Review and classification of indicators of green water availability and scarcity

Research on water scarcity has mainly focussed on blue water (groundand surface water), but green water (soil moisture returning to the atmosphere through evaporation) is also scarce, because its availability is limited and there are competing demands for green water. Crop production, grazing lands, forestry and terrestrial ecosystems are all sustained by green water. The implicit distribution or explicit allocation of limited green water resources over competitive demands determines which economic and environmental goods and services will be produced and may affect food security and nature conservation. We need to better understand green water scarcity to be able to measure, model, predict and handle it. This paper reviews and classifies around 80 indicators of green water availability and scarcity, and discusses the way forward to develop operational green water scarcity indicators that can broaden the scope of water scarcity assessments.


Introduction
Freshwater is a renewable resource that is naturally replenished over time when moving 15 through the hydrological cycle (Oki and Kanae, 2006;Hoekstra, 2013). Precipitation forms the input of freshwater on land. Subsequently, it takes the blue or the green pathway back to the ocean and atmosphere before eventually returning as precipitation again (Falkenmark, 2003;Rockström, 2006, 2010). The water that runs off to the ocean via rivers and groundwater is called the blue water flow. The 20 green water flow is formed by the water that is temporarily stored in the soil and on top of vegetation and returns to the atmosphere as evaporation instead of running off (Hoekstra et al., 2011). As suggested by (Savenije, 2004), we use in this paper the term evaporation (instead of the often used term evapotranspiration) to refer to the vapour flux from land to atmosphere, which includes soil evaporation, evaporation of A straightforward definition of water scarcity is: "an excess of water demand over available supply" (FAO, 2012). Various other definitions of water scarcity exist that aim to be more inclusive: "An imbalance between supply and demand of freshwater in a specified domain (country, region, catchment, river basin, etc.) as a result of a high rate of demand 10 compared with available supply, under prevailing institutional arrangements (including price) and infrastructural conditions." (FAO, 2015) "When an individual does not have access to safe and affordable water to satisfy her or his needs for drinking, washing or their livelihoods we call that person water insecure. When a large number of people in an area are water insecure for a significant period 15 of time, then we can call that area water scarce." (Rijsberman, 2006) Considering these definitions, we can conclude that water scarcity is not something that is experienced by a single person on a particular moment (day or week). Rather, it is experienced by a larger community within a certain geographic area (e.g. catchment or country) and relates to larger time-scales (months or years). Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | financial capital limit access to the water, the term economic water scarcity applies (Seckler et al., 1999;Molden, 2007). In a broader sense, Ohlsson (2000) defines social resource scarcity as the situation in which social resources required to successfully adapt to physical water scarcity fall short. 5 According to economic theory, water is a scarce good, because it carries opportunity costs, which are the benefits foregone from possible alternative uses of the water (FAO, 2004). This is a form of "relative scarcity" based on the assumption of substitutability of goods (Baumgärtner et al., 2006). Water can be scarce in the relative sense also in water-abundant areas, because allocating water to purpose A implies it cannot be 10 allocated to purpose B. In other words, water for purpose A is scarce in relation to water for other purposes. In common language we are inclined to say that at some times water is scarce and at other times it is not. In economic sense, water is always scarce; the degree of water scarcity can vary though, it can even be zero if alternative uses and thus competition is absent. 15 We speak of "absolute scarcity" when according to Baumgärtner et al. (2006) "scarcity concerns a non-substitutable means for satisfaction of an elementary need and cannot be levied by additional production". This means that in an area with a limited amount of water resources (that cannot be increased), at a certain level of consumption, water for elementary purposes (e.g. drinking and food production) will no 20 longer be substitutable with water use for less essential purposes. In this case, there is "absolute scarcity" of water. Whether water is scarce in the absolute or relative sense thus depends on the degree of water scarcity: relative water scarcity turns into absolute scarcity when the boundaries of water exploitation are approached. 5524

Water quantity and quality
Water scarcity is not only a function of the quantity of the water resource in relation to the demand, but also the quality of the resource in relation to the required quality for its end-purpose (Pereira et al., 2002). If there is sufficient water available for a certain purpose, but it is polluted to such an extent that it is not usable for that purpose, then 10 water can be considered scarce as long as the means are not available for cleaning the water to a desirable level. Pollution of water resources can thus aggravate water scarcity (FAO, 2012).

