capita yr− 1 compared to US$134.82 per capita yr− 1 in the poorest 20%.

## 4. Discussion

### 4.1. aHANPP at the Village Scale

Our findings indicate that aHANPP quantified at the village level is much higher than would be anticipated from previous studies. aHANPP varied from 48% in Charambira up to 113% in Makumbe, whereas previous regional studies reported a range from 12.4 to 23.0% (Imhoff et al., 2004Haberl et al., 2007Niedertscheider et al., 2012Krausmann et al., 2013Fetzel et al., 2016) and the only previous village scale study estimated aHANPP as between 34 and 38% (Bartels et al., 2017). Even in the higher resolution maps developed by Haberl et al. (2007), the majority of Zimbabwe has HANPP of between 20 and 40%, with few areas exceeding 50%.
There are several potential reasons behind the discrepancy between our results and those of published studies. Partly the high aHANPP is attributable to the choice of study site. The heterogeneity of land use in Africa is well-recognised, with some areas being underutilised while others are densely populated by smallholder farming communities (Tittonell and Giller, 2013Chamberlin et al., 2014Jayne et al., 2014). In Zimbabwe, these densely populated areas are a legacy of past inequity in land ownership, with many indigenous black farmers restricted to crowded Communal Areas while the best agricultural land was conserved for large commercial white farms (Palmer, 1990). While land reform has resulted in some population re-distribution (Scoones, 2010), the historic land ownership system still shapes the extant landscape. Although focusing on a Communal Area will undoubtedly result in higher aHANPP, there is a strong argument that, as rural populations in more marginal agricultural areas are most sensitive to environmental change (Jones and Thornton, 2009), these areas should be a priority for land use change research.
However, our higher aHANPP is also driven by a number of other factors. Firstly, use of field data rather than national statistics allowed inclusion of resources such as firewood and construction material which are only poorly recorded in national level data. Firewood, construction poles and thatching grass accounted for between 21 and 31% of total aHANPPharv, with average per capita extraction totalling 1.1 t DM yr− 1. Excluding this resource flow could therefore result in significant underestimates of aHANPPharvparticularly in rural areas of developing countries. Use of locally derived woodland survey data also meant that we could account for the contribution of woodland degradation to aHANPPluc. The majority of published studies assume aNPPpot and aNPPact to be equal in forest and woodland systems – and in rangeland systems in the case of Bartels et al. (2017) – but our data show human disturbance results in substantial variation in aNPPactover even relatively fine spatial scales. While recognising that the resource intensity of our approach would be challenging if seeking to ascertain HANPP at wider spatial scales, we suggest that detailed field data on forest and woodland structure and on rural livelihood portfolios is key to improving the accuracy of higher resolution aHANPP estimates.
The finding of extremely high aHANPP (113%) in Makumbe village has two potential explanations. The first is that the estimate includes some aNPP appropriated outside the village area due to livestock grazing. As detailed in the methods, observations of local herding patterns suggest the majority of livestock feed to be obtained within the village, but without a more detailed analysis of livestock movement patterns there is uncertainty attached to this assumption. However, even were it very conservatively assumed that only 50% of livestock feed was obtained inside the village area, aHANPP would still be high at 91%. This leads us to suggest that the high aHANPP observed in several villages is also due to harvest of production from previous years embodied in firewood and construction poles. The high environmental pressure and unsustainable use implied by a HANPP estimate of > 100% is supported by the levels of woodland degradation observed in the highest land use intensity villages.

