1 Article

2 How much environmental flow requirements are

3 transgressed? A global scale analysis
4 AV Pastor1, 2,*, H Biemans 3, W Franssen 2, D Gerten 4,5, H. Hoff 6, F Ludwig 2, P Kabat 2,7
5 1 Centre for Ecology, Evolution and Environmental Changes (CE3C), Climate Change Impacts, Adaptation
6 and Modelling (CCIAM), Faculdade de Ciências da Universidade de Lisboa, Campo Grande, 1749-016
7 Lisboa, Portugal
8 2 Water Systems and Global Change Group, Wageningen University, Wageningen UR, PO Box 47 6700 AA,

9 Wageningen, The Netherlands

10 3 Climate Change and Adaptive Land and Water Management Group, Wageningen University and Research,

11 PO Box 47, 6700AA Wageningen, The Netherlands

12 4 Research Domain of Earth System Analysis, Potsdam Institute for Climate Impact Research

13 (PIK), Telegraphenberg A62, 14473 Potsdam, Germany

14 5 Geography Department, Humboldt-Universität zu Berlin, Unter den Linden 6, 10099 Berlin, Germany

15 6 Stockholm Environment Institute, Kräftriket 2b, SE-106 91 Stockholm, Sweden

16 7 International Institute for Applied Systems Analysis, Schlossplatz 1, A-2361 Laxenburg, Austria

17 * Correspondence: amandine.pastor22@gmail.com

18 Received: date; Accepted: date; Published: date

19 Highlights
20 - Water abstraction is the main cause of environmental flow deficit globally
21 - Intermittent rivers are more prone to environmental flow deficit than perennial rivers
22 - Natural climate variability usually exacerbates environmental flow deficit especially in
23 intermittent rivers

24 Abstract: Freshwater ecosystems have degraded over the last several decades as a result of intense
25 freshwater abstraction. Therefore, environmental flow requirements (EFRs) methods have been
26 proposed to maintain healthy rivers and/or to restore river flows. In this study, we used the
27 Variable Monthly Flow method with two thresholds (normal and high) to calculate EFRs on a
28 global scale. We used refined spatial (0.5°) and temporal scales (monthly) to estimate two kinds of
29 environmental flow deficits: (1) natural deficits in which flow does not meet EFRs due to climate
30 variability, and (2) anthropogenic deficit caused by water abstractions. Results show that total
31 annual EF deficit represents 5% of the total global discharge. However, on an intra-annual scale,
32 total EF deficit can exceed monthly flow. Perennial rivers with low flow alteration such as the
33 Amazon showed an EF deficit of 2-12% of total discharge, and the natural deficit was responsible
34 for up to 75% of the total deficit. In rivers driven by high seasonality and with high water
35 abstractions such as the Indus, the total deficit represents up to 130% of its total discharge of which
36 85% is due to abstractions. Globally, the combined deficits represent between 16% and 36 % of the
37 total surface withdrawals showing potential conflicts with EFRs. With the inclusion of EFRs, we
38 show monthly global water stress indicator (WSI) ranges from 30 to 50% in temperate zones and in
39 Latin America therefore classified as “moderately exploited basins” and WSI was well above 50%
40 in the MENA and South of Asia showing basins “environmentally water stressed and/or scarce”.
41 This study highlights the need to urgently find solutions to sustainably allocate water for both,
42 humans and ecosystems.

43 Keywords: environmental flow, natural deficit, anthropogenic deficit, climate variability, water
44 abstraction, water stress index
45

Water 2019, 11, x; doi: FOR PEER REVIEW www.mdpi.com/journal/water

 Water 2019, 11, x FOR PEER REVIEW 2 of 19

46 1. Introduction
47 In the last decades, increased human water use has led to the flow alteration of major river
48 basins such as the Indus and the Colorado Rivers with 30% overuse of non-renewable water
49 resources such as groundwater from deep aquifers (Grafton et al. 2013; Wada and Bierkens 2014).
50 Flow alteration due to water abstraction and loss of river connectivity is the primary cause of
51 freshwater ecosystems damage (A H Arthington et al. 2006; Poff and Zimmerman 2010). There is a
52 growing concern about the declines of natural resources including freshwater biodiversity loss
53 (Gleick 2003; WWF/ZSL 2016) and maintaining rivers close to their natural flow regime is necessary
54 to fulfil ecosystem functions (M. C. Acreman and Ferguson 2009). Recent studies show the impact of
55 multiple anthropogenic stressors at global scale and stressors of ecological status of European rivers
56 (Vorosmarty et al. 2010; Ferreira, Globevnik, and Schinegger 2019). To limit and reduce freshwater
57 ecosystem degradation, river restoration projects have emerged with the aim to restore freshwater
58 ecosystems to acceptable ecological conditions. To maintain rivers in “acceptable” ecological
59 conditions, Environmental Flow Requirements (EFRs) have been defined as “the quantity, timing,
60 and quality of freshwater flows and levels necessary to sustain aquatic ecosystems which, in turn,
61 support human cultures, economies, sustainable livelihoods, and well-being” (Angela H Arthington
62 et al. 2018; Declaration 2007).
63 In this study we developed the concept of “environmental flow deficit” to show in time and
64 space the water flow that is lacking in a river on an intra-annual variability basis (e.g. seasonal water
65 deficit) and on an inter-annual basis (e.g. supra-seasonal water deficit). This is a new concept that
66 shows potential water stress produced on freshwater ecosystems when water is not available for
67 ecosystems due to natural climate variability or anthropogenic pressures. Contrary to previous
68 water stress indicators that were used to show potential impact of water stress on humans, this latter
69 focuses on environment and more precisely EFRs. We differentiate environmental deficits caused by
70 natural climate variability (natural deficit) and caused by water withdrawals (anthropogenic deficit),
71 as it is important to recognize the causes of water deficit in a river to anticipate better water
72 management. Natural deficit is the result of a hydrological drought usually following a
73 meteorological drought where ecosystems have a margin of manoeuvre to adapt and recover from
74 droughts (Botter et al. 2013). On the other hand, if deficit is caused by water abstraction, it can be
75 detrimental to freshwater ecosystems and even lead to collapse. Therefore, it is important to
76 differentiate the causes of EF deficits, which we do in this study for 23 major river basins with
77 different flow regimes: perennial rivers and intermittent rivers. While perennial rivers are known to
78 have a stable flow all year long, intermittent rivers are characterized with period of flow cessations
79 (Gordon, McMahon, and Finlayson 2004).
80 Our objectives are: i) differentiate natural and anthropogenic EF deficits, ii) to define global EF
81 deficits both in terms of intra-annual and inter-annual variability, iii) to deepen the understanding of
82 EF deficit in terms of timing, frequency and duration, iv) to cluster river basins in terms of flow
83 regime, flow alteration and sensitivity to EF deficit and v) and finally to calculate a new water stress
84 indicator worldwide which includes EF deficits.

85 2. Materials and Methods

86 2.1. Model
87 We used the global dynamic vegetation and hydrological model LPJmL (Bondeau et al. 2007;
88 Dieter Gerten et al. 2004; Rost et al. 2008; Jägermeyr et al. 2015) to simulate natural discharge
89 (without anthropogenic water use) and actual discharge (including anthropogenic water use for
90 irrigation, industries and households). LPJmL was developed for consistently simulating water and
91 carbon cycles globally for natural and agricultural vegetation. LPJmL calculates daily water balances
92 at 0.5 degree spatial resolution with flow components laterally routed along the Simulated
93 Topological Network (STN-30) with a constant flow velocity of 1 m/s (Biemans et al. 2011). Runoff
94 occurs when water input exceeds soil water capacity in a multi-layer soil column, which is routed
95 downstream according to the topology. For details on the simulation of crop growth and irrigation
 Water 2019, 11, x FOR PEER REVIEW 3 of 19

96 requirements see Rost et al. (2008). Each cell with crops and pasture can be irrigated and/or rainfed
97 based on a map of irrigated areas (Fader et al. 2010). Water is extracted for irrigation and other uses
98 from surface water including rivers, dams and reservoirs, calculated according to crop-specific daily
99 demand (Biemans et al. 2011). Water use for industry and households are also included in the
100 simulation rounding about 750 km3 yr-1 (Flörke et al. 2013). Conveyance losses which represent the
101 volume of water that is lost during transport, and irrigation use efficiency are included in the
102 irrigation calculations to provide gross irrigation demand (Rost et al. 2008). Net irrigation
103 requirements are computed at daily temporal resolution and 0.5 degree spatial resolution, based on
104 evaporation demand and soil water capacity requirements. In this study we used fixed land use at
105 year 2000 (Fader et al. 2010) to evaluate the impact of inter-annual climate variability. Irrigation
106 withdrawals were calculated assuming that any irrigation requirement can be met (Biemans et al.
107 2011; Rost et al. 2008). Inside a river basin, runoff and EFRs calculation was re-allocated according to
108 discharge repartition (Schewe et al. 2014).
109 We forced the model with the input dataset CRU T.S.3.10 (available online at:
110 http://badc.nerc.ac.uk/data/cru/). The input consists of monthly values of precipitation, number of
111 wet days, fraction cloud cover and average temperature for 41 years (1960-2000). The model was first
112 run with a spinup of 1000 years to put carbon and water cycle into equilibrium at year 1960. Then,
113 we ran the model for 41 years (1960-2000) to include inter-annual variability in our analyses.

