Integrated crop water management might sustainably halve the global food gap

2.1.1. Ex situ water harvesting (WH_ex)

2.1.2. In situ water harvesting (WH_in)

Micro-catchment systems, e.g. pitting, terracing, micro-basins, but also conservation tillage, and mulching can hinder water from running off the field and thus help increasing infiltration capacity. Particularly the combination of micro-catchments and mulching is observed to reduce runoff and soil evaporation considerably (Botha et al 2007, Biazin et al 2012). In LPJmL, infiltration rate In depends on soil properties, current soil moisture and the infiltration parameter p. By default, without management interventions, p = 2, but here we also simulate increased infiltration rates assuming four different intensity levels ($p=3,4,5,6$).

In is calculated for the upper soil layer as:

$$\begin{eqnarray}&&{In}={\rm{prir}}\times \sqrt[p]{1-\displaystyle \frac{{w}_{a}}{{W}_{{\rm{sat}}}-{W}_{{\rm{pwp}}}}},\end{eqnarray} \tag{ 1 }$$

where prir is daily rain and applied irrigation water, w_a is the available soil water content, and W_sat and W_pwp are soil water content at saturation and wilting point, respectively (all in mm). A sensitivity analysis for p is displayed in figure S2. Hereinafter, WH refers to the combination of WH_in and WH_ex measures at the four respective ambition levels.

2.1.3. Soil moisture conservation (SMC)

2.1.4. Irrigation improvements and expansion with saved water (IRR_exp)

Irrigation is represented through mechanistic simulation of surface, sprinkler and drip systems, depending on country and crop type. System efficiencies are directly linked to vegetation dynamics, weather and soil conditions, and water availability (Jägermeyr et al 2015). To simulate irrigation improvements, we define three theoretical transition scenarios:

• (1)  
  '50% surface', half of non-paddy surface irrigation is replaced by more efficient sprinkler systems.
• (2)  
  'Best practice', drip systems are established where applicable (based on crop suitability, (see Jägermeyr et al 2015), the remainder is under sprinkler irrigation, but paddy rice remains with surface systems.
• (3)  
  'All drip', drip irrigation on all irrigated land.

Improving irrigation performances can release water which in turn can be exploited for expanding the target area. To calculate expansion potentials, we only consider saved water that otherwise would have been consumed non-beneficially, as irrigation return-flows are often crucial for downstream water availability. Expansion of irrigated land is assumed to be further constrained by current rainfed cropland within a river basin. Table 3 presents global numbers of expansion.

Table 2.  Overview of model simulations. RCP = representative concentration pathways, GCMs = global climate models.

Reanalysis climate 1901–2009		Simulations
ACT	Current management calibrated with FAO data, reference run for all other simulations.	1
POT	Potentially achievable yields under unconstrained water availability (nutrient deficiencies remain)	1
SMC	Soil evaporation reduced by 10, 25, 50, and 85%.	4
WHex	Surface runoff collected by 10, 25, 50, and 85%	4
WHin	Infiltration rate increased in four sequential steps (equation 1)	4
WH	ex situ and in situ WH combined	4
IRR	Irrigation improvements: '50% surface', 'Best practice', 'All drip'	3
IRRexp	Irrigation expansion using saved consumptive water from IRR	3
IRRexp + SMCexp	Irrigation expansion using saved consumptive water from IRR and SMC implementation	3
Combined	Cross-combinations of SMC, WH and IRR	12
Combinedexp	Cross-combinations of SMC, WH and IRRexp + SMCexp (include the 'low', 'ambitious', 'max' scenario)	12
Climate change 2009–2099: 4 RCP scenarios, 20 GCMs each, constant and transient ${{\rm{CO}}}_{2}$ each
CC	Climate change impact	160
CC + manage	Climate change plus water management scenarios: 'low', 'ambitious', 'max'	480
Σ		688

Table 3.  Global area (Mha) of rainfed and irrigated agricultural land (including pastures) aligned with irrigation water consumption (IWC, ${{\rm{km}}}^{3}$) and irrigation withdrawal (IWD), for current land use and the three scenarios of combined water management 'low', 'ambitious', 'max', including irrigation expansion (Combined_exp in table 2).