### 4.2. Uncertainty in aHANPP Estimates

The main source of potential error in the presented method is in the estimation of aNPPpotand aNPPact in miombo woodlands. There have been few longitudinal studies of annual production in miombo woodlands, but our estimates of annual woody aNPPpot of woody production lie within the range of published studies (see SI Table 7). There are few published studies of annual leaf production in miombo woodland, and use of the allometric equations developed in Zambia by Chidumayo (1997) indicate much higher annual leaf production than that predicted by the Zimbabwean equations from Frost (1996). However, the mid-point of the two leaf production calculations used in this study falls within the range of published estimates.
An additional potential source of overestimation is that miombo woodland production is linked to precipitation levels (Frost, 1996), and although all allometric equations used in this study were derived from dry miombo systems, woodlands in Mozambique and Zambia typically receive higher rainfall than Zimbabwean woodlands and may have corresponding differences in growth rate and in proportional relations between DBH and leaf production. As there are no allometric equations developed within the study site, using a range of equations derived from similar systems was the only way to assess the scale of uncertainty introduced by choice of allometric. Furthermore, it could be argued that use of local reference plots may have resulted in overestimation of woodland production as some plots showed signs of minor disturbance. However, given that miombo woodland evolved in a context of disturbance, either by humans or by fire and megaherbivore activity (Mapaure and Moe, 2009), the use of mildly disturbed reference plots is of less substantial concern. We also do not account for the possibility that there could have been areas which were naturally clear of vegetation even in the undisturbed landscape.
Despite these limitations, our village level HANPP estimates appear robust. The different combinations of woody increment and leaf production equations resulted in varying estimates of woodland production in tonnes of dry matter, and altered the relative contributions of aHANPPharv and aHANPPluc to total aHANPP, but in no village did the final HANPP percentage estimate have a range of > 10 percentage points with all equation combinations. This indicates that aHANPP has low sensitivity to equation choice and allows a high level of confidence in the results.
An additional critique of our methods might relate to our focus on aboveground NPP. Miombo woodland soils are an important carbon store (Walker and Desanker, 2004) and changes in belowground production could significantly impact HANPPluc, while inclusion of key local crops with primarily belowground production such as groundnuts and sweet potatoes could alter both NPPact and HANPPharv. A valuable development on the present study would therefore be to explore methods of integrating belowground production into HANPP estimates.

### 4.3. aNPP Appropriation and Household Wealth

Household wealth index score was positively associated with aNPP embodied in harvested resources, partially supporting the idea that elite capture of aNPP occurs in rural Zimbabwean communities. Our data do not allow us to isolate the reasons behind the link between household wealth and crop harvests. Wealth may be a direct driver of high crop harvests, reflecting the ability of wealthier households to afford inputs such as synthetic fertiliser and paid labour (Zingore et al., 2007), or alternatively both wealth and crop harvest may co-vary with another factor such as soil fertility in household fields. Higher extraction of livestock fodder by wealthier households reflects the significant correlation of wealth with number of cattle owned. Cattle are an important multifunctional asset in many areas of Africa, used for ploughing fields and pulling carts (and thereby helping perpetuate wealth accumulation), producing manure for fertiliser (improving soil quality of private fields), and acting as a status indicator and savings bank (Dercon, 1998Hoddinott, 2005).
The lack of a clear link between household wealth and aHANPPluc is interesting. A global study found that national HANPPluc decreases with increasing development, as increasing agricultural yield compensates for loss of NPPpot in undisturbed ecosystems (Krausmann et al., 2013). However, although wealthier households in Wedza obtained higher overall crop production, this appears to be due to increased area cultivated as there was no significant relationship between household wealth and yield. Further, while cropland area cultivated each year by the household may increase with wealth, there was no significant relationship between wealth and the total cropland area owned, suggesting that land ownership in the communal area is also linked to a diversity of other factors such as length of time resident in the community, gender of the household head, and number of times the land has been divided amongst family members. These longer terms factors are important as the lowlands around Wedza Mountain have been largely deforested for over 30 years (Gumbo, 1988). Our data also only permitted calculation of aHANPPluc on household field holdings, and identifying methods of quantifying household contributions to aHANPPlucbeyond the boundaries of household property (for example due to woodland degradation through firewood collection) should be a priority for future studies, particularly in areas such as southern Africa where there is high dependence on common property resources.
Also somewhat surprising is the finding that inequality of aNPP appropriation is slightly higher in the combined households from the three lower land use intensity villages than in the three highest land use intensity villages. However, one limit of applying the HANPP paradigm at a small scale is that it only records NPP appropriated within the village area. At the national scale, wealth is associated with an ability to displace demand for natural resources (Krausmann et al., 2009Weinzettel et al., 2013). Without more detailed household consumption data we cannot calculate the quantity of displaced HANPP, but there is evidence that households in the highest land use intensity villages do displace a portion of their NPP appropriation: 15 out of 36 households in Makumbe and Pfende reported collecting or purchasing firewood outside their home village, as opposed to only one household in total out of the other four study villages. Obtaining firewood outside the study area in many cases requires either a cash payment or the possession of assets such as wooden cart and cattle, both of which are linked to household wealth. A further interesting development on the current analysis would be to determine whether inclusion of NPP embodied in resources obtained outside the study area alters the NPP appropriation inequality findings of the current study.