114 2.2. Environmental flow method

115 In this study, we use the Variable Monthly Flow method VMF (Pastor et al. 2014) to calculate
116 Environmental Flow Requirements (EFRs) globally. This method allocates a share of the natural
117 monthly flow to freshwater ecosystems. Each grid cell has a specific seasonal flow regime
118 (hydrograph) which was divided into three periods: low-flow, high flow and intermediate flow
119 periods. High flows are assigned when the mean monthly flow (MMF) is above 80% of the mean
120 annual flow (MAF), low flows are assigned when the MMF is below 40% of the MAF, and
121 intermediate flows are assigned when the MMF is between 40% and 80% of the MAF. We allocate
122 then 30% of the MMF for high flow requirements, 45% of the MMF for intermediate flow
123 requirements and 60% of the MMF for low flow requirements. The share of EFRs for low flows are
124 higher than the rest of the season to protect habitat maintenance and allow good spawning
125 conditions. In this study, we define net discharge as the total discharge minus EFRs.

126 2.3. Deficit per river basin

127 We define two types of EF deficits: the natural deficit and the anthropogenic deficit. Natural
128 deficit represents the flow that is lacking due to natural climate variability only, and anthropogenic
129 deficit represents the flow that is lacking due to water abstraction for irrigation, industry and
130 households. Explicitly, natural deficit happens when monthly flow does not meet EFRs even in the
131 absence of anthropogenic water abstractions. By using monthly flow data we define the intra-annual
132 deficit to identify the timing and the duration of the deficit. We also calculate the frequency of
133 natural deficit in terms of number of months and years. To calculate the anthropogenic deficit on the
134 global scale we differentiated water withdrawals from surface water and groundwater.
135 Groundwater withdrawals account for 851 km3 yr-1. and was calculated as a share of total potential
136 withdrawals per country (Siebert et al. 2010). The EF deficit is calculated on the grid cell and was
137 aggregated to the river basin and country scale depending on the analysis. EF deficits are defined
138 below:
𝑦,𝑚
139 𝐸𝐹_def⁡_𝑛𝑎𝑡(𝑚, 𝑦) = ∑ (𝑄𝑛𝑎𝑡(𝑚, 𝑦) − 𝐸𝐹𝑅(𝑚)) [1]
𝑖=1
𝑗=1
𝑦,𝑚
140 𝐸𝐹_def⁡_𝑎𝑛𝑡(𝑚, 𝑦) = ∑ (𝑄𝑚𝑜𝑑(𝑚, 𝑦) − 𝐸𝐹𝑅(𝑚)) [2]
𝑖=1
𝑗=1
141 Where EF_def_nat represents the natural deficit and EF_def_ant represents the anthropogenic
142 deficit. Qnat represents the natural discharge and Qmod represents the actual modified discharge
143 including water withdrawals. y represents the number of years (y=41) to iterate starting from i and
 Water 2019, 11, x FOR PEER REVIEW 4 of 19

144 m represents the number of months (m=12) to iterate starting from j. Anthropogenic and natural
145 deficits are differentiated by calculating deficits as river discharge minus EFRs with and without
146 human water abstraction.

147 2.4. Sensitivity analyses

148 For sensitivity analyses and uncertainty range, we calculated deficits for two levels of EFRs:
149 normal (VMF method) and high (VMF method + 1SD, see (Shadkam et al. 2016). The standard
150 deviation of EFRs was calculated over the 41 years’ time period.

151 2.5. Classification of river basins

152 We use a principal component analysis (PCA) to identify river basins clusters by using six
153 hydrological variables. These hydrological variables were chosen to define river flow regimes, level
154 of flow alteration, differentiation between anthropogenic and natural deficit and finally to define
155 duration, timing and frequency of deficit. A PCA of a data matrix extracts the main patterns in the
156 matrix in terms of a matching set of score and loading plots (Wold, Esbensen, and Geladi 1987). We
157 selected the hydrological indicators as input in the PCA when the highest variance was explained.
158 We selected 23 river basins worldwide (see Table 3) to represent three different river flow regimes
159 (perennial, intermittent and erratic), two different climates (tropical and temperate) and two levels
160 of flow alteration (below 50% alteration and above 50% of flow alteration).
161 To characterize flow regime, we used two variables: the baseflow index (BFI) (Beck et al. 2013)
162 and the Hydrological Variability Index (HVI) variables were defined as below:
Q90
163 BFI = ⁡
Total⁡streamflow
[3]
Q25−Q75
164 HVI =
Q50
[4]
165 Where Q90 represents the flow which is exceeded for 90% of the period of record, MAF represents
166 the mean annual flow, Q25 represents the flow which is exceeded for 25% of the period of record,
167 Q75 represents the flow which is exceeded for 75% of the period of record, and Q50 represents the
168 flow which is exceeded for 50% of the period of record. All our calculations are based on monthly
169 output where volumes of water are in km3. MOD represents the level of flow modification and is
170 defined as the mean annual actual flow (including anthropogenic water abstractions) over the mean
171 annual natural flow. The results of the PCA allowed the definition of four river categories using the
172 baseflow index (BFI) and the MOD variables:
173 1. Group 1: Perennial flow regime with low flow modification: BFI >= 0.3 and MOD>0.95
174 2. Group 2: Perennial/intermittent flow regime with low flow modification: 0.15 <= BFI <0.3
175 and MOD>0.7
176 3. Group 3: Perennial/intermittent flow regime with moderate flow modification: 0.01 <= BFI
177 <0.15 and MOD>0.7.
178 4. Group 4 : Perennial/intermittent flow regime with high flow modification: 0.01 <= BFI <0.15
179 and MOD<=0.7
180 To perform the PCA, a couple of hydrological indicators were used to define the range of origin,
181 magnitude, timing and frequency of EF deficit:
182 - Total deficit to discharge
y,m
Nat.⁡⁡Deficit⁡(m,y)+Ant.⁡⁡Deficit⁡(m,y)
183 DTD(m, y) = ∑
Qnat⁡(m,y)
[5]
i=1
j=1
184 - Natural deficit to discharge
y,m
Nat.⁡⁡Deficit⁡(m,y)
185 DTN(m, y) = ∑
Qnat⁡(m,y)
[6]
i=1
j=1
186 - Anthropogenic deficit to discharge
y,m
Ant.⁡⁡Deficit⁡(m,y)
187 DTA(m, y) = ∑
Qnat⁡(m,y)
[7]
i=1
j=1
188 - Frequency of total deficit – sum of number of months deficit
 Water 2019, 11, x FOR PEER REVIEW 5 of 19

y
189 Freq_tot(y) = ∑ MEFdef1(y) + MEFdef2(y) [8]
j=1
190 - Frequency of natural deficit – sum of number of months deficit
y
191 Freq_nat(y) = ∑ MEFdef1(y) [9]
j=1
192 - Frequency of anthropogenic deficit – sum of number of months deficit
y,m
193 Freq_ant(m, y) = ∑ MEFdef2(m,y) [10]
i=1
j=1
194 - Ratio of anthropogenic deficit over natural deficit
y,m
Ant.⁡⁡Deficit⁡(m,y)
195 ATN(m, y) = ∑
Nat.Deficit⁡(m,y)
[11]
i=1
j=1
196 - Ratio of the frequency anthropogenic deficit over natural deficit
y,m
Freq_ant.⁡⁡(m,y)
197 FATN(m, y) = ∑
Freq_nat.⁡⁡⁡(m,y)
[12]
i=1
j=1
198 - Ratio of water withdrawals over discharge
y,m
Wd⁡⁡(m,y)
199 WTD(m, y) = ∑
Qnat⁡⁡(m,y)
[13]
i=1
j=1
200 - Ratio of deficit over water withdrawals
y,m
EFdef1,2⁡⁡(m,y)
201 DTW(m, y) = ∑
Wd⁡⁡(m,y)
[14]
i=1
j=1
202 - Ratio of natural deficit over water withdrawals
y,m
EFdef1⁡⁡(m,y)
203 ATW(m, y) = ∑
Wd⁡⁡(m,y)
[15]
i=1
j=1
204 - Ratio of anthropogenic deficit over water withdrawals coming from surface water
y,m
EFdef1⁡⁡(m,y)
205 ATWs(m, y) = ∑
Wd_surf⁡⁡(m,y)
[16]
i=1
j=1
206 where MEF represents months with environmental flow deficit and Wd and Wd_surf represent
207 respectively total water withdrawals and water withdrawals coming from surface water only. I,j,m
208 and y are defined in the text above.
209 Finally, the water stress indicator (WSI) was defined such as in (Smakhtin, Revenga, and Doll
210 2004). WSI is calculated with and without EFR.
y,m
Wd(m,y)⁡(+𝐸𝐹𝑅𝑠)
211 WSI(m, y) = ∑
Water⁡available⁡(m,y)
[15]
i=1
j=1
212 If WSI exceeds 100% of discharge, the basin is classified as “environmentally water scarce”. In
213 this case, total withdrawals are outpacing total surface water availability so that there is no water left
214 for EFRs. Basins where 60% < WSI < 100% are defined as “environmentally water stressed” and
215 basins where 30% < WSI < 60% as moderately exploited. Finally, when WSI < 30%, the basins are
216 defined as “environmentally safe”.
217
218
219
220
221
222
223
224
225
226
 Water 2019, 11, x FOR PEER REVIEW 6 of 19

