Despite the increasing influence of human activities on water resources in our current Anthropocene era, the impacts of these activities on the duration, rate and timing of the recovery of drought events, known as the drought termination phase, remain unknown. Here, we present the first assessment of how different human activities (i.e. water abstractions, reservoirs, water transfers) affect drought termination. Six case studies in Europe were used to analyse the human influence on streamflow drought termination characteristics. For all case studies, we compared the drought and drought termination characteristics derived from a human-influenced time series of streamflow (observation data) and a naturalised time series (modelled data) for the same period. Overall, results clearly demonstrate the influence of human activities on drought terminations in all the studied catchments. Groundwater abstractions, reservoirs and mixed influences were all found to increase the average duration of drought termination, whereas water transfers into the catchment decreased drought termination duration. Results also show that average drought termination rates increased in all case studies due to the human influence. Furthermore, start and end months of the termination phase were more skewed to certain months in human-influenced data than in the naturalised situation. Future research could extend this new knowledge by looking to add further case studies and covering different human activities to gain a wider understanding on how human actions modify hydrological droughts and their recovery. Furthering this work could also help to improve the forecasting of drought recovery in the Anthropocene, which is important for informing drought management decisions.

## 1. Introduction

Many regions of the world will likely see an increase in drought occurrence and severity in the 21st century (Dai, 2011; Kirono et al., 2011; Wanders et al., 2015; Samaniego et al., 2018), yet there are numerous areas in which our understanding of drought processes is far from complete. This hinders our ability to forecast, manage and respond to water deficit anomalies, especially in relation to anthropogenic activities and feedbacks within drought (Van Loon et al., 2016). Arguably, a critical yet neglected phase of drought is the re-wetting stage, termed drought termination (Parry et al., 2016a). Drought termination is a characteristic of a drought event that describes its ending. It is not only a point in time denoting when a drought is said to have ended, but a quantifiable event with a temporal profile (see Figure 3 in Parry et al., 2016a). An understanding of why, when and how a drought is likely to terminate is valuable information for water managers. Such knowledge is crucial for deciding how the transition from depleted to replenished water supplies is operationally handled (Hannaford et al., 2011; Bell et al., 2013). Heim and Brewer (2012) stress the importance of including drought termination within any monitoring framework. As drought terminations are often disruptive, abrupt events (Dettinger, 2013), their study is also merited to help predict associated impacts of high flows and implications for water quality (Loecke et al., 2017). Yet the propagation, drivers, physical processes and feedbacks of drought termination through the hydrological cycle are currently poorly understood (Parry et al., 2016a).

One of the earliest studies on the topic used the Palmer Drought Severity Index (PDSI; Palmer, 1965) to calculate the amount of rainfall required for drought termination over different timeframes (Karl et al., 1987). Since then, the question of the amount of rainfall required to terminate a drought has been explored using models or climate ensembles of rainfall forecasts to calculate the likelihood of termination under given meteorological inputs (Bell et al., 2013; Pan et al., 2013; Antofie et al., 2014; Parry et al., 2018). Whilst such studies are important, this area of research is still in its infancy. Research to-date often considers events in soil moisture or subsurface storage, whereas there is a need to address hydrological drought termination more holistically.

In addition to the natural drivers of drought termination (i.e. climate, catchment type, geology, etc.), there are numerous ways in which humans modify land-use, water partitioning and hydrological regimes, which is especially crucial for the termination of hydrological drought compared to meteorological and soil moisture drought. Human-induced changes are known to have a significant effect on the hydrological cycle (Barnett et al., 2008) and are expected to increase due to a growing population (Alcamo et al., 2003). Understanding the processes impacted by human activities, such as those in the drought termination phase, is fundamental for the effectiveness of drought forecasting, monitoring and early warning systems. This information is subsequently invaluable for the formulation of adaptive responses to drought conditions and managing water scarcity, especially given that it is often human-influenced systems that we are most reliant upon for our water supply (Barker et al., 2016).

Despite the increasing influence of human activities on water resources in our current Anthropocene era (Steffen et al., 2015; Hamilton, 2016), the impacts of these activities on drought termination remain unknown and unexplored. For the first time here we assess the impact of human activities on changes in drought termination characteristics within different catchments by comparing human-influenced (observed) data with naturalised (modelled) data. The study aim is to assess if and how human modifications affect the drought termination phase by considering the human activities of reservoirs, abstraction, urbanisation and water transfer. We use a framework that can be applied to other human activities, catchments, climates etc.

It is expected that human influences will have a detectable effect on drought termination metrics, as they are known to do on other drought characteristics (Van Loon and Van Lanen, 2013, 2015; Wada et al., 2013; Tijdeman et al., 2018). The type of response seen in the termination phase is predicted to be reliant on the type of human activity dominant in the catchment.

## 2. Methods

In this section, we describe the methods used for this analysis. We introduce the case studies from across Europe, the data, the drought analysis and drought termination characteristics, and the comparisons to calculate the human influence. This work focuses on streamflow drought, defined as a sustained period of below normal water availability in river discharge (Mishra and Singh, 2010). We focus on streamflow droughts influenced by human activities (human-modified droughts, Van Loon et al. (2016)).