## 5. Conclusions

This study deploys a new approach to the assessment of HANPP at the household and village level, yielding new evidence in the study of socio-ecological interdependences and resources inequalities in small scale African farming systems. The findings from this study suggest that the low resolution of previous HANPP studies has resulted in a substantial underestimation of the intensity of land use in smallholder farming areas in southern Africa, masking the ramifications of land use intensification for rural livelihoods and the conservation of biodiversity and natural capital.
Our findings indicate that high-resolution calculations of HANPP based on field data can make a valuable contribution to understanding of patterns of land use intensity, improving the accuracy of HANPP estimates by facilitating inclusion of resources omitted from many studies such as construction materials, and also allowing finer-scale analysis of human impacts on ecosystems such as woodland degradation which may not be apparent from broader scale data sets. We suggest that such high resolution approaches mapping HANPP over larger areas can make a valuable contribution to identification of ‘hotspots’ of environmental pressure, and that linking HANPP patterns at community and household scales to characteristics of local livelihoods may assist anticipation of environmental externalities associated with livelihood change.

## Acknowledgments

We are very grateful to Nyaradzo Shayanewako, Charles Shayanewako, Wadzenai Shayanewako, Pauline Dhliwayo, Amai B Chiraya and Tapiwa Betera for their invaluable assistance with fieldwork. Particularly we wish to thank Nyaradzo and Charles Shayanewako, for allowing R. Pritchard to stay with their family while in Zimbabwe. We are grateful to the University of Zimbabwe, the Research Council of Zimbabwe, and the National Herbarium of Zimbabwe, for their support of this project. We thank the two anonymous reviewers whose comments and suggestions greatly improved the manuscript. Finally, we thank the District Administrators and Headman of Wedza District, and the Sabhukus (village heads) and residents of the six study villages.

## Funding

R. Pritchard is supported by a Natural Environment Research Council (NERC) PhD studentship, with additional funding provided by the Elizabeth Sinclair Irvine Fund and the Centenary Agroforestry 89 Fund. C.M. Ryan and I. Grundy were supported by the ACES project NE/KO10395 which is funded with support from the  (ESPA) programme. The ESPA programme is funded by the Department for International Development (DFID), the Economic and Social Research Council (ESRC) and NERC. See more at: www.espa.ac.uk.

## Appendix A. Supplementary Data

Additional detail on methods used to calculate Human Appropriation of Net Primary Productivity at the village scale.