Scope of the review and classification
This paper focuses on green water, water quantity and physical water scarcity and 15 treats both green water availability and scarcity. In the next section, we consider indicators within this scope, including indicators of aridity, agricultural and meteorological drought, vegetation drought, soil moisture and integrated green-blue water scarcity. The focus of this paper implies that several concepts and indicators fall outside the scope of the classification. Concepts and indicators focusing on blue water that are out Introduction   Drought Index (Karl, 1986); several indicators reviewed by Smakhtin (2001).
-Blue water scarcity: measures human demand for blue water resources vs. blue water availability and is thus purely about blue water. Examples of associated indicators are: the water crowding indicator (Falkenmark et al., 1989), the withdrawal-5 to-discharge ratio (Vörösmarty et al., 2000), Water Poverty Index (Sullivan et al., 2003); Water Stress Indicator (Smakhtin et al., 2004); Water Stress Index (Pfister et al., 2009); Dynamic Water Stress Index (Wada et al., 2011); Blue Water Scarcity . Note that some of these indicators also incorporate more than only physical elements of water scarcity (e.g. Water Poverty Index).

10
Concepts related to broader forms of water scarcity than physical water scarcity that are out of scope are: -Socio-economic drought: concerns imbalances in supply and demand of economic goods due to the physical characteristics of drought (Wilhite and Glantz, 1985;American Meteorological Society, 2013)

Green water availability and scarcity indicators
We have identified around eighty indicators of green water availability and scarcity, which we classify into the following categories: 1. Green water availability indicators show whether green water availability is low or high and are insensitive to actual water demand. In other words, when the 5 water demand increases, indicator values will not change. Within this category we distinguish absolute and relative green water availability indicators: a. Absolute green water availability indicators measure actual conditions of green water availability (in an absolute sense).
b. Relative green water availability indicators measure actual conditions of 10 green water availability compared to conditions that are perceived as "normal", which is often defined as the climate-average or median value of the variable of interest.
Note that this distinction between absolute and relative indicators is unrelated to and different from the concepts of relative and absolute scarcity earlier discussed in 15 Sect. 2.2.
2. Green water scarcity indicators incorporate elements of both water availability and demand and therefore respond -in contrast to green water availability indicators -to changes in water demand as well. We distinguish three different options to measure green water scarcity conceptually (explanation in Sect. 3.2): 20 a. Green water crowding.
b. Green water requirements for self-sufficiency vs. green water availability.
c. Actual green water consumption vs. green water availability.
In this paper, the term "demand" occurs in two different contexts with different meaning and hence requires some clarification. When we speak of "demand" in relation to Introduction the concept of green water scarcity, we refer to the human demand for green water, associated with the production of biomass for human purposes. In the discussion of agricultural drought indicators in Sect. 3.1, the term "crop moisture/evaporation/water demand" is used to refer to the water needs of the crop for non-water limited growth.

5
Indicators of green water availability fall apart in indicators that do so in absolute sense or in the sense of relative to normal conditions. These two categories are treated in the next two subsections, respectively. Descriptions of various specific green water availability indicators that fall in the two categories are included in Appendices A and B, respectively. The indicator acronyms used in this section are defined in these appen-10 dices.

Absolute green water availability indicators
Indicators in this category measure green water availability in a certain area (or location) and period (or moment) in an absolute sense. We find here indicators of aridity, agricultural drought, soil moisture and agricultural suitability. Aridity indicators are solely 15 based on climatic variables, while the others incorporate variables related to the soil and vegetation (or crop) as well. Agricultural drought indicators measure green water availability set against crop water demand for non-water limited growth. Absolute soil moisture indicators provide a "direct" measure of the amount of soil moisture available. Lastly, land classifications based on agricultural suitability under rain-fed conditions 20 indirectly measure if green water availability is sufficient for the production of certain crops, by taking into account climate, soil and topographic conditions and the requirements of the crop.