	Rainfed	Irrigated	Total	IWC	IWD
	[Mha]			[${{\rm{km}}}^{3}$]
Actual	3984	297	4282	1268	2507
Low
'50% surface' irrigation	3895	387	4282	1350	2379
+ 25% SMC
+ 25% WH
Ambitious
'Best practice' irrigation	3639	642	4282	1515	2059
+ 50% SMC
+ 50% WH
Max
'All drip' irrigation	3388	894	4282	1607	1818
+ 85% SMC
+ 85% WH

2.1.5. Integrated water management

The model LPJmL globally represents biogeochemical land surface processes of vegetation and soils (Bondeau et al 2007, Fader et al 2010, Jägermeyr et al 2015), simulating daily water and carbon fluxes in direct coupling with the establishment, growth, and productivity of major natural and agricultural plant types at 0.5° resolution.

Agricultural land is represented by 12 specified crop functional types (CFTs), a class 'others' that includes a suite of crops collectively parameterized as annual crops, and pastures (Bondeau et al 2007). All CFTs are either irrigated or rainfed and its spatial distribution and their irrigated fraction is prescribed as in Jägermeyr et al (2015).

Assimilated carbon (in the process of photosynthesis) is allocated to harvestable storage organs (e.g. cereal grain) and three other pools (roots, leafs, stems). Sowing dates are dynamically calculated based on climate and crop type (Waha et al 2012). Crops are harvested when they reach maturity, defined either through a CFT-specific maximum value of daily accumulated phenological heat units or expiration of the growing season. Storage organs are subsequently removed from the field. Root growth and distribution within soil layers is CFT-specific, while the soil profile is discretized into 5 hydrologically active layers (Schaphoff et al 2013).

Plant growth is currently not directly nutrient-limited in LPJmL, yet constrained by temperature, radiation, water and atmospheric CO₂ concentration. We calibrate crop yields with national FAO statistics based on three model parameters (as in Fader et al 2010) to account for CFT-specific management intensities.

LPJmL partitions precipitation and applied irrigation water into interception, transpiration, soil evaporation, soil moisture, and runoff. Surplus water that cannot infiltrate generates surface runoff. Subsurface soil water above saturation runs off in lateral direction, while remaining soil water above field capacity percolates to the layer beneath, depending on its soil water content and hydraulic conductivity. Surface and subsurface runoff are accumulated along the river network and subsequently available for downstream reuse.

A recently implemented mechanistic irrigation module provides the framework for irrigation transitions (Jägermeyr et al 2015). In addition, we account for household, industry and livestock water use and include a representation of dams and reservoirs to improve the simulation of available surface water (Biemans et al 2011).

For the time period 1901–2009, we ran LPJmL forced with the Climate Research Units (CRU) TS 3.1 monthly climatology for temperature, cloudiness (Harris et al 2014) and with the Global Precipitation Climatology Centres (GPCC) precipitation data (Schneider et al 2014). The number of monthly rain days was derived from CRU and GPCC data as described in Heinke et al (2013). To cover uncertainties in climate change simulations (2009–2099), we considered four representative concentration pathways (RCPs: 2.6, 4.5, 6.0, 8.5), each being represented by 20 global climate models (GCMs) obtained from the CMIP5 multi-model ensemble dataset (table S2) (Taylor et al 2012). Monthly GCM output was bilinearly interpolated and bias-corrected to the reference period 1970–2000 using a method adapted from Watanabe et al (2012). To analyze the CO₂ effect on crop growth, each simulation was performed with constant (at year 2000) and transient CO₂ concentration. Model runs follow a 1000-year spinup (recycling the first 30 years of input climatology) and sowing dates are fixed during the simulation period after 1960 to allow the comparison of water management potentials between different runs and otherwise they would represent a form of adaptation not intended here.

Spatially explicit global information on cropland extent is obtained from the MIRCA2000 land-use dataset (Portmann et al 2010). The extent of areas equipped for irrigation from 1900–2005 is imported from Siebert et al (2015) and the distribution of irrigation systems from Jägermeyr et al (2015). Land use patterns are fixed after the year 2005. Irrigation withdrawal is constrained by local, renewable water storage, i.e. there is no implicit assumption about contributions from fossil groundwater or diverted rivers. Only potentially achievable yields (figures 2 and 4) are simulated under unrestricted water availability.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Simulated crop water management increases global kcal production by 41% under the 'ambitious' scenario (all measures combined, including irrigation expansion), while using existing agricultural land, yet cutting irrigation abstractions. Production increases by more than 55% in many river basins between the Middle East, central Asia, China, Australia, southern Africa and North and South America (figure 3(a)). Under the 'low' and 'max' scenario global kcal production increases by 18 and 60%, respectively (figure 3(b)). Individual effects of irrigation transitions (IRR), soil moisture conservation (SMC) and water harvesting (WH) are specified below (section 3.2. and the following).