### 2.1. Case studies and data

Six case studies in Europe were used to analyse the human influence on streamflow drought termination characteristics (Figure 1; Table 1; Appendix 1). These case studies were chosen based on the availability of an observed and a naturalised streamflow time series for the same discharge gauge within the catchment. All time series used have no missing data. All case studies compared the human-influenced time series of streamflow (observation data) and the naturalised time series (representing the ‘natural’ situation) for the same period. Compared to an approach in which post-disturbance time series are compared with pre-disturbance time series, a benefit of using naturalised data is that the same time period and input data (i.e. precipitation) are used for both the human and natural situations. Any differences seen in streamflow therefore should be due to human activity, given uncertainties in both observations and model data.

Figure 1

Study catchments. Location of the six study catchments across Europe and the dominant human activity for each. DOI: https://doi.org/10.1525/elementa.365.f1

Figure 1

Study catchments. Location of the six study catchments across Europe and the dominant human activity for each. DOI: https://doi.org/10.1525/elementa.365.f1

Table 1

Summaries of case study details, their direct anthropogenic influences and naturalisation methods. DOI: https://doi.org/10.1525/elementa.365.t1

CountryBasinArea (km2)Human ActivityClimate Class1SourceDataset periodNaturalisation Method

Czech Republic Svitata 419 Groundwater abstraction for drinking water/public water supply (PWS) Dfb Van Loon and Van Lanen (2015) 1975–1990 BILAN lumped conceptual rainfall-runoff model, calibrated on a pre-disturbed period
Czech Republic Bilina 1,071 Water transport to catchment from nearby basin Cfb Van Loon and Van Lanen (2015) 1961–1990 BILAN lumped conceptual rainfall-runoff model, calibrated on a pre-disturbed period
France Ariège 1,340 Hydropower dam/reservoir, water transfer2 Cfb Banque Hydro French database (20171990–2004 Naturalisation by decomposition
Spain Upper-Guadiana 16,480 Groundwater abstraction for irrigation and artificial drainage Csa, Csb, Bsk Van Loon and Van Lanen (20131980–2001 HBV hydrological model, calibrated on a pre-disturbed period
United Kingdom Thames 9,948 Mixed influence (reservoirs, abstraction for PWS, industry and agriculture, effluent returns1 and urbanisation) Cfb NRFA (CEH, 20191900–2017 Naturalisation by decomposition
United Kingdom Lee 1,036 Mixed influence (reservoirs, abstraction for PWS, industry and agriculture, effluent returns1 and urbanisation) Cfb NRFA (CEH, 20191903–2016 Naturalisation by decomposition
CountryBasinArea (km2)Human ActivityClimate Class1SourceDataset periodNaturalisation Method

Czech Republic Svitata 419 Groundwater abstraction for drinking water/public water supply (PWS) Dfb Van Loon and Van Lanen (2015) 1975–1990 BILAN lumped conceptual rainfall-runoff model, calibrated on a pre-disturbed period
Czech Republic Bilina 1,071 Water transport to catchment from nearby basin Cfb Van Loon and Van Lanen (2015) 1961–1990 BILAN lumped conceptual rainfall-runoff model, calibrated on a pre-disturbed period
France Ariège 1,340 Hydropower dam/reservoir, water transfer2 Cfb Banque Hydro French database (20171990–2004 Naturalisation by decomposition
Spain Upper-Guadiana 16,480 Groundwater abstraction for irrigation and artificial drainage Csa, Csb, Bsk Van Loon and Van Lanen (20131980–2001 HBV hydrological model, calibrated on a pre-disturbed period
United Kingdom Thames 9,948 Mixed influence (reservoirs, abstraction for PWS, industry and agriculture, effluent returns1 and urbanisation) Cfb NRFA (CEH, 20191900–2017 Naturalisation by decomposition
United Kingdom Lee 1,036 Mixed influence (reservoirs, abstraction for PWS, industry and agriculture, effluent returns1 and urbanisation) Cfb NRFA (CEH, 20191903–2016 Naturalisation by decomposition

1 Köppen climate classification used (Kotek et al., 2006).

2 Denotes the lesser of the two (or more) human influences where they are not significant in terms of dominance in the catchment.

Data was naturalised prior to acquisition using the naturalisation technique most suitable for each case study, determined by the type of data and local information available (see Appendix 1). Monthly data was used, therefore where required, specific discharge values were converted from a daily to a monthly time-step by calculating monthly sum values. Although data lose finer details at this resolution, using a monthly time-step has the advantage of negating the need for pooling, as minor droughts (<1 month) are removed (Fleig et al., 2006). During a leap year streamflow in the month of February was corrected to remove the effect of having an additional day of flow by multiplying by a factor of 28/29. This was applied to all naturalised and human-influenced time series.