## References

Abbott and Lowore, 1999
P.G. Abbott, J.D. LoworeCharacteristics and management potential of some indigenous firewood species in Malawi
For. Ecol. Manag., 119 (1999), pp. 111-121
Abdi et al., 2014
A.M. Abdi, J. Seaquist, D.E. Tenenbaum, et al.The supply and demand of net primary production in the Sahel
Environ. Res. Lett., 9 (2014), Article 094003
Ambrose-Oji, 2003
B. Ambrose-OjiThe contribution of NTFPs to the livelihoods of the ‘forest poor’: evidence from the tropical forest zone of south-west Cameroon
Int. For. Rev., 5 (2003), pp. 106-117
Angelsen et al., 2014
A. Angelsen, P. Jagger, R. Babigumira, et al.Environmental income and rural livelihoods: a global comparative analysis
World Dev., 64 (2014), pp. S12-S28
Bartels et al., 2017
L.E. Bartels, A. Mayer, K.-H. ErbExploring potential socio-ecological impacts of changes to the Loliondo Gamed Controlled Area, Northern Tanzania: the case of the pastoral village Ololosokwan
J. Land Use Sci., 12 (2017), pp. 87-103
Cavendish, 2000
W. CavendishEmpirical regularities in the poverty-environment relationship of rural households: evidence from Zimbabwe
World Dev., 28 (2000), pp. 1979-2003
Chamberlin et al., 2014
J. Chamberlin, T.S. Jayne, D. HeadeyScarcity amidst abundance? Reassessing the potential for cropland expansion in Africa
Food Policy, 48 (2014), pp. 51-65
Chapin et al., 2000
F.S. Chapin, E.S. Zavaleta, V.T. Eviner, et al.Consequences of changing biodiversity
Nature, 405 (2000), pp. 234-242
Chave et al., 2009
J. Chave, D. Coomes, S. Jansen, et al.Towards a worldwide wood economics spectrum
Ecol. Lett., 12 (2009), pp. 351-366
Chen et al., 2015
A. Chen, R. Li, H. Wang, et al.Quantitative assessment of human appropriation of aboveground net primary production in China
Ecol. Model., 312 (2015), pp. 54-60
Chidumayo, 1988
E.N. ChidumayoEstimating fuelwood production and yield in regrowth dry miombo woodland in Zambia
For. Ecol. Manag., 24 (1) (1988), pp. 59-66
Chidumayo, 1997
E. ChidumayoMiombo Ecology and Management: An Introduction
Stockholm Environment Institute, London, UK (1997)
CIFOR-PEN, 2008
CIFOR-PENPEN prototype questionnaire, version 4.4
Coates Palgrave, 2002
M. Coates PalgraveTrees of Southern Africa
(Third edition), Random House Struick, Cape Town, South Africa (2002)
Cotula et al., 2009
L. Cotula, S. Vermeulen, R. Leonard, et al.Land grab or development opportunity?
Agricultural Investment and International Land Deals in Africa, IIED/FAO/IFAD, London/Rome (2009)
De Merode et al., 2004
E. De Merode, K. Homewood, G. CowlishawThe value of bushmeat and other wild foods to rural households living in extreme poverty in Democratic Republic of Congo
Biol. Conserv., 118 (2004), pp. 573-581
Deininger et al., 2011
K. Deininger, D. Byerlee, J. Lindsay, et al.Rising Global Interest in Farmland: Can It Yield Sustainable and Equitable Benefits?
World Bank, Washington D.C, USA (2011)
Dercon, 1998
S. DerconWealth, risk and activity choice: cattle in Western Tanzania
J. Dev. Econ., 55 (1998), pp. 1-42
Ellis and Ramankutty, 2008
E.C. Ellis, N. RamankuttyPutting people on the map: anthropogenic biomes of the world
Front. Ecol. Environ., 8 (2008), pp. 439-447
Erb et al., 2013
K.-H. Erb, H. Haberl, M.R. Jepsen, T. Kuemmerle, M. Lindner, D. Müller, P.H. Verburg, A.ReenbergA conceptual framework for analysing and measuring land-use intensity
Curr. Opin. Environ. Sustain., 5 (2013), pp. 464-470
FAO, 2015a
FAOFAOSTAT Statistics Database
FAO, Rome, Italy (2015)
FAO, 2015b
FAOGlobal Forest Resources Assessment 2015
FAO, Rome, Italy (2015)
Fetzel et al., 2016
T. Fetzel, M. Niedertscheider, H. Haberl, et al.Patterns and changes of land use and land-use efficiency in Africa 1980–2005: an analysis based on the human appropriation of net primary production framework
Reg. Environ. Chang., 16 (2016), pp. 1-14
Flack, 2013
S. FlackSpecies Composition and Interspecific Variation in Growth; Implications for Above Ground Carbon Sequestration in a Miombo Woodland
