A methodology for quantifying global consumptive water use of coffee for sustainable production under conditions of climate change

Coffee is the second most traded commodity in the world after oil. A sustainable coffee industry is crucial to maintaining global agriculture, trade, human and environmental well-being, and livelihoods. With increasing water scarcity and a changing climate, understanding and quantifying the risks associated with water, a primary input in coffee production, is vital. This methodological paper examines the means of quantifying: (a) ‘current’ consumptive water use (CWU) of green coffee (coffee beans at harvest time) globally; (b) coffee ‘hot spots’ and ‘bright spots’ with respect to levels of CWU, yields and water stress; and (c) possible impacts of climate change on the CWU of coffee. The methodology employs satellite-derived monthly evapotranspiration data and climate projections from two global circulation models for three future scenarios. Initial estimates suggest that currently (on average) 18.9 m/kg of water is consumed in producing one unit of green coffee. The same estimate for irrigated coffee is 8.6 m/kg, while that for rain fed coffee is 19.6 m/kg. Climate scenarios show that effective mean annual rainfall in many major coffee areas may decrease by the 2050s. The generic methodology presented here may be applied to other crops, too, if crop data are


INTRODUCTION
Coffee is an extremely popular beverage worldwide with over 1,400 million cups being consumed every day. It is also one of the most traded commodities after oil (Nestlé ). However, it is noted that studies focussed on the sustainability of the coffee industry with respect to consumptive water use (CWU), a primary input in coffee production, are few and far between.
The sustainability of the coffee crop itself depends on the availability of adequate water, while coffee production needs to ensure the sustainability of the water resources that it depends on by not overusing and over-polluting it.
In the face of increasing water scarcity (CA ) and a changing climate, understanding and quantifying the risks to water, a primary input in coffee production, and coffee itself, is both timely and essential. This paper attempts to develop a methodology to: estimate the 'current' CWU of green coffee (coffee beans at harvest time) at the global scale; assess implications of the 'current' CWU on waterstressed locations of the world; and project the possible impacts of climate change (CC) on the CWU of coffee, with a view to understanding the influence of global water resources on the long-term sustainability of the coffee industry and vice versa.
The CWU, also referred to as the 'water footprint' or the 'virtual water content' (Chapagain & Hoekstra ; Siebert & Döll ), is represented by the total evapotranspiration (ET) requirement during a crop cycle, which is met from two sources, rainfall (green water) and irrigation (blue water). This study does not account for 'grey water' (the volume of water depleted in terms of quality deterioration during the crop production process and another component of the water footprint) because, generally, the 'grey water' component is much smaller than the ET component. In order to assess CWU, a novel approach (also discussed by Romaguera et  A handful of attempts have been made to visualize the impacts of rising temperature and changing patterns of precipitation (caused by CC) on coffee production quantities, areas suitable for coffee growth (loss of current areas and migration into new areas), and the coffee economy (e.g. Therefore, there is a need for a comprehensive global assessment of the current and future CWU of coffee, to facilitate the identification of sustainable production practices, especially with respect to water use, as well as the risks to the sustainability of the coffee industry due to water stress and CC. The scope of this paper is limited to a methodological development of such a global study and a first reconnaissance application of it, due to uncertainties linked to the principal source of coffee information: the HarvestChoice database (http://harvestchoice.org/), the latest, best available, published, public global coffee dataset.
The coffee grid data extracted from this database contain questionable information for several countries. For example, in Africa, substantial coffee-growing areas are shown for: central Madagascar, northern Cameroon, northern Ivory Coast, eastern Guinea and much of Zimbabwe, while coffee production is limited to the eastern highlands; and most of Tanzania, while production is concentrated in three specific regions (Figure 1(a)). Other questionable information includes average yields (e.g. Malawian coffee is not more productive than Brazilian) and the presence or absence of irrigation (e.g. Vietnam is shown as non-irrigated but uses substantial irrigation (D'haeze et al. , ; Figure 1(a)). Thus, this paper attempts to develop a methodology (using the latest, public, spatial data) to: • quantify components (total, fractions from rainfall and irrigation) of current (represented by year 2000) CWU of coffee at grid, national, continent and global levels to ascertain differences (if any) in the CWUs of irrigated/ rain fed coffee cultivation and Arabica/Robusta coffee varieties; • identify countries operating at 'optimum' CWU levels, and assess different pathways and sustainable practices that other countries can follow to reach these optimum levels; • identify current 'hot spots' and 'bright spots' considering CWU and yields, and assess implications of water scarcity and stress on these locations; • assess the impacts of changing rainfall and temperature (caused by CC) on coffee yield and ET, and hence on CWU by 2050; and • estimate potential changes in mean annual effective rainfall by the 2050s on current coffee areas and the implications of such changes on current 'hot spots' and 'bright spots'.