227 Table 3. Hydrological indexes of 23 river basins including discharge, EFRs and natural and anthropogenic deficit.
Frequen
Mod: Water cy Share of Share of
Hydrologic EFR Share of Share of Share of Share of
ratio of Withdra total anthropo ant.
Discharg Baseflo al range EFR to Total Anthrop. total ant. ant.
River actual wals from deficit genic deficit to
e w Index Variability (low-high discharge deficit deficit deficit to deficit to deficit to
basin flow on IRR and (nb on total surface
(km3 yr-1) (BFI) Index ; km3 (Low-hig (km3 yr-1) (km3 yr-1) Discharg discharge water
natural OTH months deficit water
(HVI) yr-1) h; %) e (%) (%) Wd. (%)
flow (km3 yr-1) out of (%) Wd. (%)
41)
7495 ± 2419 -125 ±
Amazon 1.00 0.47 0.76 32-49% 7-8 0±0 25 0% -2% 0% 0% 0%
886 -3690 538
Amu
61 ±28 0.30 0.08 1.46 21-42 35-68% 86-93 -78 ±21 -56±15 360 72% -127% -91% -62% -28%
darya
Colorado 35 ±22 0.74 0.25 0.84 11-24 31-68% 8-12 -7 ±4 -3±2 156 48% -20% -10% -34% -21%
Columbia 171 ±62 0.94 0.22 0.94 54-98 32-57% 12-16 -12 ±15 -4±5 84 37% -7% -3% -32% -20%
2287 ±
Congo 1.00 0.54 0.49 722-1141 32-50% 1-2 -37 ±160 0±0 25 0% -2% 0% 0% 0%
389
Danube 207 ±73 0.99 0.38 0.71 68-118 33-57% 19-21 -13 ±18 -4±5 54 30% -6% -2% -19% -2%
Ebro 19 ±10 0.79 0.19 0.87 6-12 32-65% 4-7 -4 ±3 -2±2 184 54% -24% -13% -43% -16%
Euphrates 91 ±45 0.68 0.02 1.62 27-57 30-63% 52-60 -45 ±19 -29±12 286 65% -49% -32% -52% -23%
Ganges 1281 ±32 0.96 0.05 1.69 403-640 31-50% 183-189 -142 ±112 -116±91 227 81% -11% -9% -62% -35%
Garonne 33 ±16 0.97 0.27 0.94 11-21 33-63% 1-5 -2 ±3 -1±1 123 43% -7% -3% -34% -12%
Godavari 171 ±61 0.92 0.03 1.79 53-99 31-58% 30-36 -29 ±18 -17±10 285 59% -17% -10% -52% -29%
Guadalqui 33-91
10 ±9 0.70 0.01 1.28 3-9 3-11 -6 ±3 -3±1 314 42% -64% -27% -41% -15%
vir %
Indus 155 ±53 0.22 0.11 1.15 52-94 34-60% 255-263 -201 ±60 -172±52 461 85% -130% -111% -66% -37%
Mekong 609 ±124 0.99 0.05 1.74 190-310 31-51% 17-23 -20 ±45 -15±33 143 74% -3% -2% -76% -22%
Mississip
818 ±258 0.96 0.51 0.56 264-441 32-54% 67-68 -53 ±67 -5±7 39 10% -6% -1% -8% -5%
pi
Murray 109 ±75 0.92 0.24 0.82 33-80 30-73% 18-21 -15 ±12 -6±5 128 41% -14% -6% -32% -8%
Niger 500 ±94 0.99 0.03 1.66 153-249 31-50% 3-9 -10 ±35 -4±13 95 39% -2% -1% -72% -10%
Nile 597 ±125 0.94 0.10 1.45 243-390 41-65% 55-58 -39 ±62 -23±37 108 59% -7% -4% -41% -13%
Orange 60 ±47 0.97 0.04 1.14 21-54 35-89% 1-7 -5 ±6 -1±2 182 24% -8% -2% -28% -2%
Rhine 75 ±33 0.99 0.43 0.65 23-45 31-60% 18-20 -9 ±9 -5±6 112 60% -12% -7% -27% -14%
Yangtze 732 ±184 0.97 0.24 1.01 233-388 32-53% 98-102 -72 ±66 -42±39 78 59% -10% -6% -42% -12%
Yellow
72 ±24 0.64 0.03 1.52 24-41 33-58% 45-57 -39 ±17 -29±13 303 75% -54% -40% -57% -16%
river
Zambezi 400 ±41 0.99 0.01 1.92 127-233 32-58% 0-5 -9 ±30 -5±16 151 54% -2% -1% -220% -17%

228
 Water 2019, 11, x FOR PEER REVIEW 7 of 19

229 3. Results
230 Our simulations show that total annual discharge is about 44,000 km 3 per year of which EFRs
231 represent between 18,000 to 28,450 km3 year-1 (40-60% of global discharge) depending on the chosen
232 EFRs threshold. We show that total EF deficit represents 5% of the total global discharge and ranges
233 from 2,321 to 2,706 km3 yr-1 (Table 1). Irrigation withdrawals have almost doubled within 41 years
234 from 1,088 km3 year-1 in 1960 to 2,075 km3 year-1 in 2000. The anthropogenic deficit ranges from 1,068
235 to 1,354 km3 yr-1 and represents about 50% of total annual deficit (Figure 1; Table 1 and
236 Supplementary Figure 1). We also show that anthropogenic deficit coming from surface water only
237 represents between 260 to 545 km3 yr-1 (about 25% to 50% of the total anthropogenic deficit). While
238 total annual deficit does not sound very dramatic, we show here that deficit can be particularly high
239 at specific temporal and regional scales (Fig. 2). For example, the highest monthly deficits are
240 observed in South and East Asia (up to 80 km3 month-1) with a duration of at least five months
241 (between December and May). Among them, the Indus River shows the maximum annual deficit
242 (200 km3 per year representing 130% of total discharge). the fact that the deficit can exceed the river
243 flow, is explained by additional withdrawals coming from groundwater from shallow and deep
244 aquifers (or other sources). The frequency of the deficit can reach up to 90% of the time (Fig 2, Table
245 3) indicating that in the Indus, monthly flows meets EFRs for only 10% of the time. North African
246 and Mediterranean river basins also show large deficits especially in the summer. For example, the
247 Euphrates has a deficit of up to 25 km3 per month and up to 60 km3 per year (deficit equals 50% of
248 the discharge). In the last case, a deficit occurs at least half of the time. Large deficits are also
249 observed in the Western part of the United States such as in the Colorado and the Columbia rivers.
250 Finally, perennial rivers such as the Congo and the Amazon rivers show a very low EF deficit (< 2%
251 of the total discharge) coming mainly from natural deficit and occurring during less than 5% of the
252 time.

253
254 Figure 1. Spatial representation of monthly deficit on the global scale for four months: Jan, Apr, Jul,
255 Oct averaged over 41 years. Units are in km3 month-1. Negative values represent deficit of flow (in
256 yellow/orange/red) whereas positive values (in blue) represent surpluses of flow.
 Water 2019, 11, x FOR PEER REVIEW 8 of 19

257 Table 1. Global numbers of discharge, EFRs, deficit and irrigation (km3 year-1) as a share of surface
258 water and groundwater withdrawals (%).

Absolute
Share of Irrigation
values
Wd (%)
(km3 yr-1)
Discharge (±sd) 44387±1619 -
18089
EFR range (EFR-EFR+2sd) -
-28450
Irrigation withdrawals range (1960-2000) 1088-2075 -
Total-annual deficit range (1960-2000) 2321-2706 -
Anthropogenic annual deficit range (1960-2000) 1068-1353 66-84%
Anthropogenic annual deficit (considering 50% irrigation from
260-545 16-34%
groundwater) (1960-2000)

259
260 Figure 2. Principal Component Analysis of 23 river basin using 6 hydrological variables: EF_def2
261 (Total deficit), Freq_nat (Natural deficit), Freq_ant (Anthropogenic deficit), BFI (Baseflow index),
262 HVI (hydrological variability index) and MOD (the rate of flow alteration).

