Global evaluation of the effects of agriculture and water management adaptations on the water-stressed population

Abstract

Fresh water is one of the most important resources required for human existence, and ensuring its stable supply is a critical issue for sustainable development. The effects of a general set of agriculture and water management adaptations on the size of the world’s water-stressed population were assessed for a specific but consistent scenario on socio-economic development and climate change during the 21st century. To maintain consistency with agricultural land use change, we developed a grid-based water supply–demand model integrated with an agro-land use model and evaluated the water-stressed population using a water withdrawals-to-availability ratio for river basins. Our evaluation shows that, if no adaptation options are implemented, the world’s water-stressed population will increase from 1.8 billion in 2000 to about 3.3 billion in 2050, and then remain fairly constant. The population and economic growth rather than climate change will be dominant factors of this increase. Significant increase in the water-stressed population will occur in regions such as North Africa and the Middle East, India, Other South Asia, China and Southeast Asia. The key adaptation options differ by region, depending on dominant crops, increase in crop demand and so on. For instance, ‘improvement of irrigation efficiency’ and ‘enhancement of reclamation water’ seem to be one of important options to reduce the water stress in Southeast Asia, and North Africa and the Middle East, respectively. The worldwide implementation of adaptation options could decrease the water-stressed population by about 5 % and 7–17 %, relative to the scenario without adaptations, in 2050 and 2100, respectively.

Appendices

Appendix 1. Outline of the agro-land use model

The agro-land use model developed by Kii et al. (2011) is a grid-based model with 15 × 15 min resolution, in which the amounts of cropland required to meet the crop demands are estimated for 32 separate regions, for every decade from 2000 to 2050, and at time points for 2070 and 2100. Crop types include wheat, rice, maize, sugar cane, soybeans, oil palm fruit, rapeseed and others, based on their importance in terms of principal food, favorite food, vegetable oil and feed considerations. The amounts of available land for crop production are allocated in order to meet the regional crop demands. If there is a shortage of crop production due to a lack of available land, production is reallocated to regions where land is still available and any shortage is filled by means of crop trade. The shares of domestic productions relative to the regional demand for each of crops, and the export shares in a global market are maintained to those in 2000 (FAO 2010) except for the case, in which the available land in some regions is not sufficient to meet the crop demand and then the reallocation of production is examined.

Although the model is simplified so that only one selected crop is produced per grid in each analysis year, the regional mean yield for each crop in 2000 is adjusted to reduce the discrepancy with the FAO (2010) statistics less than a few % for most of the 32 regions, and then this simplification is not likely to cause significant estimation errors. It is well justified to utilize in order to quantitatively estimate impacts on the yield and required area due to changes in crop demand, productivity progress, changes on regulation for land use, etc. Furthermore, impacts due to climate change and adaptations for planting can be taken into account through changes in the potential production, which can be calculated using the information on crop characteristics and soil types provided in the GAEZ model for both rain-fed and irrigated conditions. The grids available for irrigation are based on an irrigation map for the year 2000 (Siebert et al. 2007). The regional areas for irrigation maintain the consistency with those of the original map, although the share of irrigation area for each of grids is simplified to be either 0 or 100%. No expansion of the irrigation available grids is allowed in the future according to the assumptions by Alcamo et al. (2007). Our assumption is thought as one of very plausible scenarios, in cases where crop productivities in rain-fed cultivation areas will be robustly improved and the regulations regarding the quality of wastewater will be tightened. The possibility of double cropping is considered for some developing regions, of which the double-cropping area for rice productions are reported (FAO 2009). The double-cropping grids are allocated on the grids with high production potential for rice so that the regional areas may be equal to the statistics. In the world total, approximately 300,000 km² in 2000 is considered; and for simplicity, it is assumed that the allocations of double cropping grids will be maintained in the future.

In this study, a partially revised model based on the original agro-land use model is used. Thus, the grid data for arable land in 2000 is updated in accordance with Fischer et al. (2008) and available grids for arable land after 2010 are assumed to can be slightly expanded by including land adjacent to the arable land grids in the previous analysis year (an increase of up to 3 % for each of the 32 regions). Cereals, which are especially important as a principal food crop, are treated preferentially by allocating them to the grids with higher productivity. Furthermore, the assumptions regarding the ‘yield factor’ are revised.