(Unpublished MSc Thesis)
University of Edinburgh (2013)
Foley et al., 2005
J. Foley, R. DeFries, G. Asner, et al.Global consequences of land use
Science, 309 (2005), pp. 570-574
Frost, 1996
P. FrostThe ecology of miombo woodlands
B. Campbell (Ed.), The Miombo in Transition: Woodlands and Welfare in Africa, CIFOR, Bogor, Indonesia (1996), pp. 11-55
Gebhardt and Thomas, 2002
S.E. Gebhardt, R.G. ThomasNutritive Value of Foods
USDA, Washington D.C., USA (2002)
Gibbs et al., 2010
H.K. Gibbs, A.S. Ruesch, F. Achard, et al.Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s
PNAS, 107 (2010), pp. 16732-16737
Goldsmith and Carter, 1981
B. Goldsmith, D. CarterThe Indigenous Timbers of Zimbabwe
Forestry Commission, Harare, Zimbabwe (1981)
Grundy, 1995
I. GrundyRegeneration and Management of Brachystegia spiciformis Benth. and Julbernardia globiflora(Benth.) Troupin in Miombo Woodland, Zimbabwe
(D. Phil Thesis
University of Oxford (1995)
Grundy et al., 1993
I. Grundy, B.M. Campbell, S. Balebereho, et al.Availability and use of trees in Mutanda Resettlement Area, Zimbabwe
For. Ecol. Manag., 56 (1993), pp. 243-266
Grundy et al., 2000
I. Grundy, J. Turpie, P. Jagger, et al.Implications of co-management benefits from natural resources for rural households in north-western Zimbabwe
Ecol. Econ., 33 (2000), pp. 369-381
Gumbo, 1988
D. GumboAn integrated land-use plan for Wedza Mountain, Zimbabwe
Mt. Res. Dev., 8 (1988), pp. 289-295
Haberl et al., 2004
H. Haberl, N.B. Schulz, C. Plutzar, et al.Human appropriation of net primary production and species diversity in agricultural landscapes
Agric. Ecosyst. Environ., 102 (2004), pp. 213-218
Haberl et al., 2007
H. Haberl, K.H. Erb, F. Krausmann, et al.Quantifying and mapping the human appropriation of net primary production in earth's terrestrial ecosystems
PNAS, 104 (2007), pp. 12942-12947
Haberl et al., 2014
H. Haberl, K.-H. Erb, F. KrausmannHuman appropriation of net primary production: planets, trends and planetary boundaries
Annu. Rev. Environ. Resour., 39 (2014), pp. 363-391
Hackman et al., 2014
H. Hackman, S. Moser, A.L. St ClairThe social heart of global environmental change
Nat. Clim. Chang., 4 (2014), pp. 653-655
Hansen et al., 2013
M.C. Hansen, P.V. Potapov, R. Moore, et al.High-resolution global maps of 21st-century forest cover change
Science, 342 (2013), pp. 850-853
Hoddinott, 2005
J. HoddinottShocks and their consequences across and within households in rural Zimbabwe
J. Dev. Stud., 42 (2005), pp. 301-321
Hoffa et al., 1999
E. Hoffa, D. Ward, W. Hao, et al.Seasonality of carbon emissions from biomass burning in a Zambian savanna
J. Geophys. Res., 104 (1999), pp. 13841-13853
Hyde et al., 2016
M. Hyde, B. Wursten, P. Ballings, et al.Flora of Zimbabwe
www.zimbabweflora.co.zw (2016), Accessed 23rd Sep 2016
Imhoff et al., 2004
M.L. Imhoff, L. Bounouna, T. Ricketts, et al.Global patterns in human consumption of net primary production
Nature, 429 (2004), pp. 870-873
Jayne et al., 2014
T.S. Jayne, J. Chamerblin, D.D. HeadeyLand pressures, the evolution of farming systems, and development strategies in Africa: a synthesis
Food Policy, 48 (2014), pp. 1-17
Jones and Thornton, 2009
P.G. Jones, P.K. ThorntonCroppers to livestock keepers: livelihood transitions to 2050 in Africa due to climate change
Environ. Sci. Pol., 12 (2009), pp. 427-437
Kamanga et al., 2009
P. Kamanga, P. Vedeld, E. SjaastadForest incomes and rural livelihoods in Chiradzulu District, Malawi
Ecol. Econ., 68 (2009), pp. 613-624
Kastner, 2009
T. KastnerTrajectories in human domination of ecosystems; human appropriation of net primary production in the Philippines during the 20th century
Ecol. Econ., 69 (2009), pp. 260-269
Kates et al., 2001
R.W. Kates, W.C. Clark, R. Corell, et al.Sustainability Science
Science, 27 (2001), pp. 641-642
Kolheb and Kraussman, 2009
N. Kolheb, F. KraussmanLand use change, biomass production and HANPP: the case of Hungary 1961–2005
Ecol. Econ., 69 (2009), pp. 292-300