263 A PCA was performed to categorize four groups of river basins (Figure 2, Table 2) and
264 localization of river basins is represented in supplementary Figure S2. The PCA was defined by two
265 components that explain 79% of the variance of the 23 river basins. Six out of 16 hydrological
266 variables were selected when we reached the maximum variance explanation. The first component
267 explains 52% of the variance which is characterized by the BFI and the MOD variables on the
 Water 2019, 11, x FOR PEER REVIEW 9 of 19

268 positive axis (+0.4) and by the frequency of the natural deficit (Freq_nat), HVI and the frequency of
269 the anthropogenic deficit (Freq_ant) on the negative side of the axis (up to -0.4). The second
270 component is explained by the anthropogenic EF deficit (EF_def_ant) as positive value (up to 0.4)
271 and the Freq_ant and MOD values as negative values (up to -0.2). We notice that river basins groups
272 are following the component 1 with groups starting from the least altered flow (group 1) up to the
273 most altered flow (group 4).

274
275 Figure 3. Temporal representation of natural and actual flow (left plot) with intra-annual monthly
276 deficit (km3 month-1) and frequency of deficit (number of months) for four river basins representing
277 each group of river basins. Y axis represent absolute flow and deficit have the same scale to enable
278 comparison, Barplots (right plot) represent the frequency of natural deficit (black) and anthropogenic
 Water 2019, 11, x FOR PEER REVIEW 10 of 19

279 deficit (grey). Total Environmental Deficit (EF) (yellow range-right plot) was shown with positive
280 values for comparison with the flow (blue/grey range – left plot).

281 Table 2. Classification of 23 river basins into four groups according to hydrological indexes. Values
282 are showing range of minimum and maximum of each group.
Tot. Anthrop.
Tot. Tot.
Flow Freq. to
BFI deficit Wd.
(km3 HVI (nb. total DTD WTD DTW
(km3 (km3
yr-1) of deficit
yr-1) yr-1)
months) ratio
75
Group 1 0.47-0.54 0.49-0.76 9-125 1-67 25-112 0-27% 2-12% 0-25% 0-27%
-7495
Group 2 19-732 0.1-0.25 0.82-1.45 2-72 3-100 78-184 37-59% 7-24% 8-31% 32-42%
60-128
Group 3 0.01-0.05 1.14-1.92 5-142 2-186 95-285 24-81% 3-17 1-19% 28-72%
1
Group 4 10-155 0.01-0.11 1.15-1.62 6-201 7-259 314-461 42-85% 49-130% 62-168% 41-66%
283 Note: BFI = Baseflow index; HVI = Hydrological variability index; Tot. deficit = Total deficit; Tot.
284 Wd.= Total withdrawals; Tot. Freq.= Total frequency; Anthrop. To nat. deficit = anthropogenic to
285 natural deficit ratio; DTD = Deficit to Discharge ratio; WTD= Withdrawals to discharge ratio; DTW=
286 Deficit to withdrawals ratio. Group 1 contains Amazon, Congo, Danube, Mississipi and Rhine.
287 Group 2 contains Colorado, Columbia, Ebro, Garonne, Nile and Yangtze. Group 3 contains ganges,
288 Godaravi, Mekong, Niger, Orange and Zambezi. Group 4 contains Amu Darya, Euphrates, Guadalkivir
289 and Yellow River.

290 Group 1 is along the axis of BFI (Amazon, Congo, Mississippi, Rhine, Danube). This group is
291 characterized by rivers with low flow variability: BFI >0.38 and HVI <0.76 (Table 2; Figure 3a and
292 supplementary Figure S3). This group is dominated with perennial rivers of which the deficit is
293 mainly natural (25-45 months out of 492 months) compared to the anthropogenic deficit (0 to 87
294 months out of 492 months). Group 1 has also the smallest Deficit To Discharge (DTD) ratio (2-12%).
295 Anthropogenic deficit to discharge ratio represents 0 to 7% of the discharge and Deficit to
296 Withdrawals (DTW) reaches up to 27% of withdrawals (e.g. Rhine). For example, the Congo River
297 has the most stable flow regime (BFI=0.54, HVI=0.49) and shows barely any flow modification except
298 between January and March. Finally, the Congo River is characterized with a low deficit frequency
299 (2 years out of 41) with no anthropogenic deficit.
300 Group 2 is along the axis of MOD variable (Yangtze, Columbia, Garonne, Colorado, Nile). This
301 group is characterized by less stable flows (BFI=0.1-0.27 and HVI=0.74-0.97) (Table 2; Figure 3b and
302 supplementary Figure S3). Frequency of deficit is moderate with 78 months to 184 months deficit out
303 of 492 months of which 37 to 60% is due to anthropogenic deficit. DTD represents between 7 to 24%
304 (e.g. Ebro) and anthropogenic deficit represents 3-13% of total discharge. Potential affected
305 withdrawals (deficit to withdrawals) represent 34% to 43% of withdrawals and potential affected
306 withdrawals from surface water can reach up to 31%. For example, the total flow of the Columbia
307 River equals 171 km3 year-1 of which 8% are withdrawn for irrigation and other users. Total deficit
308 equals 12 km3 (7% of total discharge) of which natural deficit represents 63% of total deficit and is
309 occurring between August and February. Anthropogenic deficit equals 3% of total flow and shows a
310 frequency of 31 months out of 492 (between June and October). Potentially affected water
311 withdrawals represent up to 32% (of which 20% affect surface water withdrawals).
312 Group 3 is characterized by the variables Freq_ant and MOD and includes the following rivers:
313 Godavari, Niger, Zambezi, Orange, Mekong. This group represents a cluster of river basins with
314 high seasonality and with a moderate flow modification (BFI=0.01-0.05, HVI=1.14-1.92) (Table 2;
315 Figure 3c and supplementary Figure S3). The frequency of the deficit is 95 to 285 out of 492 months,
316 and 18% to 58% of the deficits are anthropogenic. DTD ranges between 2 and 17% (e.g. Godavari).
317 Potentially affected withdrawals represent 28 to 72% of total withdrawals (4-66% for withdrawals
318 from surface water only). For example, the Ganges has an intermittent flow regime with a BFI of 0.05
319 and a HVI of 1.69. Withdrawals represent 186 km3 (15% of total flow). Total deficit equals 142 km 3
320 year-1 (11% of total flow) and spreads over 227 months (<50% of the time) of which 185 months are
321 due to anthropogenic deficit (81% of total deficit). The duration of the deficit is about 4 months
 Water 2019, 11, x FOR PEER REVIEW 11 of 19

322 (between November and April) and potentially affected withdrawals represent 62% (of which 26%
323 affect surface water withdrawals).
324 Group 4 is highly correlated with EF_def_ant (especially the Indus), with Freq_nat (Amu
325 Darya) and with HVI and Freq_ant (Yellow river Euphrates and Guadalquivir). This group
326 represents a cluster of river basins with an intermittent to ephemeral flow regime and with a high
327 flow modification (BFI=0.01-0.11, HVI=1.15-1.62, MOD=0.22-0.7) (Table 2; Figure 3d and
328 supplementary Figure S3). The frequency of the deficit is 286-461 months out of 492 months, and
329 42% to 85% of the deficit is anthropogenic. DTD ranges between 49 and 130% (the Indus). Potentially
330 affected withdrawals represent 41 to 66% of total withdrawals (of which 8-36% affect surface water
331 only). For example, the Guadalquivir has an intermittent flow regime with a BFI of 0.01 and a HVI of
332 1.28. Withdrawals represent 7 km3 (66% of total flow). Total deficit equals 6 km3 year-1 (64% of total
333 flow) and spreads over 314 months (64% of the time) of which 133 months are due to anthropogenic
334 deficit (42% of total deficit). Duration of the deficit is about six months (between May and October)
335 and potentially affected withdrawals represent 41% (of which 8% affect surface water withdrawals).
336 We show that the higher the BFI is, the lower the frequency of natural deficit is (R 2=0.35),
337 meaning that perennial rivers encounter less deficit caused by climate variability than intermittent
338 rivers do. Similarly, the BFI is negatively correlated with the frequency of anthropogenic deficit
339 (R2=0.32), meaning that intermittent rivers encounter more deficit due to anthropogenic sources then
340 perennial rivers do.
341 We present one river basin per group in Fig. 4 showing the magnitude of monthly natural and
342 actual flows (left plot) and the magnitude and frequency of natural and anthropogenic deficits (right
343 plot). The rest of the river basins and respective deficits are shown in supplementary Figure S3.
344 Group 1 is represented by the Congo River (Figure 3a) with a low modification of the natural
345 flow (in blue) compared to the actual flow (in grey). The absolute monthly deficit is below 15 km 3
346 and frequency of the monthly deficit is low (below 5 months over 41). The total deficit is coming
347 from climate variability only.
348 (a)

349
350 (b)
 Water 2019, 11, x FOR PEER REVIEW 12 of 19

351
352 Figure 4. Spatial representation of monthly water stress index (WSI) for January, April, July and
353 October on the global scale without EFRs (upper figure) and considering EFRs (down figure)
354 averaged over 41 years aggregated on a river basin level. Units are in percentage of water withdrawn
355 per water available. Levels of WSI are defined as such: low in blue (between 0 and 20%), medium in
356 green (20% to 45%), high in yellow (45% to 80%) and scarce in red (above 80%).