The ‘yield factor’ defined as K in Eq. A1 has been introduced by Kii et al. (2011) in order to reduce discrepancies between the regional mean of crop yield, estimated from potential production by the model, and the FAO’s statistics (FAO 2009), and to express future improvements in yield cased by factors except for changes on climate, planting area and times, etc..

$$ {{Y}} = \left( {\frac{{\sum\limits_i {{{P}_i} \bullet {{A}_i}} }}{{\sum\limits_i {{{A}_i}} }}} \right) \bullet K $$

(A1)

where Y is a regional mean yield for each crop, and i, P and A mean a cultivation grid, potential production (estimated using the GAEZ base module taking climate, soil, slope, varieties of crops and planting times into consideration), and its land area, respectively.

In this study, the scenario of the yield factor (K) is developed taking into account the technological progress associated with economic growths (e.g., mechanization, use of chemical fertilizer) and changes caused by other factors such as land use constraints, agricultural policies, adjustments to price etc. (RITE 2010). The future scenarios on the technological progress and other factors are assumed based on their historical trend during the period of 1961–2006 (FAO 2010). The developed scenario of the yield factor is shown in Table 5. For all crops, a greater growth is expected in developing regions than in developed regions.

Table 5 Scenario describing the yield factors. (Each value is normalized by comparison with 2000)

Appendix 2. Domestic water withdrawal per ‘access-person’ in urban and rural areas

The relationships between per ‘access-persons’ withdrawal and per capita GDP are formulated by Eq. A2, which was estimated by regression analysis using several country-statistics for the period of 1990–2004. That is the statistics for domestic water withdrawal (FAO 2009), per capita GDP (World Bank 2008) and ‘access-persons’ in urban and rural areas (World Bank 2008)). In order to separate the domestic water withdrawal reported by FAO into those for urban and rural areas, the per capita domestic water demand for each of areas in several regions of the world (refer to Table 4.4 of Rosegrant et al. 2002) were adopted:

$$ \left. \begin{gathered} D{{W}_{{t,c}}} = {{k}_c} \times DW{{0}_{{t,c}}} \hfill \\ \log DW{{0}_{{t,c}}} = \left( {a \times per\,capita\,GD{{P}_{{t,c}}}} \right)/\left( {b + per\,capita\,GD{{P}_{{t,c}}}} \right) \hfill \\ \end{gathered} \right\} $$

(A2)

where the subscript t and c indicate the time point and country, respectively, DW is per ‘access-persons’ withdrawal, a and b are regression coefficients (a = 1.98 and b = 83 for urban areas, and a = 1.88 and b = 127 for rural areas), and k is a factor to reduce discrepancies between the per ‘access-persons’ withdrawal estimated by the regression function and the amount by statistics for the year of 2000. Based on the relationships studied, per ‘access-persons’ withdrawal is expected to increase in both of urban and rural areas associated with the economic growth.

Appendix 3. Scenarios for the industrial water requirement and water-use efficiency

The ‘weighted production’ and the energy-use efficiency for the production of crude steel by blast furnaces are used as proxies for the industrial water requirements and the water-use efficiency, respectively, as mentioned in Section 2.1.3. Table 6 shows values used in this study for these proxies. The data up to the year 2050, have been obtained from the DNE21+ model and, from 2050 on, estimates are extrapolated based on the assumption that the annual change rate will gradually decrease.