### 2.2. Identifying drought events

The method chosen to define drought conditions is important as it can influence the results obtained (Fleig et al., 2006; Heudorfer and Stahl, 2016). Here, the threshold level method was used to identify drought events (Yevjevich, 1967; Hisdal et al., 2004). The threshold level method identifies drought events as periods when discharge is below a predefined threshold, calculated using multiple years of streamflow data at a certain percentile of the flow duration curve (Van Loon, 2015). We used the 80th percentile (Q80), frequently used for identifying drought (Hisdal and Tallaksen, 2000; Fleig et al., 2006). As all catchments used here have flow regime seasonality, the variable threshold level method was deemed most appropriate for use (Fleig et al., 2006) rather than a fixed threshold. The Q80 threshold for each catchment was calculated using the full length of the natural dataset, which was then used to identify drought events in both the natural and human-influenced data. This was important to effectively show the impact of human activity on hydrological droughts (Liu et al., 2016; Rangecroft et al., 2019) and their termination characteristics. Ideally, a time period of at least 30 years of data is used to establish the drought threshold (McKee et al., 1993). As it is often difficult to obtain appropriate datasets of this length, time series used here varied from 14 to 117 years in length, depending on the case study.

Drought event analysis produced several drought characteristics (frequency, timing, duration, deficit volume, drought maximum intensity) (Table 2 and Figure 2). We subsequently calculated drought termination characteristics.

Table 2

Summaries of drought and drought termination metrics used in this study and their calculations. DOI: https://doi.org/10.1525/elementa.365.t2

Drought CharacteristicDescription

Drought Events

Frequency (DfreqNumber of droughts identified in time series.
Timing Date of drought onset and drought termination for each event.
Duration (DdurNumber of months spent in drought for each event.
Deficit (DdefWater volume (mm) below Q80 lacking for each event.
Drought Maximum Intensity (MI) Most intense point for each drought event. Calculated as follows:
Drought Maximum Intensity = Max(Q(t) – Threshold(i)) if Q(t) < Threshold(i)
Q is the flow value for month t, Threshold is Q80 for the same month in the year and i is the timestep. The MI using the (variable) threshold level method is therefore the largest distance between Q and the Q80 during an event.
Drought Termination

Drought Termination Start (DTstartThe first month of the drought termination phase (DTstart) is the month where the maximum intensity (MI) is reached. DTstart is calculated for each event.
Drought Termination End (DTendThe last month of the drought termination phase (DTend) is the last month of the drought event (i.e. the last month when discharge <Q80 drought threshold). DTend is calculated for each event.
Drought Termination Duration (DTdurNumber of months encompassed between DTstart and DTend (inclusive) for each event.
Drought Termination Rate (DTrateMaximum intensity divided by drought termination duration (mm/month) for each event.
Drought CharacteristicDescription

Drought Events

Frequency (DfreqNumber of droughts identified in time series.
Timing Date of drought onset and drought termination for each event.
Duration (DdurNumber of months spent in drought for each event.
Deficit (DdefWater volume (mm) below Q80 lacking for each event.
Drought Maximum Intensity (MI) Most intense point for each drought event. Calculated as follows:
Drought Maximum Intensity = Max(Q(t) – Threshold(i)) if Q(t) < Threshold(i)
Q is the flow value for month t, Threshold is Q80 for the same month in the year and i is the timestep. The MI using the (variable) threshold level method is therefore the largest distance between Q and the Q80 during an event.
Drought Termination

Drought Termination Start (DTstartThe first month of the drought termination phase (DTstart) is the month where the maximum intensity (MI) is reached. DTstart is calculated for each event.
Drought Termination End (DTendThe last month of the drought termination phase (DTend) is the last month of the drought event (i.e. the last month when discharge <Q80 drought threshold). DTend is calculated for each event.
Drought Termination Duration (DTdurNumber of months encompassed between DTstart and DTend (inclusive) for each event.
Drought Termination Rate (DTrateMaximum intensity divided by drought termination duration (mm/month) for each event.
Figure 2

Definition of drought termination. Conceptual diagram of drought termination characteristics and definitions used. DOI: https://doi.org/10.1525/elementa.365.f2

Figure 2

Definition of drought termination. Conceptual diagram of drought termination characteristics and definitions used. DOI: https://doi.org/10.1525/elementa.365.f2

### 2.3. Drought termination

Drought termination can be characterised by its duration, rate of recovery, and seasonality (Nkemdirim and Weber, 1999; Mo, 2011; Parry et al., 2016a). Our definitions for the drought termination phase for each drought event are described in Table 2. Figure 2 shows a conceptual diagram of how the termination phase was defined and the metrics calculated for each drought event. Drought events were divided at the drought maximum intensity (MI) (Bravar and Kavvas, 1991). The first month of the drought termination phase (DTstart) is therefore the month where the MI is reached (Figure 2). A drought ends when discharge exceeds the Q80 threshold, and the last month of the drought termination phase (DTend) is the last month of the drought event. A drought event and its termination phase therefore end at the same point in time. The drought termination duration (DTdur) is the number of months encompassed between DTstart and DTend (inclusive). The drought termination rate (DTrate) is the rate at which the system changes from being in a state of most intense drought state to non-drought conditions.