357 Figure 3b shows the flow dynamics of the Columbia River representing group 2. The range of
358 anthropogenic flow was reduced compared to the natural flow and the actual flow is below the
359 natural flow from April to August. The absolute monthly deficit is about 5 km 3 with summer peaks
360 (in June and July). The frequency of the monthly deficit is less than 10 months over 41 for natural
361 deficit and anthropogenic deficit frequency is mainly occurring between August and September (up
362 to 24 months over 41). Natural deficit frequency is spread over six months and anthropogenic deficit
363 is spread over three months with a higher share of absolute natural deficit over anthropogenic deficit
364 (63%). Total deficit to withdrawals represents 32% of which 20% are impacting surface water only.
365 Group 3 is represented by the Ganges (Figure 3c) showing a high flow modification compared
366 with the actual flow and the absolute monthly deficit reaches up to 5 km3 in June (31-50% of annual
367 discharge). Total frequency deficit is nearly 100% between December and April where
368 anthropogenic deficit frequency represents more than 80% of the total deficit frequency. Total deficit
369 to withdrawals represents 62% of which 35% are impacting surface water only.
370 Finally, Group 4 is represented by the Guadalquivir River that shows a lower actual flow than
371 natural flow all year long (Figure 3d). Total monthly deficit can reach up to 20 km3 in June-July and
372 outpace natural and actual flows in summer months. Total deficit frequency occurs every month
373 with at least 20 months of deficit out of 41. Natural frequency deficit is between 10 and 30 months
374 out of 41 with peaks in late summer and anthropogenic deficit frequency is about 10 to 25 months
375 out of 41 occurring mainly between April and September. The Guadalquivir has a total deficit of
376 about 64% of the total discharge. Total deficit to withdrawals represents 66% of which 15% are
377 impacting surface water only.
378 Figure 4 presents water stress indicator (WSI) at global scale with and without the inclusions of
379 EFRs. Without EFRs, regions with a low WSI (value < 40%) encompass the Baltics, Latin America,
380 Europe, Sub-Saharan Africa, the Pacific Islands and Oceania. In January, mostly countries of Asia
381 and around the Equateur belt show a high WSI while in July, we show that mainly countries from
382 the MENA are affected by water stress both corresponding to the dry period and usually marked by
383 high irrigation requirements. When EFRs are considered, WSI is increased by more than 20% and
384 with the same spatial distribution of water stress than without EFRs included. While Latin America,
 Water 2019, 11, x FOR PEER REVIEW 13 of 19

385 North America, Oceania and SSA present low to moderate water stress, other regions of Asia, CIS
386 and MENA are seasonally under water scarcity (WSI > 80%).

387 4. Discussion
388 We provide a global analysis yielding classified patterns of EFR transgression (natural vs.
389 anthropogenic) which tentatively points to different ways of their possible management/avoidance
390 and update the global water stress indicator with the integration of EFRs. Concerning the EF deficit,
391 we identified how frequent the deficit occurs over a period of 41 years (inter-annual variability), the
392 timing, the duration of the deficit (intra-annual variability) and on which category of river basins the
393 deficit was more frequent. We used the VMF method to calculate EFRs at a monthly time-step and at
394 a spatial resolution of 0.5°; then aggregated on the river basin scale. This is the first study showing
395 differences between natural and anthropogenic monthly EF deficits including intra and inter-annual
396 variabilities. We showed that total EF deficit represents a low share of the total global discharge
397 (5%). However, at regional scale, we demonstrated that monthly deficit can reach significant shares
398 of the discharge or even exceed the available flow (usually during low flow periods) in locations
399 where high irrigation abstractions occur. Correlations were found between river flow regime and
400 level of flow alteration. For example, tropical perennial flow regimes such as the Amazon and the
401 Congo tend to have low EF deficits with a high proportion coming from natural deficit, while
402 intermittent flow regimes located in dry areas (and usually requiring high irrigation demand) show
403 high absolute deficit with a high proportion coming from anthropogenic deficit and occurs more
404 than 50% of the time period. Natural deficit can either follow a meteorological drought or can be
405 overestimated due to the non-consideration of non-renewable sources of water in this analysis. From
406 an ecological point of view, natural deficit can be beneficial or detrimental to freshwater ecosystems.
407 For example, some species have become resilient to periodic droughts (Wen and Saintilan 2015). In
408 some cases, natural deficit can even be required for the removal of invasive species (Bond, Lake, and
409 Arthington 2008; Lake 2003). Natural disturbances such as floods and droughts events can also
410 contribute to the regulation of population size and species diversity (Lytle and Poff 2004). The
411 recovery of freshwater species is linked to three kinds of adaptation mechanisms: life history,
412 behaviour and morphology which are used by species to deal with natural disturbances. However,
413 an excessive deficit with long duration could lead to species extinction due to low provision of
414 refugia for specie recovery and due to consequent abrupt changes in biological community structure
415 and ecosystem processes (Humphries and Baldwin 2003).
416 Anthropogenic deficit can lead to a shift in flow regime jeopardizing the natural integrity of
417 freshwater ecosystems (Bunn and Arthington 2002). Anthropogenic deficit can also exacerbate the
418 effect of drought and water deficit on freshwater ecosystems by “reducing the resistance and the
419 resilience of species” (Bond, Lake, and Arthington 2008). For example, important components of the
420 flow of the river of the Murray-Darling such as low flow duration were shown to be heavily
421 modified after water abstraction (Wen and Saintilan 2015; Maheshwari, Walker, and McMahon
422 1995). In addition, increase in low-flow duration can decrease riparian vegetation and can lead to
423 physiological stress (Poff et al. 1997). Furthermore, dams and reservoirs can harm freshwater
424 ecosystems via the disruption of the existing movement of water and sediments that provide food
425 and refugia to most of river taxa (Bunn and Arthington 2002).
426 Our results showed that anthropogenic and natural deficits are usually combined and seldom
427 independent because water demands for agriculture (e.g. evapotranspiration and irrigation) are
428 usually higher during and after a natural deficit is occurring. Only in the case of perennial tropical
429 rivers with high discharge such as the Amazon and the Congo, natural deficit accounted for most of
430 the deficit. Natural deficit is usually linked with hydrological droughts which are responses to
431 meteorological droughts. According to Keyantash and Dracup (2002), the response of a hydrological
432 drought (and natural deficit) depends on the location of the climatic drought and on how long the
433 flow requires to reach downstream parts. For example, in the Amazon River, we found that drought
434 years and natural deficit are occurring on the same years (1973, 1982 and 1991) linked to El Niño
435 events (Sheffield and Wood 2012). Further, meteorological droughts leading to higher
 Water 2019, 11, x FOR PEER REVIEW 14 of 19

436 evapotranspiration and irrigation requirements can exacerbate the anthropogenic deficit (Gómez
437 and Blanco 2012).
438 In this study, we examined the cause of deficit and we defined the timing, volume, duration
439 and frequency of the deficit for 23 river basins. The concept of environmental flow deficit is a
440 forward step in the eco-hydrology field providing key hydrological tools to water managers and
441 freshwater ecologists on when, where and why freshwater ecosystems are at risk. Xenopoulos et al
442 (2009) have shown that flow alteration could be linked with the loss of freshwater species richness.
443 Similarly, (Pracheil, McIntyre, and Lyons 2013) managed to define discharge thresholds to maintain
444 most of specialist species of the Mississippi River. However, more research is needed on “how the
445 trait composition of stream communities varies along geographical and environmental gradients”
446 (Heino, Schmera, and Erős 2013). Using seasonal and supra-seasonal EF deficit definitions could be
447 helpful for the monitoring of rivers and for the adoption of the most appropriate solution. We used
448 the VMF method because it is a global EF method with high spatial and temporal resolutions that
449 was validated with local study cases (Pastor et al. 2014). Moreover, it was largely implemented and
450 accepted in the latest global assessments (Boulay et al. 2015; D. Gerten et al. 2013; Sadoff 2015; Grill et
451 al. 2015; Steffen et al. 2015). It is also possible to give a range of EFRs to water managers with its the
452 standard deviation (this study and in (Shadkam et al. 2016). Other studies providing sensitivity
453 analyses of EF methods show small differences between methods on water resources and food
454 production and a lower impact on these latter than with drivers such as climate change and/or
455 increase in population (Jägermeyr et al. 2017; Bonsch et al. 2015; Steffen et al. 2015). EFRs can also be
456 calculated for wet and dry years such as in (Richter et al. 2006) but this requires climate and river
457 monitoring. If ecological data on the related rivers is collected, EFRs can also be defined according to
458 the level of flow alteration and to the desired ecological outcome of the river basin community (Pyne
459 and Poff 2016; M. Acreman et al. 2014).
460 In this study, regions with high level of water stress were characterized in term of percentage of
461 water abstraction-to-availability in time and space. Hot-spots of the MENA and South-Asian regions
462 were found to be always under high level of water stress in magnitude and frequency such as in
463 Kummu et al. (2016). They show that global water consumption and stress have increased four times
464 since 1900’s. Therefore, solutions to restore and/or protect rivers should be targeted to specific group
465 of rivers and their related type of pressure and type of EF deficit. For example, Mediterranean areas
466 are subject to both natural and anthropogenic deficits due to long dry seasons combined with high
467 irrigation abstractions. Tackling these two different types of deficits might require different
468 measures, e.g. for natural deficits, it might be necessary to increase water storage (Gaupp, Hall, and
469 Dadson 2015) and for anthropogenic deficits to decrease the production of water intensive crops
470 (Aldaya, Martínez-Santos, and Llamas 2010). The type and the level of intervention can be informed
471 by EFR assessments which aim at avoiding ecosystem collapse. In this study, we show that surface
472 water withdrawals for irrigation are responsible for 16-34 % of the anthropogenic deficit and this is
473 also shown in Jägermeyr et al. (2017). If up to 14% of the irrigated production were lost, food security
474 would be at risk in the most affected regions and potentially also globally, unless measures and
475 policy levies are used. Similarly, other studies show that reducing water use for irrigation to
476 preserve ecosystems may lead to water conflicts between users (Kingsford and Auld 2005; Poff et al.
477 2003; Richter et al. 2006). A specific example is shown in (Blanco-Gutiérrez, Varela-Ortega, and
478 Purkey 2013) where respecting EFRs for Water Framework Directive in the Guadiana River would
479 cut consequent water use for irrigation and would affect rice growers. Another study shows the
480 importance to estimate the deficit according to its flow regime and flow alteration such as in the
481 example of the Segura River. This latter is characterized by an intermittent river with high water
482 abstractions where four levels of adaptation to drought management were defined: normal,
483 pre-alert, alert and emergency of which the resulting irrigation withdrawals from surface water
484 were respectively reduced by 100%, 90%, 75% to 50%. This kind of drought management was shown
485 to be successful when groundwater rights were defined (Gómez and Blanco 2012). Finally, payments
486 for river ecosystem services such as natural purification of wastewater, erosion control, habitat for
 Water 2019, 11, x FOR PEER REVIEW 15 of 19