On current farmland, we calculate the average transpiration coefficient (TC) at 46% for rainfed and 33% for irrigated systems (figure 1). Simulated water management significantly shifts water towards transpiration through alleviating soil evaporation, surface runoff and irrigation losses. Therefore, the TC (combined for rainfed and irrigated systems) increases from 42% to 49%, 54%, and 61%, respectively, in the 'low', 'ambitious', and 'max' scenario.

Production gains are particularly steep in regions currently experiencing large water gaps. Figure 2(a) highlights basins where the water gap exceeds 50%, and include large parts of the Middle East, central Asia, North China plains, Australia, southern Africa, and the western United States. Based on LPJmL, current global rainfed farming operates with a water gap of 29% relative to its unconstrained water potential (yet neglecting nutrient deficiencies). In the 'low', 'ambitious' and 'max' scenario, this gap could be closed up to 17%, 11% and 5%, respectively (figure 4).

The irrigation water gap is necessarily smaller than the rainfed, as irrigation largely closes the gap. Under current conditions, global irrigated farming is simulated to be only 6% short of its unconstrained water potential (spatial patterns in figure 2(b)). While better water management can further narrow this gap (local significance), important benefits at the global level are associated with water savings (figure 4).

Figure 5(a) confirms that improved irrigation and SMC implementations can only marginally increase irrigated production at the global level (by <2%). More importantly however, these measures ('low', 'ambitious', 'max' scenario without expansion) show the potential to cut consumptive losses (i.e. soil evaporation, interception, and evaporative conveyance losses) by respectively 24%, 48%, and 85% (figure 5(b)). This results in significant reductions of global irrigation withdrawal from currently 2507 ${{\rm{km}}}^{3}$ to 2071, 1248 and 808 ${{\rm{km}}}^{3}$ (figure 4), because alleviated soil evaporation and higher conveyance and application efficiencies strongly reduce irrigation requirements.

Zoom In Zoom Out Reset image size

These water savings would theoretically allow for an additional 90, 345, and 597 Mha expansion into rainfed cropland, respectively, for the 'low','ambitious', and 'max' scenario. These numbers are substantial in perspective of current irrigated land of about 300 Mha and the expected slow expansion pace. But future irrigation expectations are curbed due to land constraints under current system efficiencies (Alexandratos and Bruinsma 2012); farmers who pursue efforts to save water often use it to expand their irrigated share of cropland (Fishman et al 2015).

Global total kcal production (rainfed + irrigated) could thereby increase by 7%, 26%, and 43% with considerably higher numbers in specific basins particularly between the middle East, large parts of Asia, and Central to North America (figure 5(b); aggregated to the basin level, as upstream irrigation improvements can have water trade offs downstream). Note that irrigation expansion (with higher efficiencies), but also SMC and WH, lead to higher productive plant transpiration, which increases global irrigation water consumption from currently 1268 ${{\rm{km}}}^{3}$ to 1350, 1515, and 1607 ${{\rm{km}}}^{3}$, respectively (table 3, 'low', 'ambitious', 'max' scenario), while non-productive losses still decrease (not all saved water used up for expansion as some basins lack sufficient available rainfed cropland, figure 5(a)). Overall, the total global withdrawal amount is simulated to decrease by 128, 448, and 689 ${{\rm{km}}}^{3}$ for the three respective scenarios, despite the growth of irrigated areas (table 3).

SMC shows considerable potential to amplify rainfed kcal production (3%–14% globally, figure 6(a)). Regions with high sensitivity are concentrated in semi-arid to arid regions such as the Sahel, southern Africa, central Asia, and Australia, where production increases reach >20% (figure 6(b), no downstream effect and thus displayed at the grid cell level). As for water harvesting, WH_ex exhibits much higher production potentials compared to WH_in, but combining both measures appears especially beneficial at low intensity levels, and could increase global kcal production by 7%–24% (figure 6(c)). Figure 6(d) shows spatial patterns of WH (50% level) with high sensitivity also in semi-arid regions, but in addition in sub-humid regions with high rainfall variability and runoff excess, across tropical and temperate regions.