DEVELOPMENT OF METHODOLOGY AND DATA USED Estimating (current) CWU
Estimates of CWU are presented here as: (a) the quantity of water consumed in producing one unit of crop (CWU1 in cubic meters (m 3 )/kilogram (kg); Equation (1)); and (b) quantity of water consumed per year per unit of land (CWU2 in m 3 /(hectare (ha).year); Equation (2)).
Considering that optimum production of a coffee tree occurs from years 5 to 15, the CWU of coffee in terms of total water consumed per unit of crop product (CWU1: where ET ¼ actual crop ET per year (mm/year); Y ¼ coffee yield (kg/(ha.year)); 10 is a factor balancing units; and 1.5 is a factor reflecting the productive fraction of a coffee tree's life span (10 out of 15 years approximately).
The CWU of coffee in terms of total water consumed per unit of land per year by coffee trees (CWU2: m 3 /(ha.year)) is expressed as The factor 1.5 is not applied in estimating CWU2 because it is only estimated for the cropped area in a certain year.
The ET in the above equations is sourced either from soil moisture (rainfall, green water) or from irrigation (blue water). In the case of irrigated coffee, the contribution from irrigation (IR) to ET in any given month is assessed as FAO ). Arabica accounts for over 60% of the world's coffee production and is grown throughout Latin America,   Table 2.
Analysis of (current) 'hot spots' and 'bright spots' This paper attempts to gauge the 'productivity' of coffee with respect to water use at different locations of the world, and thereby identify current coffee hot spots (locations with low productivity) and bright spots (locations with high productivity  noted though that, in these areas, coffee yields may increase with small supplemental irrigation as opposed to other seasonal crops. The marginal value productivity of that additional irrigation could be much higher for coffee than for other irrigation-intensive crops.

Projecting CC impacts
Assessment of the impacts of changing rainfall and temperature (caused by CC) on ET, yield and the CWU of coffee is made for three scenarios specified in the Special Report  and CWU2 may be projected on these new areas using the same methodology set out in this paper.
Actual ET in 2050 is projected at 5 × 5 min grid cell resol-  Table 2. CWU1 and CWU2 in the 2050s are estimated using Equations (1) and (2)  The world average CWU1 is 18.9 m 3 /kg (which includes both irrigated and rain fed coffee), but the world average CWU1 of irrigated coffee alone is as low as 8.6 m 3 /kg while that of rain fed coffee is as high as 19.6 m 3 /kg (Table 3). The contribution from irrigation to the global average CWU1 is a mere 1% (Table 5), while the rest is provided by effective rainfall. This estimate is arrived at by separating and summing up the contribution from irrigation to CWU (IR in Equation (3)   (a) optimum water management practices (water is not limiting yield production by lack or excess and water inputs are timed appropriately); and (b) optimum farming practices  Raghuramulu ) as discussed earlier in this section.
A second group of countries, provisionally named   (1) and (2), respectively. on Vietnam (Figure 4). The potential for increasing yields through better water (quantity as well as timing) management is very high for this group of countries, through which they may move into the Farming Practices group and subsequently into the Optimum Group, hence the name 'Water Management'. In their quest for higher yields, these countries also have a unique opportunity to test out more sustainable agronomic practices, such as estab- ively. The first reason why this paper's estimate is higher than those by the other three studies is the accounting for non-productive years of the coffee tree (assuming that a coffee tree generally produces coffee beans only for 10 years out of its optimum productive life span of 15 years).
If only one productive year is considered as in the case of the other studies, CWU1 is 12.6 m 3 /kg, which is lower than the estimates from the other studies. Another reason is due to our use of actual ET estimates derived through remote sensing, rather than estimating potential ET by multiplying reference ET and the crop factor (Kc), the method used in the other three studies. Actual ET is generally less than potential ET. Last, but not the least, inherent deficiencies in the main coffee database used in this study The differences between CWU1 of Arabica and Robusta coffee varieties are not apparent at first glance. Table 3 shows the world average CWU1 for the two coffee varieties and a comparison across the three continents. The global average CWU1 of Arabica is 19 m 3 /kg while that of Robusta