Aridity indicators
Aridity is seen as a permanent feature of a climate, consisting of low average annual precipitation and low soil moisture availability (Pereira et al., 2002;Heim, 2002;Kallis, 2008). As such, one can say that an aridity map shows the preconditions for vegetation (Falkenmark and Rockström, 2004). Aridity indicators are usually based on long-term 5 average comparisons of precipitation vs. potential evaporation, temperature or saturation deficit, whereby the latter two were often used in the 20th century as proxies for potential evaporation due to lack of data. Aridity indicators are reviewed by Walton (1969), Wallén (1967) and more recently by Stadler (2005). Two indicators we classify as aridity indicators require a note. First, Peixoto and Oort 10 (1992) take the long-term average ratio of actual (instead of potential) evaporation over precipitation as a measure of aridity (ER). Second, the SCMD by  shows the probability of seasonal crop moisture deficiency based on a combination of long-term precipitation records and area-weighted evaporation of the mixture of crops grown in the study area.  apply the SCMD to assess agricul-Introduction Therefore, the bulk of agricultural drought indicators measure crop available water compared to crop water needs for non-water limited growth (i.e. potential evaporation) and are usually applied on a daily, weekly, monthly or seasonal basis (Woli et al., 2012). Some indicators only look at the difference between actual and potential transpiration (e.g. DTx and WDI).

5
Drought is typically a relative-to-normal phenomenon as will be discussed in Sect. 3.1.2. Agricultural drought indicators, which measure actual relative to potential evaporation, are "relative" indicators in another way, though. They do not compare actual with "normal" conditions. Instead, they compare moisture supply with a crop water demand in the ideal case of non-water limited growth. Therefore these indicators actually measure absolute green water availability (actual evaporation), set against this crop water demand. In fact, these indicators say more about the demand for blue water (irrigation) to ensure non-water limited crop growth than they do about green water availability.
Here, three indicators need an extra note. Both the DSI by Mu et al. (2013) and 15 the GrWSI by Wada (2013) compares the actual to potential evaporation ratio with the long-term average of this ratio. Therefore, these indicators are in essence relative indicators according to our classification. However, they are classified as agricultural drought indicators because they, like most of the others, measure actual to potential evaporation. The name of the GWSI by Nunez et al. (2013) suggests that it is a green 20 water scarcity indicator. Nevertheless, we classify it as an agricultural drought indicator, because it measures actual moisture supply vs. crop-specific reference evaporation, albeit on a larger time-scale (three-year crop rotation) than most other agricultural drought indicators. 25 Several indicators provide a measure of the absolute amount of soil moisture available at a given location and moment (or summed over a period), be it on the basis of field measurements (e.g. SMIX, SMI) and/or modelling of the soil water balance (e.g. Avg-GWS and SD-GWS) or remote sensing data (e.g. TVDI, MPDI). Many of these indicators have been introduced and applied as indicators of agricultural drought (e.g. ADD, SMDI, SMIX, SMI), analysing the correlation between soil moisture availability and crop yields. Therefore, they are typically calculated on intra-annual time-scales.

Absolute soil moisture indicators
Agricultural suitability under rain-fed conditions 5 Maps that classify land according to agricultural suitability under rain-fed conditions (green water only) are indirect measures of green water availability in the absolute sense. Up to date, two global studies have made such land suitability classifications for rain-fed crop production for climate-average temperature and precipitation conditions and taking into account various soil parameters and terrain slope: GAEZ 10 (IIASA/FAO, 2012) and GLUES (Zabel et al., 2014). Both studies classify lands as "not suitable", "marginally suitable", "moderately suitable" or "highly suitable". This classification shows where the climate, soil and topographic conditions are more or less suitable for agricultural production with green water only. In other words, where aridity maps show the preconditions for vegetation in general (Falkenmark and Rockström,15 2004), these maps show the preconditions for rain-fed crop production, therein considering soil and terrain parameters in addition to climate.

Relative green water availability indicators
Indicators in this category measure green water availability relative to a "normal" condition and are usually calculated on intra-annual scales. As opposed to aridity, drought 20 is often defined as a condition relative to what is perceived as a "normal" amount of precipitation or balance between precipitation and evaporation (World Meteorological Organization, 1975;Wilhite and Glantz, 1985). Droughts are often termed temporary, uncertain and difficult to predict features characterized by lower-than-average precipitation (Pereira et al., 2002;Heim, 2002;Kallis, 2008;Mishra and Singh, 2010 and Singh (2010) and those indicators applied in the US by Heim (2002) and Hayes (2007).