### 2.4. Quantifying the human influence

#### 2.4.1. Human influence on drought, drought termination duration and drought termination rate

Drought and drought termination characteristics per event were averaged over all events in the time period for each case study. Although it would be interesting to look at individual events, droughts identified in human-influenced and naturalised time series could not be matched and compared like-for-like, therefore comparisons were achieved based on averages. Results were compared between the drought or drought termination characteristic of the naturalised data (Xnat) and the same characteristic of the human-influenced data (Xhum) to quantify the size and direction of change in characteristics using Equation 1. Percentages reported are the change in the human-influenced conditions relative to the naturalised situation.

$\%\ {\rm{ change\ due\ to\ human\ influence }} = \left[ {\left({{{\rm{X}}_{\rm hum}}-{{\rm{X}}_{\rm nat}}} \right)} \right]/\,{{\rm X}_{{\rm nat}}}*\,\,100$
1

#### 2.4.2. Human influence on timing of drought termination

Differences in the timing of termination in human-influenced and naturalised data were presented visually. Histograms showing the frequency of termination start and end month were compared for both datasets for each catchment. To ensure valid comparison, as the number of drought events were not equal in human and natural data, the number of drought terminations starting/ending in each month were converted to a percentage of the total number of drought events.

## 3. Results

### 3.1. Human influence on drought events

The influence of human activities is evident in all catchments when visually comparing the two time series plots for each catchment (Figure 3). With the exception of the Bilina case study (Figure 3b), droughts events are more severe, frequently last longer and have greater deficit volumes in human-influenced data. Boxplots show that upper values in drought duration, deficit and maximum intensity are markedly higher due to human activities (Figure 4a, cf). For example, maximum drought durations are longer for all catchments except Bilina. The human-influenced time series for the Bilina catchment show the mitigation of drought compared to the naturalised time series, with a decrease in the range of drought duration, deficit and magnitude (Figure 4b).

Figure 3

Human-influenced and natural droughts and the timing of drought termination. Time series and drought events identified (red areas) for a) Svitata and b) Bilina case studies. Black triangles mark the point of DTstart for the five largest drought termination events (ranked on DTdur first, then on DTrate if DTdur was the same). Relating histograms display the frequency (as % of drought events) of termination start (light blue) and end months (dark blue) for naturalised (top) and human-influenced (bottom) drought events. DOI: https://doi.org/10.1525/elementa.365.f3

Figure 3

Human-influenced and natural droughts and the timing of drought termination. Time series and drought events identified (red areas) for a) Svitata and b) Bilina case studies. Black triangles mark the point of DTstart for the five largest drought termination events (ranked on DTdur first, then on DTrate if DTdur was the same). Relating histograms display the frequency (as % of drought events) of termination start (light blue) and end months (dark blue) for naturalised (top) and human-influenced (bottom) drought events. DOI: https://doi.org/10.1525/elementa.365.f3

The overall direction of change in drought characteristics due to anthropogenic activities is negative (drought characteristics are aggravated, Table 3; absolute values provided in Appendix 2). Average drought duration is on average up to 408% longer and average maximum intensity is up to 1047% greater in human-influenced data compared to naturalised data. Average drought deficit volumes in human-influenced data also exceed those of naturalised data by over 100% in four out of the six study catchments (reaching up to 6135% higher). Only the Bilina case study shows drought amelioration. Statistics in Table 3 reveal that larger increases were observed in the average maximum intensity and deficits of drought events in human-influenced data than in average drought duration.

Table 3

Percentage change results from naturalised to human-influenced data for drought and drought termination metrics. Any percentage change of 100% or above is in bold to highlight the high magnitude of change. DOI: https://doi.org/10.1525/elementa.365.t3

Case studyDominant human activityDrought metrics: % change due to human influenceDrought termination metrics: % change due to human influence

Drought frequency (Dfreq)Mean drought duration (Ddur)Mean drought duration (Ddef)Mean maximum intensity (MI)Mean termination duration (Ddur)Mean termination duration (Drate)

Svitata, Czech Republic Abstraction –24 +408 +6135 +1047 +247 +363
Bilina, Czech Republic Water transfer –71 –72 –85 –8 –65 +113
Ariège, France Hydropower dam –40 +25 +196 +189 +36 +131
Upper-Guadiana, Spain Abstraction +56 +95 +203 +83 +16 +38
Thames, UK Dam, abstraction, urbanisation +63 +36 +83 +23 +26 +2
Lee, UK Dam, abstraction, urbanisation +47 +41 +127 +56 +12 +54
Case studyDominant human activityDrought metrics: % change due to human influenceDrought termination metrics: % change due to human influence

Drought frequency (Dfreq)Mean drought duration (Ddur)Mean drought duration (Ddef)Mean maximum intensity (MI)Mean termination duration (Ddur)Mean termination duration (Drate)

Svitata, Czech Republic Abstraction –24 +408 +6135 +1047 +247 +363
Bilina, Czech Republic Water transfer –71 –72 –85 –8 –65 +113
Ariège, France Hydropower dam –40 +25 +196 +189 +36 +131
Upper-Guadiana, Spain Abstraction +56 +95 +203 +83 +16 +38
Thames, UK Dam, abstraction, urbanisation +63 +36 +83 +23 +26 +2
Lee, UK Dam, abstraction, urbanisation +47 +41 +127 +56 +12 +54

### 3.2. Human influence on drought termination

Overall, the influence of human activities on drought termination is present in all catchments (Figure 3 and 4, Table 3, Appendix 2). There is an increase in drought termination duration in all case studies except Bilina and higher termination rates in all case studies due to the influence of human impacts.