Table 6 Scenario describing the ‘weighted production’ and the energy-use efficiency for the production of crude steel by blast furnaces. The values in 2000 and the annual change rates are listed for the top twenty ‘weighted production’ regions in 2050 among the 54 regions defined for the DNE21+ model, which are expected to occupy the approximately 90 % of the world ‘weighted production’ during the 21st century

Appendix 4. The ratio of urban population relative to total population

According to the UN survey (2009), the specific conditions and terms used to describe the ‘urban’ population can differ from country to country. For instance, although the ratio of non-agricultural workers is considered in some countries such as China, India and Russia, such occupations are not specified in most countries and other descriptive terms are used, such as ‘the population density is above a certain level’ and ‘several public infrastructure constructions have been carried out’. We assumed that the urban population ratio (UP) was represented by a function of the population density (D) and per capita GDP (used as a proxy for the construction of infrastructure), as denoted by Eq. A3, and the parameters α ₁ and α ₂ are estimated for each country using regression analyses based on statistical data from 1960–2005 (i.e., data for the urban population ratio (UN 2009), area per country, and per capita GDP (World Bank 2008)).

$$ U{{P}_{{rt}}} = {{\alpha }_{{1r}}} \bullet {{\left( {\log \left( {per\,capita\,GDP{}_{{rt}} \bullet {{D}_{{rt}}}} \right)} \right)}^{{{{\alpha }_{{2r}}}}}} $$

(A3)

where r and t are the country and the time point, respectively. This equation shows that the urban population ratio increases with growth in per area GDP. Determination coefficients for the past of about 50 years are over 0.8 for 97 out of the 135 countries which comprised 91 % of the world’s population in 2000.

Appendix 5. Grid-based distributions of population and urban areas

Scenarios for population distribution and urban areas (for every 10 years from 2000 to 2100) were developed based on population maps for the year 2000 (PBL 2009) with 5 × 5 min of resolution. First, the original PBL’s total population map was adjusted to agree with the ALPS country-level population data, by applying the same factor to the grids within one country. Then, population distributions after 2010 were estimated so that the population in each grid will change in proportion to its population density as denoted by the following Eq. (A4).

$$ {{p}_{{i,t + 1}}} = {{p}_{{i,t}}} + ({{p}_{{i,t}}}/{{P}_t}) \times ({{P}_{{t + 1}}} - {{P}_t}) $$

(A4)

where the P is a number of population for a specific country, and subscript t is the time point. The p _i,t indicates the number of population of grid i, which is included in the country. This allocation rule is based on a study by Grübler et al. (2007), which reflects a well-observed pattern of the concentration to urban areas and the spread to areas located in close to proximity.

Grids for urban areas in 2000 were specified based on the PBL’s urban population map so that the number of population included in the urban grids agrees with the ALPS country-level urban population data, by slightly shifting the PBL’s original boundaries between urban-rural areas. For the estimations of urban grids after 2010, it was assumed that a grid classed as urban will be the most likely to become urban in the next future time point. The urban-rural boundaries were adjusted for each of the time points so that population included in the urban grids agrees with the ALPS country-level urban population data.

Appendix 6. Scenarios for food demand

The scenario for per capita dietary energy demand was developed based on the logistic functions of per capita GDP, which were estimated by regression analysis for each of the 32 regions assuming that the per capita food demand will increase associated with the economic growth, and will be saturated at a specific amount. For the analysis, the data during the period of 1961–2005 (FAO 2010; World Bank 2008) were utilized. Figure 11(a) shows the developed scenario for major regions. It should be noted that this demand includes household.

Figure 11

(a) Per capita food demand for major regions and (b) Shares of dietary energy demand by the four food classes in the world. In Figure (a), solid lines are based on the FAO’s statistic, and dashed line show scenarios developed in this study. In Figure (b), values after 2005 are based on the developed scenarios

It was assumed that the demand was divided into the demand for four classes of food (i.e., cereals and vegetables, animal products, sugar, and oil). The increases in the shares for animal products, sugar and oil associated with economic growth were projected based on the logistic functions of per capita GDP, which were estimated by regression analysis for the each region. Figure 11(b) shows the scenario for the shares by food classes in the world. It is assumed that preference of animal products, oil and sugar will be enhanced throughout this century.

Furthermore, demand of the animal products was converted to the feed crop demand, and the oil and sugar demand were converted to the ingredient crop demand. Finally, the demands of the four food classes were aggregated to the demand by each of the eight specific crops adopted in the agro-land use model (i. e. wheat, rice, maize, sugar cane, soybeans, oil palm fruit, rapeseed and others) (RITE 2010, 2011a).