487 fish and wildlife, and recreation was a good incentive to limit water stress in some communities
488 (Loomis et al. 2000).
489 To protect rivers, it is necessary to improve environmental river monitoring and anticipate
490 responses to flow changes by choosing appropriate methods (Palmer et al. 2009). For example, our
491 results show that irrigation abstractions might need to be reduced to meet EFRs especially in
492 modified intermittent rivers and that specific flow releases from storage could be applied to meet
493 EFRs and preserve ecosystems (Gaupp, Hall, and Dadson 2015). Therefore, when river basin
494 communities decide to implement EFRs, trade-offs between water users can be informed by
495 assessments like ours. Additionally, decision-support models and hydro-economic models could be
496 helpful in meeting both irrigation water requirements and EFRs (Acutis et al. 2009; Blanco-Gutiérrez,
497 Varela-Ortega, and Purkey 2013; Scott et al. 2014).
498 Solutions to save water locally and other water demand and supply side measures should be
499 included in future studies and assessments such as increasing green water storage, purifying grey
500 water, reducing export of virtual water, increasing water use efficiency and by desalinating brackish
501 water (Lal 2015). Another option to decrease local and regional deficits in water-scarce regions
502 would be by decreasing irrigation via increasing food imports from water-abundant regions and to
503 use more dam regulation (Dalin and Rodríguez-Iturbe 2016). Attention should also be paid in
504 ensuring return flows to recharge groundwater and downstream rivers (Scott et al. 2014; Ward and
505 Pulido-Velazquez 2008).
506 Future research could also address the effects of climate change, land-use change and future
507 socio-economic scenarios to evaluate the intensity, duration, frequency and spatial distribution of
508 deficits under global change. In fact, it was shown that future climate is likely to be more variable
509 due to intensification of the hydrological cycle. This latter might exacerbate pressure on water
510 provision for human and ecosystem (Davis et al. 2015; Palmer et al. 2009). For example, a study from
511 (Rajagopalan et al. 2009) showed that a 20% reduction of the Colorado River flow would imply a
512 tenfold reduction in reservoir storage. Including other parameters than flow alteration in
513 eco-hydrological assessments could also be beneficial such as the inclusion of changes in water
514 temperature or water quality (M. T. H. van Vliet et al. 2013). In fact, future water temperature were
515 shown to increase the most in US, Europe and China and will likely be exacerbated by decreased
516 summer low flows in most regions (M. H. Vliet, Ludwig, and Kabat 2013).

517 5. Conclusions
518 This study presents a new concept of water scarcity, which include environmental flow
519 requirements. We evaluated water deficit for the environment (EF deficit) in terms of duration,
520 timing and frequency for 23 major river basins in the world. EF deficit was shown to be rather a
521 regional than a global concern (globally EF deficits only represented 5% of discharge). The study
522 highlights the hot-spots regions, such as in South-Asia, in the Mediterranean area and in the West
523 coast of US, where immediate action needs to be taken to restoire stream flows and avoid further
524 degradation of riverine ecosystems In this study we also identified the causes of EF deficit (natural
525 and anthropogenic) which allows being more specific in the choice and the level of intervention of
526 river restoration. For example, a river with a high anthropogenic deficit such as the Indus is likely to
527 require a water allocation agreement between different water users. We found correlations between
528 the cause of deficit (natural or anthropogenic), the level of flow modification (in magnitude) and the
529 type of flow regime (perennial or intermittent). Intermittent rivers with moderate to high flow
530 alterations are likely to encounter anthropogenic deficit. We show that specific targeted actions
531 might be required in heavily modified rivers where EF deficit exceeds the available flow. A warning
532 should also be put on pristine rivers and river basins with actual low EF deficit to be cautious with
533 future water allocation and avoid stream degradation. Finally, the inclusion of EFRs show a higher
534 water stress index globally varying between 30% in temperate zones to maximum WSI values in the
535 MENA and in South-East Asia.
536 Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1, Figure S1: Conceptual
537 framework to calculate EFR deficit on the global scale, Figure S2: Localisation of river basin groups after PCA
 Water 2019, 11, x FOR PEER REVIEW 16 of 19

538 analysis from the least modified and stable flow to the most modified and variable flow, Figure S3: Spatial
539 representation of monthly deficit on the global scale for 12 months. Units are in km 3 month-1, Figure S4:
540 Temporal representation of natural and actual flow (left plot) with intra-annual monthly deficit (km3month-1)
541 and frequency of deficit (number of months) for four river basins representing each group of river basins.
542 Barplots represent the frequency of natural deficit (black) and anthropogenic deficit (grey), Figure S5: Spatial
543 representation aggregated on a country level of monthly water stress index (WSI) for January, April, July and
544 October on the global scale without EFRs (upper figure) and considering EFRs (down figure) averaged over 41
545 years. Units are in percentage of water withdrawn per water available. Levels of WSI are defined as such: low in
546 blue (between 0 and 20%), medium in green (20% to 45%), high in yellow (45% to 80%) and scarce in red (above
547 80%), Figure S6: Boxplot of global WSI indexes with and without EFRs in percentage (%).