Meteorological drought indicators
Meteorological drought indicators are based on climate factors. They fall apart in indicators that are solely based on precipitation (e.g. SPI) and those that consider both 10 precipitation and potential evaporation (e.g. PDSI, RDI, SPEI). These indicators show whether there is relatively little precipitation or whether the normal balance between precipitation and evaporation is distorted.

Vegetation drought indicators
Vegetation drought indicators show the impact of relatively low green water availability 15 by measuring the greenness of vegetation relative to historical observations of greenness. Hence, they reflect whether vegetation is deviating from regular conditions. Since the vegetation drought indicators we have identified are all based on remote-sensing observations, the indicators do not show whether deviations are caused by relatively dry weather (i.e. meteorological drought) or by other factors influencing vegetation 20 growth (e.g. plant diseases or human interference such as pruning and clearing).

Relative soil moisture indicators
In contrast to the absolute soil moisture indicators discussed in Sect. 3.1.1, these indicators measure the moisture conditions at a given location relative to a normal condition. Identified examples are the PZI, SMAI and SD. They are also considered suitable HESSD 12, 2015 Review and classification of indicators of green water availability and scarcity J. F. Schyns et al. for measuring agricultural droughts (Keyantash and Dracup, 2002;Narasimhan and Srinivasan, 2005).

Green water scarcity indicators
As put forward in Sect. 2, water scarcity pertains to a situation with a high water demand compared to water availability, which is experienced by a community (numerous 5 people) within a certain geographic area (e.g. catchment or country) over a significant period of time (months or years). We can then define green water scarcity as the degree of competition over limited green water resources, whereby the demand for green water resources to sustain the production of a desirable level of biomass-based products within a certain geographic area is somehow compared to the available green 10 water resources in space and time.
Since production of biomass-based products (food, fibres, biofuels, timber) generally takes place in cycles of one year (or more in case of perennials and forestry), this definition of green water scarcity incorporates the "significant period of time" element in the imbalance between green water availability and demand. Furthermore, limited 15 production of biomass-based products affects numerous people, both producers and consumers.
As opposed to the indicators discussed in Sect. 3.1, indicators of green water scarcity thus need to include a measure of green water demand, associated with the production of biomass for human purposes, compared to green water availability. In other 20 words, they should measure green water availability in relation to human needs for green water: crop production, grazing lands, forestry. Note that the term "green water availability" here refers to the part of the green water flow available for biomass production for human purposes (in space and time); it thus excludes green water flows that are effectively unavailable, for instance green water flows in unsuitable areas (e.g. 25 because of steep slopes) or green water flows in cold parts of the year unsuitable for growth.
We distinguish three different options to measure green water scarcity conceptually: 5533 Introduction a. Green water crowding: per capita available green water resources in an area compared to a global average threshold representing the amount of green water required to sustain a person's "standard consumption pattern of biomass-based products".
b. Green water requirements for self-sufficiency vs. green water availability: green 5 water requirements for producing the consumed biomass-based products within a certain geographic area, assuming self-sufficiency within the geographic area, compared to the green water resources in the geographic area.
c. Actual green water consumption vs. green water availability: actual green water consumption in a certain geographic area (associated with the actual production 10 of biomass for human purposes) compared to green water availability in the area. This type of indicator thus acknowledges the possibility of virtual water trade as opposed to assuming self-sufficiency as in the previous two types of indicators.
In Sects. 3.2.1 and 3.2.2, we discuss existing indicators that measure aggregated green-blue water scarcity and reflect on how these indicators could be adapted to 15 measure green water scarcity according to above-mentioned options (a) and (b). In Sect. 3.2.3, we elaborate upon a comprehensive indicator of the third form of measuring green water scarcity that has yet to be brought into practice. The challenges for operationalization of these green water scarcity indicators are discussed in Sect. 3.2.4. 20 Rockström et al. (2009) introduced a combined green-blue water shortage index, which compares the sum of green and blue water availability with a global average threshold of 1300 m 3 cap −1 yr −1 . This threshold represents the green and blue water requirements for sustaining a global average "standard diet". When green-blue water availability drops below the threshold, this indicates a shortage of green-blue water re-HESSD 12, 2015 Review and classification of indicators of green water availability and scarcity J. F. Schyns et al. sources. The green-blue water shortage index is an indicator of water crowding, similar to Falkenmark's blue-water focused water crowding indicator (Falkenmark et al., 1989). Similar to the indicator by Rockström et al. (2009), an indicator of green water crowding could be defined as the per capita available green water resources in an area compared to a global average threshold representing the amount of green water required to sustain a person's "standard consumption pattern". We intentionally speak here of a consumption pattern, because green water is not only required to produce food, but also to produce other biomass-based products humans consume, such as fibres, biofuels and forestry products. As such, the measure of green water requirements we propose here is broader than the definition of a "standard diet" according to Rockström et al. (2009) (and Gerten et al., 2011Kummu et al., 2014), which only pertains to water requirements for food production. Rockström et al. (2009) define green water availability as "the soil moisture available for productive vapour flows from agricultural land". Technically, they calculate green water availability as actual evaporation from existing cropland and permanent pasture, 15 reduced by a factor 0.85 that accounts for minimum evaporation losses that are unavoidable in agricultural systems (Rockström et al., 2009). This definition is dependent on the extent of agricultural land and excludes available green water on lands that are currently uncultivated, but have potential to be used productively in a sustainable manner. blue availability in each country of the world. The resulting green-blue water scarcity indicator, computed for each country, is defined as the ratio between green-blue water availability and green-blue water requirements for producing the standard diet.  Rockström et al. (2009), but a bit more conservative: they do not assume year-round evaporation from areas covered with perennial "other" crops (excl. the major food crops) they parameterized as perennial grass (Gerten et al., 2011). Whereas the studies by Rockström et al. (2009)  green-blue water requirements, on which they base their classification of green-blue scarcity: no scarcity; occasional scarcity (subdivided in four levels); or chronic scarcity. The green-blue water scarcity indicator shows the potential of a geographic area (e.g. country or food producing unit) to reach food self-sufficiency and reflects its dependency on trade in agricultural commodities and associated virtual water (Kummu 15 et al., 2014). A similar indicator for green water could show an area's green water demand for self-sufficiency in producing biomass-based products for sustaining the "standard consumption pattern" to green water availability in the area.