Krausmann et al., 2009
F. Krausmann, H. Haberl, K.H. Erb, et al.What determines geographical patterns of the global human appropriation of net primary production?
J. Land Use Sci., 4 (2009), pp. 15-33
Krausmann et al., 2013
F. Krausmann, K.H. Erb, S. Gingrich, et al.Global human appropriation of net primary production doubled in the 20th century
PNAS, 110 (2013), pp. 10324-10329
Lambin and Meyfroidt, 2011
E.E. Lambin, P. MeyfroidtGlobal land use change, economic globalisation, and the looming land scarcity
PNAS, 108 (2011), pp. 3465-3472
Licker et al., 2010
R. Licker, M. Johnston, J.A. Foley, et al.Mind the gap: how do climate and agricultural management explain the ‘yield gap’ of croplands around the world?
Glob. Ecol. Biogeogr., 19 (2010), pp. 769-782
Malimbwi et al., 1994
R.E. Malimbwi, B. Solberg, E. LuogaEstimation of biomass and volume in miombo woodland at Kitulangalo Forest Reserve, Tanzania
J. Trop. For. Sci., 7 (1994), pp. 230-242
Mamo et al., 2007
G. Mamo, E. Sjaastad, P. VedeldEconomic dependence on forest resources: a case from Dendi District, Ethiopia
Forest Policy Econ., 9 (2007), pp. 916-927
Mapaure and Moe, 2009
I. Mapaure, S.R. MoeChanges in the structure and composition of miombo woodlands mediated by elephants (Loxodonta africana) and fire over a 26-year period in north-western Zimbabwe
Afr. J. Ecol., 47 (2009), pp. 175-183
Mather, 2005
A.S. MatherAssessing the world's forests
Glob. Environ. Chang., 15 (2005), pp. 267-280
Millenium Ecosystem Service, 2005
Millenium Ecosystem ServiceEcosystems and Human Well-being: Synthesis
Island Press, Washington D.C., USA (2005)
Monfreda et al., 2008
C. Monfreda, N. Ramankutty, J.A. FoleyFarming the planet: 2. Geographic distribution of crop areas, yields, physiological types, and net primary production in the year 2000
Glob. Biogeochem. Cycles, 22 (1) (2008)
Mullin, 2006
L.J. MullinA New Zimbabwean Botanical Checklist of English and African Names
(1st edition), The Tree Society of Zimbabwe, Harare (2006)
Musel, 2009
A. MuselHuman appropriation of net primary production in the United Kingdom, 1800–2000 Changes in society's impact on ecological energy flows during the agrarian-industrial transition
Ecol. Econ., 69 (2009), pp. 270-281
Niedertscheider et al., 2012
M. Niedertscheider, S. Gingrich, K.H. ErbChanges in land use in South Africa between 1961 and 2006: an integrated socio-ecological analysis based on the human appropriation of net primary production framework
Reg. Environ. Chang., 12 (2012), pp. 715-727
G. Nyamadzawo, M. Wuta, J. Nyamangara, et al.Burning, biomass removal and tillage effects on soil organic carbon and nutrients in seasonal wetlands (Dambos) of Chiota smallholder farming area, Zimbabwe
Arch. Agron. Soil Sci., 60 (2014), pp. 1411-1427
OMAFRA, 2016
OMAFRA
http://www.omafra.gov.on.ca/english/index.html (2016), Accessed 29th Sep 2016
Palmer, 1990
R. PalmerLand reform in Zimbabwe, 1980–1990
Afr. Aff., 89 (1990), pp. 163-181
Parr et al., 2014
C.L. Parr, C.E. Lehmann, W.J. Bond, W.A. Hoffman, A.N. AndersenTropical grassy biomes: misunderstood, neglected, and under threat
Trends Ecol. Evol., 29 (2014), pp. 205-213
Potter et al., 2010
P. Potter, N. Ramankutty, E. Bennett, S.D. DonnerCharacterising the spatial patterns of global fertiliser application and manure production
Earth Interact., 14 (2010), pp. 1-22
V.K. Prasad, K.V.S. BadarinthLand use changes and trends in human appropriation of above ground net primary production (HANPP) in India (1961–1998)
Geogr. J., 170 (2004), pp. 51-63
QGIS Development Team, 2016
QGIS Development TeamQGIS geographic information system. Open Source Geospatial Foundation project
R Core Team, 2014
R Core TeamR: A Language and Environment for Statistical Computing
R Foundation for Statistical Computing, Vienna, Austria (2014)
Ramankutty et al., 2002
N. Ramankutty, J.A. Foley, J. Norman, et al.The global distribution of cultivated lands: current patterns and sensitivity to possible climate change