548 Acknowledgments: This research was funded by Netherlands Organization for scientific research (NWO).

549 References
550 1. Acreman, M, Angela H Arthington, Matthew J Colloff, Carol Couch, Neville D Crossman, Fiona Dyer, Ian
551 Overton, Carmel A Pollino, Michael J Stewardson, and William Young. 2014. “Environmental Flows for
552 Natural, Hybrid, and Novel Riverine Ecosystems in a Changing World TT - Frontiers in Ecology and the
553 Environment.” Frontiers in Ecology and the Environment 12: 466–73. https://doi.org/10.1890/130134.
554 2. Acreman, M C, and A J D Ferguson. 2009. “Environmental Flows and the European Water Framework
555 Directive.” Freshw Biol 55: 32–48.
556 3. Acutis, M, M Rinaldi, F Mattia, and A Perego. 2009. “Integration of a Crop Simulation Model and Remote
557 Sensing Information.” In Crop Modeling and Decision Support, 307–16. Springer.
558 4. Aldaya, Maite M, Pedro Martínez-Santos, and M Ramón Llamas. 2010. “Incorporating the Water Footprint
559 and Virtual Water into Policy: Reflections from the Mancha Occidental Region, Spain.” Water Resources
560 Management 24: 941–58.
561 5. Arthington, A H, S E Bunn, N L Poff, and R J Naiman. 2006. “The Challenge of Providing Environmental
562 Flow Rules to Sustain River Ecosystems.” Ecological Applications 16: 1311–18.
563 https://doi.org/doi:10.1890/1051-0761(2006)016[1311:TCOPEF]2.0.CO;2.
564 6. Arthington, Angela H, Anik Bhaduri, Stuart E Bunn, Sue E Jackson, Rebecca E Tharme, Dave Tickner, Bill
565 Young, et al. 2018. “The Brisbane Declaration and Global Action Agenda on Environmental Flows (2018).”
566 Frontiers in Environmental Science 6: 45. https://doi.org/10.3389/fenvs.2018.00045.
567 7. Beck, Hylke E, Albert I J M van Dijk, Diego G Miralles, Richard A M de Jeu, L A Sampurno Bruijnzeel, Tim
568 R McVicar, and Jaap Schellekens. 2013. “Global Patterns in Base Flow Index and Recession Based on
569 Streamflow Observations from 3394 Catchments.” Water Resources Research 49 (12): 7843–63.
570 8. Biemans, H, I Haddeland, P Kabat, F Ludwig, R W A Hutjes, J Heinke, W von Bloh, and D Gerten. 2011.
571 “Impact of Reservoirs on River Discharge and Irrigation Water Supply during the 20th Century.” Water
572 Resources Research 47: W03509.
573 9. Blanco-Gutiérrez, Irene, Consuelo Varela-Ortega, and David R Purkey. 2013. “Integrated Assessment of
574 Policy Interventions for Promoting Sustainable Irrigation in Semi-Arid Environments: A Hydro-Economic
575 Modeling Approach.” Journal of Environmental Management 128: 144–60.
576 10. Bond, Nicholas R, P S Lake, and Angela H Arthington. 2008. “The Impacts of Drought on Freshwater
577 Ecosystems: An Australian Perspective.” Hydrobiologia 600: 3–16.
578 11. Bondeau, Alberte, Pascalle C Smith, Sonke Zaehle, Sibyll Schaphoff, Wolfgang Lucht, Wolfgang Cramer,
579 Dieter Gerten, Hermann Lotze Campen, Christoph Müller, and Markus Reichstein. 2007. “Modelling the
580 Role of Agriculture for the 20th Century Global Terrestrial Carbon Balance.” Global Change Biology 13:
581 679–706.
582 12. Bonsch, Markus, Alexander Popp, Anne Biewald, Susanne Rolinski, Christoph Schmitz, Isabelle Weindl,
583 Miodrag Stevanovic, Kathrin Högner, Jens Heinke, and Sebastian Ostberg. 2015. “Environmental Flow
584 Provision: Implications for Agricultural Water and Land-Use at the Global Scale.” Global Environmental
585 Change 30: 113–32.
586 13. Botter, Gianluca, Stefano Basso, Ignacio Rodriguez-Iturbe, and Andrea Rinaldo. 2013. “Resilience of River
587 Flow Regimes.” Proceedings of the National Academy of Sciences 110: 12925–30.
588 14. Boulay, A.-M., J. Bare, C. De Camillis, P. Döll, F. Gassert, D. Gerten, S. Humbert, et al. 2015. “Consensus
589 Building on the Development of a Stress-Based Indicator for LCA-Based Impact Assessment of Water
 Water 2019, 11, x FOR PEER REVIEW 17 of 19

590 Consumption: Outcome of the Expert Workshops.” International Journal of Life Cycle Assessment 20 (5).
591 https://doi.org/10.1007/s11367-015-0869-8.
592 15. Bunn, S E, and A H Arthington. 2002. “Basic Principles and Ecological Consequences of Altered Flow
593 Regimes for Aquatic Biodiversity.” Environmental Management 30: 492–507.
594 https://doi.org/doi:0.1007/s00267-002-2737-0.
595 16. Dalin, Carole, and Ignacio Rodríguez-Iturbe. 2016. “Environmental Impacts of Food Trade via Resource
596 Use and Greenhouse Gas Emissions.” Environmental Research Letters 11 (3): 35012.
597 17. Davis, Jenny, Anthony P O’Grady, Allan Dale, Angela H Arthington, Peter A Gell, Patrick D Driver, Nick
598 Bond, et al. 2015. “When Trends Intersect: The Challenge of Protecting Freshwater Ecosystems under
599 Multiple Land Use and Hydrological Intensification Scenarios.” Science of The Total Environment 534: 65–
600 78. https://doi.org/http://dx.doi.org/10.1016/j.scitotenv.2015.03.127.
601 18. Declaration, The Brisbane. 2007. “The Brisbane Declaration: Environmental Flows Are Essential for
602 Freshwater Ecosystem Health and Human Well-Being.” Declaration of the 10th International River
603 Symposium. 3–6 September 2007, Brisbane, Australia.
604 19. Fader, Marianela, Stefanie Rost, Christoph Müller, Alberte Bondeau, and Dieter Gerten. 2010. “Virtual
605 Water Content of Temperate Cereals and Maize: Present and Potential Future Patterns.” Journal of
606 Hydrology 384: 218–31.
607 20. Ferreira, Teresa, Lidija Globevnik, and Rafaela Schinegger. 2019. “Chapter 8 - Water Stressors in Europe:
608 New Threats in the Old World.” In Multiple Stressors in River Ecosystems, edited by Sergi Sabater, Arturo
609 Elosegi, and Ralf Ludwig, 139–55. Elsevier.
610 https://doi.org/https://doi.org/10.1016/B978-0-12-811713-2.00008-X.
611 21. Flörke, Martina, Ellen Kynast, Ilona Bärlund, Stephanie Eisner, Florian Wimmer, and Joseph Alcamo.
612 2013. “Domestic and Industrial Water Uses of the Past 60 Years as a Mirror of Socio-Economic
613 Development: A Global Simulation Study.” Global Environmental Change 23: 144–56.
614 22. Gaupp, Franziska, Jim Hall, and Simon Dadson. 2015. “The Role of Storage Capacity in Coping with
615 Intra-and Inter-Annual Water Variability in Large River Basins.” Environmental Research Letters 10:
616 125001.
617 23. Gerten, D., H. Hoff, J. Rockström, J. Jägermeyr, M. Kummu, and A.V. Pastor. 2013. “Towards a Revised
618 Planetary Boundary for Consumptive Freshwater Use: Role of Environmental Flow Requirements.”
619 Current Opinion in Environmental Sustainability 5 (6). https://doi.org/10.1016/j.cosust.2013.11.001.
620 24. Gerten, Dieter, Sibyll Schaphoff, Uwe Haberlandt, Wolfgang Lucht, and Stephen Sitch. 2004. “Terrestrial
621 Vegetation and Water Balance—hydrological Evaluation of a Dynamic Global Vegetation Model.” Journal
622 of Hydrology 286: 249–70.
623 25. Gleick, Peter H. 2003. “Global Freshwater Resources: Soft-Path Solutions for the 21st Century.” Science
624 302: 1524–28.
625 26. Gómez, Carlos Mario Gómez, and Carlos Dionisio Pérez Blanco. 2012. “Do Drought Management Plans
626 Reduce Drought Risk? A Risk Assessment Model for a Mediterranean River Basin.” Ecological Economics
627 76: 42–48.
628 27. Gordon, N D, Thomas A McMahon, and Brian L Finlayson. 2004. Stream Hydrology: An Introduction for
629 Ecologists. John Wiley and Sons.
630 28. Grafton, R Quentin, Jamie Pittock, Richard Davis, John Williams, Guobin Fu, Michele Warburton, Bradley
631 Udall, et al. 2013. “Global Insights into Water Resources, Climate Change and Governance.” Nature
632 Climate Change 3: 315–21. http://dx.doi.org/10.1038/nclimate1746.
633 29. Grill, Günther, Bernhard Lehner, Alexander E Lumsdon, Graham K MacDonald, Christiane Zarfl, and
634 Catherine Reidy Liermann. 2015. “An Index-Based Framework for Assessing Patterns and Trends in River
635 Fragmentation and Flow Regulation by Global Dams at Multiple Scales.” Environmental Research Letters
636 10: 15001.
637 30. Heino, Jani, Dénes Schmera, and Tibor Erős. 2013. “A Macroecological Perspective of Trait Patterns in
638 Stream Communities.” Freshwater Biology 58: 1539–55.
639 31. Humphries, Paul, and Darren S Baldwin. 2003. “Drought and Aquatic Ecosystems: An Introduction.”
640 Freshwater Biology 48: 1141–46.
641 32. Jägermeyr, J., A. Pastor, H. Biemans, and D. Gerten. 2017. “Reconciling Irrigated Food Production with
642 Environmental Flows for Sustainable Development Goals Implementation.” Nature Communications 8.
643 https://doi.org/10.1038/ncomms15900.
 Water 2019, 11, x FOR PEER REVIEW 18 of 19