Human influences alter drought termination duration compared to the naturalised situation. Drought aggravating anthropogenic activities (i.e. water abstractions, mixed influences, hydropower reservoir; Figure 3a, cf) all increase the mean duration of drought termination (Table 3). This means that on average, these human-influenced systems take longer to transition from the point of a drought’s maximum intensity to reaching non-drought conditions. The boxplots in Figure 4 show that upper quartile results for drought termination duration are consistently higher for all human-influenced time series except Bilina (Figure 4a, cf). Variability was observed across all case studies in the magnitude of change, likely due to the differences in the dominant human activity. The increase in average and maximum drought termination duration was remarkably pronounced in the Svitata case study, a catchment impacted by abstraction (Table 3 and Figure 4a). The human activity of water transfers into the catchment that alleviated droughts themselves (Figure 3b, Bilina case study) also decreased average and maximum drought termination duration (Table 3 and Figure 4b).

Figure 3 (continued)

Human-influenced and natural droughts and the timing of drought termination. Time series and drought events identified (red areas) for c) Ariège and d) Upper-Guadiana case studies. Hydrograph peaks in the Upper-Guadiana reach up to 21.5 mm/month and 8.9 mm/month for naturalised and human-influenced time series, respectively. Black triangles mark the point of DTstart for the five largest drought termination events (ranked on DTdur first, then on DTrate if DTdur was the same). Relating histograms display the frequency (as % of drought events) of termination start (light blue) and end months (dark blue) for naturalised (top) and human-influenced (bottom) drought events. DOI: https://doi.org/10.1525/elementa.365.f3a

Figure 3 (continued)

Human-influenced and natural droughts and the timing of drought termination. Time series and drought events identified (red areas) for c) Ariège and d) Upper-Guadiana case studies. Hydrograph peaks in the Upper-Guadiana reach up to 21.5 mm/month and 8.9 mm/month for naturalised and human-influenced time series, respectively. Black triangles mark the point of DTstart for the five largest drought termination events (ranked on DTdur first, then on DTrate if DTdur was the same). Relating histograms display the frequency (as % of drought events) of termination start (light blue) and end months (dark blue) for naturalised (top) and human-influenced (bottom) drought events. DOI: https://doi.org/10.1525/elementa.365.f3a

Figure 3 (continued)

Human-influenced and natural droughts and the timing of drought termination. Time series and drought events identified (red areas) for e) Thames and f) Lee case studies. Black triangles mark the point of DTstart for the five largest drought termination events (ranked on DTdur first, then on DTrate if DTdur was the same). Relating histograms display the frequency (as % of drought events) of termination start (light blue) and end months (dark blue) for naturalised (top) and human-influenced (bottom) drought events. DOI: https://doi.org/10.1525/elementa.365.f3b

Figure 3 (continued)

Human-influenced and natural droughts and the timing of drought termination. Time series and drought events identified (red areas) for e) Thames and f) Lee case studies. Black triangles mark the point of DTstart for the five largest drought termination events (ranked on DTdur first, then on DTrate if DTdur was the same). Relating histograms display the frequency (as % of drought events) of termination start (light blue) and end months (dark blue) for naturalised (top) and human-influenced (bottom) drought events. DOI: https://doi.org/10.1525/elementa.365.f3b

Figure 4

Drought and drought termination results boxplots. Illustrates the average, upper and lower quartiles, and upper values in results for natural and human-influenced droughts in the a) Svitata and b) Bilina case studies c) Ariège and d) Upper-Guadiana case studies e) Thames and f) Lee case studies. DOI: https://doi.org/10.1525/elementa.365.f4

Figure 4

Drought and drought termination results boxplots. Illustrates the average, upper and lower quartiles, and upper values in results for natural and human-influenced droughts in the a) Svitata and b) Bilina case studies c) Ariège and d) Upper-Guadiana case studies e) Thames and f) Lee case studies. DOI: https://doi.org/10.1525/elementa.365.f4

Average drought termination rates increase in all case studies due to the human activities (Table 3). This means that on average, human activities increase the rate at which a system recovers from its most intense month of drought to non-drought conditions. This is true for all human activities, although the magnitude of change is larger in the Svitata, Bilina and Ariège catchments. The difference between human-influenced and naturalised data is not only apparent from the average values; the upper (and often lower) quartiles of the DTrate results are also higher in human-influenced droughts (Figure 4).