Zoom In Zoom Out Reset image size

Climate change is simulated to have adverse effects on global crop production, while high uncertainty is associated with the direct effect of carbon dioxide on plant growth. In simulations with constant CO₂ concentration (performed to isolate the climate change effect), global kcal production is projected to change by −3% (RCP 2.6) to −18.2% (RCP 8.5, table 4, median of 20 GCMs), mostly due to increased water deficiencies. With transient CO₂ concentration, a strong fertilization effect (in LPJmL not directly constrained by nutrient limitation) actually increases global production by 4.3% (RCP 2.6) to 13% (RCP 8.5) despite concurrent climate impacts. A 'moderate' CO₂ fertilization (mean of constant and transient CO₂ simulations) suggests marginal global production changes (−2.6% to 1.6%). Regionally however, India, Pakistan, west Australia, African Sahel, and east Brazil face negative changes from −5% to < −20% under RCP 2.6 ('moderate' CO₂), while strong increases occur in large parts of Russia, east and southern Africa, and parts of central and south America (5% to >20%). In an RCP 8.5 world ('moderate' CO₂) the Mediterranean region, major parts of the United States and Mexico, and southern Asia appear additionally on the map with distinct negative changes, and many basins show kcal declines from −10 to < −30% (figure 7(c)).

Zoom In Zoom Out Reset image size

In the 'low' scenario, most adverse climate change impacts are simulated to be buffered in a RCP 2.6 world (figure 7, see figure S4 and S5 for constant and transient CO₂). The 'ambitious' scenario can ease negative impacts in an RCP 8.5 world in many basins, but some regions, notably east Brazil and west Africa, remain with negative impacts. Despite large uncertainties associated with the CO₂ effect, global crop production is simulated to increase by >40% under 'ambitious' water management for all but the most severe climate change scenario (35% with RCP8.5, table 4).

Supplemental irrigation, mulching, and conservation tillage demonstrably increased yields in case studies by 56%, 44%, and 30%, respectively (Araya and Stroosnijder 2010, Welderufael et al 2008, Fox and Rockström 2003). A major case study (286 projects in 57 countries) documents average yield increases by 79% through a number of conservation agriculture interventions, including water harvesting and conservation tillage (Pretty et al 2006). A wider collection of case studies shows similar ranges (table S1). Lebel et al (2015) simulate for maize in sub-Saharan Africa (SSA) an average yield increase with WH intervention of 9%-39%. We arrive at 0%–33% across SSA basins. In our simulation (50% level), on average 57 mm supplemental irrigation are applied during the growing season (in 95% of all cases less than 100 mm, figure S3), Fox and Rockström (2003) document 60-90 mm in a Burkina Faso case study. More generally we can reproduce the documented scale of observed yield gains using LPJmL, and our analysis extends case study insights to a broader set of climates, locations, and crops and thus refines management potentials at the global level (Rost et al 2009).

We point out that it is critical to evaluate the local feasibility of WH catchment and storage systems. This depends on different factors (e.g. terrain type, soil structure, hydroclimatic setting, social and financial capital (Studer and Linger 2013, Falkenmark et al 2001, Mati et al 2007)), addressing those in detail is beyond the scope of this broad-scale study of biophysical potentials. Although case studies show that often only 10%-20% of the land is unsuitable for WH and SMC adoption (Barron et al 2015, Kahinda et al 2008, van Rensburg et al 2012), we might exaggerate WH suitability. However, this is faced by our very conservative assumption on the catchment area that is limited to existing cropland only, and which is often much larger in reality. But it appears infeasible to delineate additional suitable catchment areas with a sufficient degree of detail globally.

Therefore, we regard the 'low' and 'ambitious' scenario as potentially achievable, while the 'max' scenario, locally proven though, indeed appears unlikely to become implemented globally. Nevertheless, it provides important insights into planetary biophysical capacities. Furthermore, it is difficult to quantify the extent to which farmers already adopted WH and SMC measures. Although initial up-scaling projects prove successful regionally (e.g. Zhu and Yuanhong 2006), they still represent only marginal areas at the global level.

Our estimates of irrigation withdrawal and consumption agree well with previous estimates (FAO 2014, Döll et al 2014, Wada and Bierkens 2014), albeit featuring much more process detail. Irrigation expansion adds a noticeable share to production increases simulated in this study (figures 3(b) and 5(a)). The expansion of irrigated crops replaces rainfed crops, which results in a propagation of irrigated cropland into pastures in some cells, as the share of irrigated pastures is generally low. However, in SSA only 5% of the cropland is under irrigation today, which first explains our flat irrigation improvement potentials in SSA (figure 5(b)), and second outlines scope for irrigation expansion using currently untapped water resources (FAO 2005, Burney et al 2013, Xie et al 2014).

Finally, it is important to note that upstream IRR and WH interventions can lead to reduced return-flows and runoff, which can negatively affect water availability downstream. Despite noticeable impacts locally, gains at the basin level over-compensate losses in all basins (figure S4). This appears beneficial from a food production perspective, but there is a clear need for policies and institutional orders to regulate water reallocations. In this context it is crucial to quantify contributions from groundwater and water diversions, given the complex recharge and transboundary issues involved.