644 33. Keyantash, John, and John A Dracup. 2002. “The Quantification of Drought: An Evaluation of Drought
645 Indices.” Bulletin of the American Meteorological Society 83: 1167.
646 34. Kingsford, R T, and K M Auld. 2005. “Waterbird Breeding and Environmental Flow Management in the
647 Macquarie Marshes, Arid Australia.” River Research and Applications 21: 187–200.
648 https://doi.org/doi:10.1002/Rra.840.
649 35. Kummu, Matti, J H A Guillaume, Hans de Moel, S Eisner, Martina Flörke, Miina Porkka, Stefan Siebert,
650 Ted I E Veldkamp, and Philip J Ward. 2016. “The World’s Road to Water Scarcity: Shortage and Stress in
651 the 20th Century and Pathways towards Sustainability.” Scientific Reports 6: 38495.
652 36. Lake, P S. 2003. “Ecological Effects of Perturbation by Drought in Flowing Waters.” Freshwater Biology 48:
653 1161–72.
654 37. Lal, Rattan. 2015. “World Water Resources and Achieving Water Security.” Agronomy Journal 107: 1526–
655 32.
656 38. Loomis, John, Paula Kent, Liz Strange, Kurt Fausch, and Alan Covich. 2000. “Measuring the Total
657 Economic Value of Restoring Ecosystem Services in an Impaired River Basin: Results from a Contingent
658 Valuation Survey.” Ecological Economics 33: 103–17.
659 39. Lytle, David A, and N LeRoy Poff. 2004. “Adaptation to Natural Flow Regimes.” Trends in Ecology &
660 Evolution 19: 94–100.
661 40. Maheshwari, B L, K F Walker, and T A McMahon. 1995. “Effects of Regulation on the Flow Regime of the
662 River Murray, Australia.” Regulated Rivers: Research & Management 10: 15–38.
663 41. Palmer, Margaret A, Dennis P Lettenmaier, N LeRoy Poff, Sandra L Postel, Brian Richter, and Richard
664 Warner. 2009. “Climate Change and River Ecosystems: Protection and Adaptation Options.”
665 Environmental Management 44: 1053–68.
666 42. Pastor, A.V., F. Ludwig, H. Biemans, H. Hoff, and P. Kabat. 2014. “Accounting for Environmental Flow
667 Requirements in Global Water Assessments.” Hydrology and Earth System Sciences 18 (12).
668 https://doi.org/10.5194/hess-18-5041-2014.
669 43. Poff, N LeRoy, J David Allan, Mark B Bain, James R Karr, Karen L Prestegaard, Brian D Richter, Richard E
670 Sparks, and Julie C Stromberg. 1997. “The Natural Flow Regime.” BioScience 47: 769–84.
671 44. Poff, N LeRoy, J David Allan, Margaret A Palmer, David D Hart, Brian D Richter, Angela H Arthington,
672 Kevin H Rogers, Judy L Meyer, and Jack A Stanford. 2003. “River Flows and Water Wars: Emerging
673 Science for Environmental Decision Making TT - Frontiers in Ecology and the Environment.” Frontiers in
674 Ecology and the Environment 1: 298–306.
675 https://doi.org/10.1890/1540-9295(2003)001[0298:rfawwe]2.0.co;2.
676 45. Poff, N LeRoy, and Julie K H Zimmerman. 2010. “Ecological Responses to Altered Flow Regimes: A
677 Literature Review to Inform the Science and Management of Environmental Flows.” Freshwater Biology
678 55: 194–205.
679 46. Pracheil, Brenda M, Peter B McIntyre, and John D Lyons. 2013. “Enhancing Conservation of Large‐river
680 Biodiversity by Accounting for Tributaries.” Frontiers in Ecology and the Environment 11: 124–28.
681 47. Pyne, Matthew I, and N LeRoy Poff. 2016. “Vulnerability of Stream Community Composition and
682 Function to Projected Thermal Warming and Hydrologic Change across Ecoregions in the Western United
683 States.” Global Change Biology.
684 48. Rajagopalan, Balaji, Kenneth Nowak, James Prairie, Martin Hoerling, Benjamin Harding, Joseph Barsugli,
685 Andrea Ray, and Bradley Udall. 2009. “Water Supply Risk on the Colorado River: Can Management
686 Mitigate?” Water Resources Research 45.
687 49. Richter, B D, A T Warner, J L Meyer, and K Lutz. 2006. “A Collaborative and Adaptive Process for
688 Developing Environmental Flow Recommendations.” River Research and Applications 22: 297–318.
689 50. Rost, Stefanie, Dieter Gerten, Alberte Bondeau, Wolfgang Lucht, Janine Rohwer, and Sibyll Schaphoff.
690 2008. “Agricultural Green and Blue Water Consumption and Its Influence on the Global Water System.”
691 Water Resources Research 44: W09405. https://doi.org/10.1029/2007wr006331.
692 51. Sadoff, Claudia W. 2015. Securing Water, Sustaining Growth: Report of the GWP/OECD Task Force on
693 Water Security and Sustainable Growth.
694 52. Schewe, Jacob, Jens Heinke, Dieter Gerten, Ingjerd Haddeland, Nigel W Arnell, Douglas B Clark, Rutger
695 Dankers, Stephanie Eisner, Balázs M Fekete, and Felipe J Colón-González. 2014. “Multimodel Assessment
696 of Water Scarcity under Climate Change.” Proceedings of the National Academy of Sciences 111: 3245–50.
 Water 2019, 11, x FOR PEER REVIEW 19 of 19

697 53. Scott, Christopher A, S Vicuña, Irene Blanco Gutiérrez, F Meza, and Consuelo Varela Ortega. 2014.
698 “Irrigation Efficiency and Water-Policy Implications for River-Basin Resilience.” Hydrology and Earth
699 System Sciences 18: 1339–48.
700 54. Shadkam, S., F. Ludwig, M.T.H. van Vliet, A. Pastor, and P. Kabat. 2016. “Preserving the World Second
701 Largest Hypersaline Lake under Future Irrigation and Climate Change.” Science of the Total Environment
702 559. https://doi.org/10.1016/j.scitotenv.2016.03.190.
703 55. Sheffield, Justin, and Eric F Wood. 2012. Drought: Past Problems and Future Scenarios. Routledge.
704 56. Siebert, Stefan, Jacob Burke, Jean-Marc Faures, Karen Frenken, Jippe Hoogeveen, Petra Döll, and Felix
705 Theodor Portmann. 2010. “Groundwater Use for Irrigation–a Global Inventory.” Hydrology and Earth
706 System Sciences 14: 1863–80.
707 57. Sitch, S, Benjamin Smith, I Colin Prentice, Almut Arneth, A Bondeau, W Cramer, J O Kaplan, S Levis, W
708 Lucht, and M Thonicke Sykes. 2003. “Evaluation of Ecosystem Dynamics, Plant Geography and Terrestrial
709 Carbon Cycling in the LPJ Dynamic Global Vegetation Model.” Global Change Biology 9: 161–85.
710 https://doi.org/doi: 10.1046/j.1365-2486.2003.00569.x.
711 58. Smakhtin, V, C Revenga, and P Doll. 2004. “A Pilot Global Assessment of Environmental Water
712 Requirements and Scarcity.” Water International 29: 307–17.
713 59. Steffen, Will, Katherine Richardson, Johan Rockström, Sarah E Cornell, Ingo Fetzer, Elena M Bennett,
714 Reinette Biggs, Stephen R Carpenter, Wim de Vries, and Cynthia A de Wit. 2015. “Planetary Boundaries:
715 Guiding Human Development on a Changing Planet.” Science 347: 1259855.
716 60. Vliet, Michelle T H van, Wietse H P Franssen, John R Yearsley, Fulco Ludwig, Ingjerd Haddeland, Dennis
717 P Lettenmaier, and Pavel Kabat. 2013. “Global River Discharge and Water Temperature under Climate
718 Change.” Global Environmental Change 23: 450–64.
719 https://doi.org/http://dx.doi.org/10.1016/j.gloenvcha.2012.11.002.
720 61. Vliet, MichelleT H, Fulco Ludwig, and Pavel Kabat. 2013. “Global Streamflow and Thermal Habitats of
721 Freshwater Fishes under Climate Change TT - Climatic Change.” Climatic Change 121: 739–54.
722 https://doi.org/10.1007/s10584-013-0976-0.
723 62. Vorosmarty, C J, P B McIntyre, M O Gessner, D Dudgeon, A Prusevich, P Green, S Glidden, et al. 2010.
724 “Global Threats to Human Water Security and River Biodiversity .” Nature 468: 334.
725 https://doi.org/doi:10.1038/Nature09549.
726 63. Wada, Yoshihide, and Marc F P Bierkens. 2014. “Sustainability of Global Water Use: Past Reconstruction
727 and Future Projections.” Environmental Research Letters 9: 104003.
728 64. Ward, Frank A, and Manuel Pulido-Velazquez. 2008. “Water Conservation in Irrigation Can Increase
729 Water Use.” Proceedings of the National Academy of Sciences 105: 18215–20.
730 65. Wen, Li, and Neil Saintilan. 2015. “Climate Phase Drives Canopy Condition in a Large Semi-Arid
731 Floodplain Forest.” Journal of Environmental Management 159: 279–87.
732 66. Wold, Svante, Kim Esbensen, and Paul Geladi. 1987. “Principal Component Analysis.” Chemometrics and
733 Intelligent Laboratory Systems 2: 37–52. https://doi.org/http://dx.doi.org/10.1016/0169-7439(87)80084-9.
734 67. WWF/ZSL. 2016. “Living Planet Index 2016.” Living Planet Report 201. WWF.

© 2019 by the authors. Submitted for possible open access publication under the terms
and conditions of the Creative Commons Attribution (CC BY) license
(http://creativecommons.org/licenses/by/4.0/).
735