Since termination rate is calculated from maximum intensity (MI) and termination duration (DTdur, see Figure 2), the change is termination rate is the result of changes in these two variables. For case studies with drought-aggravating human influences (all except Bilina), the increase in drought termination rate is caused by the increase in MI in the human-influenced situation, despite the smaller increase in DTdur. For the Bilina case study, however, the human activity decreased MI, but DTdur decreased more, also resulting in an increased termination rate. This is discussed further in Section 4.

The influence of human activities is also seen in the timing of drought termination. The month of DTstart is indicated for five drought events in each time series in Figure 3. Termination start and end months in human-influenced catchments are often concentrated in a smaller number of months and/or in one season (Figure 3). Peaks in most frequent months for the start and end of termination are amplified and shifted in human-influenced droughts. For example, in the Upper-Guadiana, August was the most frequent month for termination beginning in 25% of events in the natural data, whereas in the human data over 70% of terminations began in April (Figure 3d). Results from the two UK catchments indicate that human activities make drought termination more likely to start towards the end of the dry season (summer) and end at the beginning of the wet season (winter).

## 4. Discussion

### 4.1. Drought termination duration

Human activities that aggravate drought manifested in longer drought terminations in all affected catchments as a result of a decrease in water availability compared to the natural situation. Water abstractions influence streamflow drought and the severity of the influence is related to the amount of abstraction (Tijdeman et al., 2018). Similarly, reservoirs have also been found to aggravate streamflow drought downstream (Firoz et al., 2018; Tijdeman et al., 2018). This likely explains the results obtained for drought termination duration, as the human-induced decrease in water availability means longer is required for the system to recharge and return to non-drought levels. In contrast, water transfer activity occurring in the Bilina catchment decreased drought termination duration. In this catchment, the low flows are artificially increased (Rolls, et al., 2012). As this activity ameliorates drought conditions (Soulsby et al., 1999; Xu et al., 2016), shorter, less intense human-influenced droughts recover faster compared to the naturalised situation without water augmentation.

This research implies that drought termination duration is therefore related to the change in drought duration and severity (i.e. maximum intensity and deficit) as a result of human activities (Figure 5). However, due to the small sample size of this study, we cannot determine a valid relationship between drought termination duration and drought characteristics. More data would be needed to test the validity of the suggested relation between the drought and drought termination metrics.

Figure 5

Relation between drought termination duration and drought metrics. Relation between the % changes in drought termination duration and a) drought duration; b) drought deficit; and c) maximum intensity. DOI: https://doi.org/10.1525/elementa.365.f5

Figure 5

Relation between drought termination duration and drought metrics. Relation between the % changes in drought termination duration and a) drought duration; b) drought deficit; and c) maximum intensity. DOI: https://doi.org/10.1525/elementa.365.f5

### 4.2. Drought termination rate

Anthropogenic activities increase the average rate at which a system changes from drought at its maximum intensity to non-drought conditions in all case studies. The most likely explanation for this is that the drought maximum intensity where the drought termination phase begins is typically lower in human-influenced situations (Figure 3 and 4 Table 3), resulting in larger drought termination rates.

It was expected that drought termination rate would be lower in case studies where drought is aggravated if water is being removed from the system for human use. If human-influenced droughts are more intense than natural droughts, the magnitude of change in streamflow required between the maximum intensity and non-drought conditions is greater. However, our results show a quicker transition from drought at its worst back to normal conditions. This could be due to active management during times of drought (Scanlon et al., 2016), a change in hydrological pathways (i.e. more surface runoff (Alaoui et al., 2018)), or a change in seasonal water demand over different times of year, allowing the system to recover when pressure on the hydrological system is low (i.e. wet season) (Ho et al., 2016). Additionally, it is important to note that the natural processes driving drought termination may be consistently great enough to compensate for a higher drought maximum intensity due to human activities.

The scatterplots of drought and drought termination rate metrics (Figure 6) show a possible relationship between drought termination rate and drought characteristics, but again more data is required to test validity.

Figure 6

Relation between drought termination rate and drought metrics. Relation between the % changes in drought termination duration and a) drought duration; b) drought deficit; and c) maximum intensity. DOI: https://doi.org/10.1525/elementa.365.f6

Figure 6

Relation between drought termination rate and drought metrics. Relation between the % changes in drought termination duration and a) drought duration; b) drought deficit; and c) maximum intensity. DOI: https://doi.org/10.1525/elementa.365.f6

Importantly, it is worth examining the question of what denotes a desirable drought termination rate. Whether a faster rate is a ‘desirable’ effect of human influences will likely depend on the circumstances of individual events or the intended group of interest (i.e. an ecosystem, specific biota or a human focus). The nature of the system in drought and whether the drivers behind the termination phase cause any further issues (i.e. intense rainfall leading to flooding) are factors that would contribute to a (possibly contradictory) definition of a desirable termination rate. For instance, water managers who are eager to lift water restrictions might favour a quicker drought termination rate, however a rapid rate caused by high precipitation events following dry spells could compromise water quality (via high levels of accumulated sediment and/or pollutants in runoff) and its suitability for drinking water (Parry et al., 2016a). Further work is recommended to refine suppositions on this topic.