Green water crowding
For the potential green water scarcity indicators discussed in Sects. 3.2.1 and 3.2.2, a more comprehensive definition of green water availability is advised than the one 20 applied by Rockström et al. (2009), Gerten et al. (2011 and Kummu et al. (2014). An example of a more comprehensive definition is discussed in the following section.

Actual green water consumption vs. green water availability
The green water scarcity indicator by Hoekstra et al. (2011) compares the actual green water consumption in an area associated with the actual biomass production pattern 25 (hence considering virtual water trade as opposed to assuming self-sufficiency) with green water availability in the area. Green water scarcity is defined as the ratio of the total green water footprint in a catchment x in a period t (e.g. a year) over green water availability. The sum of green water footprints equals all actual evaporation (E act ) related to biomass production for human purposes (i.e. agriculture and forestry) excluding the part of the vapour flow that originates from blue water resources (irrigation). Green 5 water availability is defined as total E act over the catchment minus E act from land reserved for natural vegetation (so called "environmental green water requirement") and minus E act from land that cannot be made productive, e.g. in areas or periods of the year that are unsuitable for crop growth (Hoekstra et al., 2011). In fact, green water availability defined like this, represents the maximum sustainable green water footprint 10 in the catchment and period under consideration. Hence, the green water scarcity ratio shows the extent to which the green water footprint has reached its maximum sustainable level. Of course, this definition can also be applied to other geographical units than a catchment.
The definition of green water availability by Hoekstra et al. (2011) is more compre- 15 hensive than the one used by Rockström et al. (2009), Gerten et al. (2011) and Kummu et al. (2014). However, this is also the reason why the indicator has not been made operational yet. Difficulties remain in estimating the amount of land that needs to be reserved for nature and when and where the green water flow (E act ) cannot be made productive (Hoekstra et al., 2011). These challenges are discussed in the following 20 section. Furthermore, this indicator does not overcome the problem of dealing with the productivity of green water use (Rockström et al., 2009). Transpiration is a productive form of green water use, contributing to biomass production, while other components of the evaporative flow are regarded as unproductive (Rockström, 2001; Rockstrom et al., water scarcity indicator by Hoekstra et al. (2011). A green water scarcity assessment based on both will give insight into the severity of green water scarcity: areas that are considered highly green-water scarce, but have a low transpiration efficiency, may have options to improve the latter and thereby yields, which may lower the green water scarcity.