Glob. Ecol. Biogeogr., 11 (2002), pp. 377-392
Rojstaczer et al., 2001
S. Rojstaczer, S.M. Sterling, N.J. MooreHuman appropriation of photosynthesis products
Science, 294 (2001), pp. 2549-2552
Ryan et al., 2011
C.M. Ryan, M. Williams, J. GraceAbove- and belowground carbon stocks in a Miombo woodland landscape of Mozambique
Biotropica, 43 (2011), pp. 423-432
Sanchez, 2010
P.A. SanchezTripling crop yields in tropical Africa
Nat. Geosci., 3 (2010), pp. 299-300
Sanderson et al., 2002
E.W. Sanderson, M. Jaiteh, M.A. Levy, et al.The human footprint and the last of the wild
Bioscience, 52 (2002), pp. 891-904
Scholes et al., 1996
R.J. Scholes, J. Kendall, C.O. JusticeThe quantity of biomass burned in southern Africa
J. Geophys. Res., 101 (1996), pp. 23667-23676
Schwarzlmuller, 2009
E. SchwarzlmullerHuman appropriation of aboveground net primary production in Spain, 1955–2003: an empirical analysis of the industrialisation of land use
Ecol. Econ., 69 (2009), pp. 282-291
Scoones, 1991
I. ScoonesWetlands in drylands: key resources for agricultural and pastoral production in Africa
Ambio, 20 (1991), pp. 366-371
Scoones, 1995
I. ScoonesExploiting heterogeneity: habitat use by cattle in dryland Zimbabwe
J. Arid Environ., 29 (1995), pp. 221-237
Scoones, 2010
I. ScoonesZimbabwe's Land Reform: Myths and Realities
James Currey/Boydell & Brewer, Suffolk, UK (2010)
Shea et al., 1996
R.W. Shea, B.W. Shea, J.B. Kaufmann, et al.Fuel biomass and combustion factors associated with fires in savanna ecosystems of South Africa and Zambia
J. Geophys. Res., 101 (1996), pp. 23551-23568
Tittonell and Giller, 2013
P. Tittonell, K.E. GillerWhen yield gaps are poverty traps: the paradigm of ecological intensification in African smallholder agriculture
Field Crop Res., 143 (2013), pp. 76-90
Turner et al., 2007
B.L. Turner, E.F. Lambin, A. ReenbergThe emergence of land change science for global environmental change and sustainability
PNAS, 104 (2007), pp. 20666-20671
USDA, 1992
USDAWeights, Measures, and Conversion Factors for Agricultural Commodities and Their Products
USDA, Washington D.C., USA (1992)
Vačkář and Orlitova, 2010
D. Vačkář, E. OrlitovaHuman appropriation of aboveground photosynthetic production in the Czech Republic
Reg. Environ. Chang., 11 (2010), pp. 519-529
Vincent and Thomas, 1960
V. Vincent, R.G. ThomasAn Agro-ecological Survey of Southern Rhodesia Part 1: Agro-ecological Survey
Government Printers, Salisbury, Southern Rhodesia (now Harare, Zimbabwe) (1960)
Vitousek et al., 1986
P.M. Vitousek, P.R. Ehrlich, A.H. Ehrlich, et al.Human appropriation of the products of photosynthesis
Bioscience, 36 (1986), pp. 368-373
Vitousek et al., 1997
P.M. Vitousek, H.A. Mooney, J. Lubchenco, et al.Human domination of the earth's ecosystems
Science, 277 (1997), pp. 494-499
Walker and Desanker, 2004
S.M. Walker, P.V. DesankerThe impact of land use on soil carbon in miombo woodlands of Malawi
For. Ecol. Manag., 203 (2004), pp. 345-360
Weinzettel et al., 2013
J. Weinzettel, E.G. Hertwich, G.P. Peters, et al.Affluence drives the global displacement of land use
Glob. Environ. Chang., 23 (2013), pp. 433-438
Williams et al., 2008
M. Williams, C.M. Ryan, R.M. Rees, et al.Carbon sequestration and biodiversity of re-growing miombo woodlands in Mozambique
For. Ecol. Manag., 254 (2008), pp. 145-155
Woollen et al., 2016
E. Woollen, C.M. Ryan, S. Baumert, et al.Charcoal production in the mopane woodlands of Mozambique: what are the trade-offs with other ecosystem services?
Philos. Trans. R. Soc. B, 371 (2016), Article 20150315
Young, 1999
A. YoungIs there really spare land. A critique of estimates of available cultivable land in developing countries
Environ. Dev. Sustain., 1 (1999), pp. 3-18
Zingore et al., 2007
S. Zingore, H.K. Murwira, R.J. Delve, et al.Influence of nutrient management strategies on variability of soil fertility, crop yields and nutrient balances on smallholder farms in Zimbabwe
Agric. Ecosyst. Environ., 119 (2007), pp. 112-126