### 4.3. Timing of termination

Human influences studied here create stronger seasonal patterns of when drought termination is most likely to occur. Patterns in the timing of termination were found to either become amplified or shift depending on the human activity the catchment was subjected to. This effect is visible in the Ariège and Bilina basins, where the artificial flow regime created by the reservoirs and water transfers (Firoz et al., 2018) alters the termination start and end months. Changes are highly pronounced in the Ariège catchment, however it is likely that the utility (hydropower vs irrigation or drinking water supply) and seasonal operation of the reservoir is an important factor in the resulting flow regime and consequent timings of downstream drought events (Rangecroft et al., 2019). Catchments influenced by abstractions tend to have naturally-existing patterns in their timing of termination amplified, seen especially in the Svitata and somewhat in the Upper-Guadiana, Thames and Lee, possibly relating to increasing drought occurrences during low-flow periods.

It has previously been stated that drought termination depends on the probability of a given region to receive large positive precipitation anomalies needed to end drought (Karl et al., 1987; Antofie et al., 2015). Mo (2011) established that meteorological and agricultural drought is more likely to end at the beginning of the wet season. In addition to precipitation anomalies, human activities are known to play a key part in moderating the seasonal cycle of water availability. For example, a demand for water for agriculture that has a strong seasonal cycle that differs between summer (growing season in Europe) and winter often determines seasonal abstraction regimes (Wada et al., 2013). Flow regulated by a hydropower reservoir will likely depend on the seasonal demand for electricity and the flow regime required to maintain the storage or release water to generate power when needed (Kern et al., 2012; Ashraf et al., 2018). Industrial demands have patterns determined by economic factors (Flörke et al., 2013) and the degree of seasonal patterns in drinking water requirements and municipal demands can depend on the degree of outdoor water use (du Plessis and Jacobs, 2014). It is possible that the seasonal cycle of a catchment is enhanced due to the human-induced water shortages that can mirror the seasonality of precipitation and potential evaporation patterns (Apruv et al., 2017), causing more synchronicity in drought termination start and end months (i.e. due to the most water use during the low flow period).

### 4.4. Limitations and uncertainties

Any study using a hydrological model will have associated uncertainties (Beven, 1993; Gupta et al., 1998; Walker et al., 2003). This is an important consideration as uncertainty determines model confidence in simulating the naturalised situation, which was the basis of the comparisons in this work. Where results are published, levels of uncertainty are within acceptable limits and model performance is good (Van Loon and Van Lanen, 2013, 2015). However, re-wetting and recovery are not always simulated well in hydrological models (Birkel et al., 2011) and therefore the processes most relevant to drought termination (i.e. soil moisture or subsurface stores) may have larger uncertainties. This could lead to an over- or under-estimation of the effect of human influences on streamflow and consequently drought termination metrics. When naturalisation uses information on abstraction licence limits, this assumes that water users are using the full allowance of their licence. Potential inaccuracies are introduced if full licence limits are not being used, or if unrecorded and/or illegal abstraction takes place. However, information on actual abstracted volumes is often sensitive and thereby very difficult to obtain.

A variety of naturalisation methods were used to establish the naturalised time series for the case studies used in this study (Appendix 1). Different methods of naturalisation introduce some uncertainty into cross-catchment comparisons due to the lack of uniformity in the processes considered and calculations used to produce the naturalised flow record.

The brevity of available hydrological records is a constraint for most studies of drought (Barker et al., 2016). A small sample size due to the limited number of appropriate catchments with an available human-influenced (observed) and naturalised time series, and also in terms of the total number of drought events identified (i.e. the Bilina case study had only five drought events in the human-influenced time series), could limit results. This could be addressed by expanding the number of case studies in future research.

Hydrometric accuracy (particularly at low flows) and the evolution of a gauge station over time means it is often impossible to have an accurate, homogeneous flow record. This creates uncertainty, particularly affecting stations with longer records such as those of the Thames (Marsh and Harvey, 2012) and the Lee. However, it was recognised that the benefit of using these longer records in analysis outweighs the disadvantages of uncertain hydrometric accuracy.

The period of record could influence the results. Here, we used the full period of record available in analyses, as, even though human influences were highly unlikely to have been equal throughout the time period (particularly for the Thames and Lee catchments due to their >100-year record), it was felt that this would give the greatest representation of drought and drought termination metrics when using averages in the results. It was noted however that the period chosen can indeed influence the results obtained. For instance, when using only the 1951–2017 human-influenced and naturalised flow records for the Thames, the % change due to human influences in average drought termination rate was 36% (results not shown), compared to just 2% using the period 1900–2017. It is likely that different record periods will have experienced different natural (i.e. climatic) and anthropogenic (i.e. abstraction activity) factors that will be driving differences in results. In addition, changes in how humans intervene within the hydrological system in response to the introduction of policy concepts and regulations, such as the maintenance of environmental flows (EA, 2013), could shift the active management of drought and alter termination characteristics. Further work is therefore needed to explore the use of different time periods in drought termination analyses to illustrate effects of non-stationarity in human influences.