Current hot spots and bright spots
The favourable conditions for coffee identified above are applied at grid cell level in this section in order to produce a spatially distributed global map of current coffee hot spots and bright spots (Figure 6(a)). All grid cells having yields less than 600 kg/ha, CWU1 higher than 25 m 3 /kg, and CWU2 below 6,000 as well as above 11,300 m 3 / (ha.year) are regarded as having highly unfavourable con- Higher levels of productivity are more widespread in Asia than in the other two continents.
The identified hot spots and bright spots are further compared with the spatial distribution of net annual irrigation requirement and irrigation water availability (considering that the IWMI water stress maps serve as proxies for irrigation water availability). By doing so, an attempt is made to identify the role that irrigation/nonirrigation plays in yield levels at these locations.  In as Vietnam (Robusta), but reap vastly lower yields (283 and 287 kg/ha against 1,690 kg/ha). This may be due to improper timing of water delivery to the coffee tree. It is generally established that a period of water stress, induced either by dry soil or dry air, is needed to prepare flower buds for blossoming that is then stimulated by rain or irrigation. Water must also be freely available during the period of rapid fruit expansion to ensure large, high-quality seed yields (Carr ). It is presumed that the practice of timing water delivery at such critical periods is established at an advanced state in the bright spots on Figure 6(a) as well as in the optimum group of countries (see section 'Current CWU, the role of irrigation and differences between Arabica/Robusta coffee varieties' and Figure 5), resulting in higher outputs through lower water inputs.
The need for irrigation and its role in controlling the timing of flowering also depends on other factors such as the monthly rainfall distribution, the severity of the dry season, and the soil type and depth (Carr ). Carr () elaborates that these factors vary within the two geographic regions where coffee is present: areas close to the equator with a bi-modal rainfall pattern and those at higher latitudes with a single rainy season and an extended dry season.
Further in-depth ground-based studies, at least within the two broad geographic regions, are required to quantify (as far as possible) the degree to which each of these factors influences coffee yields.

CWU under CC and implications for present hot spots
The median global average CWU1 of coffee in 2050, estimated using future ET and yield projections for the three SRES scenarios (A1B, A2 and B2) of the two GCMs forest. Since coffee is generally grown at forested sites and is a tree (although short), ET from coffee trees is assumed to be similar to that from natural forest in this methodology.
However, it is acknowledged that irrigated coffee will generally have higher ET than natural forest, especially in the dry areas (which are less humid), and that irrigated ET may be underestimated in this procedure.
Actual ET in 2050 was projected using a simplified how CC may impact water requirements during critical periods of crop growth, and scheduling of irrigation, will also help to enrich results derived from a global study. It is hoped that the procedures illustrated in this paper will lead the way to enhanced understanding of current water use and future risks in global coffee production to ensure the crop's continued sustainability. The same generic methodology may be applied to other crops too if sufficiently detailed spatial information on crop area, production and yield are available.