Challenges for operationalization of green water scarcity indicators
Operationalization of green water scarcity indicators faces three major challenges, particularly regarding the quantification of green water availability. First, the determination of which areas and periods of the year the green water flow can be used productively is not straightforward. Absolute green water availability indi-10 cators, in particular land classifications of agricultural suitability, can provide insight in the availability of green water in the spatial dimension. Relative green water availability indicators can enrich the picture by showing which areas are prone to large inter-and intra-annual variations in green water availability, making these areas less suitable for (certain types of) biomass production. To estimate which part of the green water flow 15 can be used productively in time, advanced crop growth models (like APSIM (McCown et al., 1995;Holzworth et al., 2014), AquaCrop (Steduto et al., 2009), CropSyst (Stöckle et al., 2003), EPIC (Jones et al., 1991) or SWAP/WOFOST, van Dam et al., 2008) can be used to simulate water-limited yields and actual evaporation for various cropping periods and different types of soil, crop and agricultural water management (e.g. adding 20 blue water in the form of deficit irrigation during a dry spell, might make it possible for the crop to survive and use the green water flow later in the year productively).
Second, estimating green water consumption of forestry is difficult, because it entails separation of production forest evaporation into green and blue parts. This is problematic, because trees generally root so deep that, by means of capillary rise, they directly Third, research is required to determine the environmental green water requirements, i.e. the green water flow that should be preserved for nature, similar to the environmental flow requirements for blue water. Key here is the identification of areas that need to be reserved for nature and biodiversity conservation. It is known that the current network of protected areas is insufficient to conserve biodiversity (Rodrigues 5 et al., 2004a, b;Venter et al., 2014;Butchart et al., 2015) and that attention should be paid to conservation of biodiversity in production landscapes that are shared with humans (Baudron and Giller, 2014). The 11th Aichi Biodiversity Target is to expand the protected area network, which currently has a terrestrial coverage of about 14.6 % (Butchart et al., 2015), to at least 17 % terrestrial coverage by 2020 (Convention on Biological Diversity, 2010). However, to properly assess the limitations to green water availability, spatially explicit information on the additional areas to be preserved is required. The best-available data regarding this is recently published work by Montesino

Conclusions and future research
In this paper we have reviewed and classified around eighty indicators of green water availability and scarcity. This list of indicators is extensive, but not exhaustive. Nevertheless, we are confident to have identified the most widely used and cited indicators. 20 The number of green water availability indicators by far outnumbers the existing green water scarcity indicators. This reflects that the concept of green water scarcity is still largely unexplored. Indicators of green-blue water crowding and scarcity have been developed by Rockström et al. (2009), Gerten et al. (2011 and Kummu et al. (2014). These have potential to be tailored to measure green water crowding and green water requirements for self-sufficiency vs. green water availability. The green water scarcity indicator by Hoekstra et al. (2011)  water availability, but has not yet been operationalized due to several challenges discussed in Sect. 3.2.4. The biggest challenge is to determine which part of the green water flow can be made productive in space and time. Application of both absolute and relative green water availability indicators will provide insight in which areas the green water flow can be made productive for human purposes. Simulations with crop growth 5 models for different management strategies can be used to assess which parts of the year the green water flow can be made productive. Future research should be aimed at overcoming these challenges to make the green water scarcity indicators discussed in this paper operational. We also encourage the development of additional definitions of green water scarcity indicators to the ones 10 discussed here. The conceptual definition of green water scarcity we introduced in Sect. 3.2 can be a starting point for this.

Bhalme and Mooley
Gommes and Petrassi (1994) Effective Drought Index EDI Ratio of the difference between effective precipitation (EP, calculated from equations based on precipitation) and its 5 day running mean over the standard deviation of this difference.