### 4.5. Water management applications

The framework used here to assess the termination phase allows application to any climatic/geographical region. This type of research (and further work) is invaluable for the formulation of adaptive responses to drought conditions, especially given that it is often human-influenced systems that we are most reliant upon for our water supply (Barker et al., 2016). Given that human influences are generally set to increase (Alcamo et al. 2003), attempting to quantify their impacts on aspects of drought is essential for improving knowledge on how to improve drought management.

Longer drought recoveries have consequences for water companies under increasing pressure to meet potable water demands and ensure sustainability in their abstractions. The increased time available for the accumulation of pollutants (which might otherwise be present in streamflow in diffuse concentrations in non-drought conditions) during a longer termination phase may lead to water quality concerns when recovery and contaminant re-mobilisation does occur. Further to this, water companies have legal obligations to meet flow thresholds and regimes that maintain critical environmental thresholds for water-dependent species (EA, 2013). If longer time periods are expected for termination, strategies in water resource management will have to adapt to minimise these impacts.

Understanding the link between drought termination duration and drought severity as a result of human water use and modifications enables water managers to balance supply and demand and the manner in which termination is likely to proceed. Knowledge of the rate at which the system recovers and the human influence on seasonal drought/termination patterns are therefore an inherent part of improving probabilistic drought forecasting (Yuan et al., 2013). Information of this kind is also valuable for reservoir managers who have greater levels of control over stored resources (Návar and Lizárraga-Mendiola, 2016). Water managers may opt to implement short-, medium-, or long-term water-saving measures if they are aware of actions known to delay the recovery of the system (Bell et al., 2013). This also assists policy makers overseeing legislation (i.e. abstraction licences) with the aim of increasing sustainability and resilience to drought risks (DEFRA, 2013).

Knowledge of the seasonal cycle of water availability within a catchment is essential for those handling water resource management decisions. The findings of this study suggest that the timings associated with human-influenced drought termination could be predicted with more confidence than those in natural systems. Although it is arguably generally unfavourable to have fewer opportunities for drought termination, it could allow more robust decision-making if there is less uncertainty on its timing.

### 4.6. Future Research Directions

Work has begun exploring the predictability of drought recovery using global circulation models and large-scale meteorological forcings (Mo, 2011) or water-balance approaches (Bell et al., 2013). However, the significant impact of human activities is highly likely to be responsible in explaining some of the variation not captured by such climate/hydrological models. It is therefore important to consider the influence, non-stationarity and synergistic/antagonistic effects of our activities.

The specific effect of management practices on the termination phase, including large-scale proactive measures (i.e. managed aquifer recharge) and water saving technologies (Scanlon et al., 2016) merit investigation. Assessing the impact of human activities where more than one is occurring in a catchment, with possible synergistic or non-linear effects, is yet another challenge that needs addressing (Parry et al., 2016b). A comparison between drought onset/development and termination phases may help reveal the significance of mechanisms or management strategies at work over the course of an event (Vernon-Kidd et al., 2017). As management practices change over short (daily, monthly) and longer (decadal) time periods, it would be interesting to test their impact on the termination phase as they evolve. This is no easy task however, as the type of information required (i.e. abstraction records) from governments/water managers is often sensitive or unreliable (Bromley et al., 2001; Novo et al., 2015).

Finally, the current sample size is small, therefore further work should aim to build on these results by adding more case studies and covering other human activities not included here, with a view to create a typology of termination characteristics under prevalent human influences. Examples include other reservoir purposes, urbanisation, deforestation and afforestation and other large-scale land-use changes that are yet to be studied in this context (Laurence and Williamson, 2001; Tallaksen and Van Lanen, 2004; Zhang et al., 2017). This will help to produce a deeper understanding of how human activities are impacting the hydrological system, drought termination and ultimately help to guide water resource management in the future.

## 5. Summary and conclusions

Here we present the first study that analyses anthropogenic influences on the streamflow drought termination phase. Our example case studies in Europe have shown that human activities have a notable effect on drought termination characteristics. In response to human activities and compared to the naturalised situation, most case studies show a longer drought termination duration (by up to 247%). A faster termination rate (with some over two orders of magnitude higher) is observed in human-influenced data due to an often greater drought maximum intensity than in naturalised data. The time of year at which the termination phase begins and ends was observed to shift in systems under human influence, with more start and end months concentrated in a single season.

Building on these results will help gain a wider understanding on how human actions are modifying hydrological droughts and help to improve drought management policies. This is crucial for our understanding of how the transition from depleted to replenished water supplies is operationally handled (Hannaford et al., 2011; Bell et al., 2013) and should feature as a fundamental part of effective drought monitoring and early warning systems. Despite the increasing influence of human activities on water resources, their various impacts on drought termination were previously unknown. Further work should aim to build on current results by adding further case studies and covering other influential human activities not included here. Given the projected increase and non-stationarity of human influences, knowledge of this kind is subsequently invaluable for the formulation of adaptive responses to drought and managing the changing threat of water scarcity. This is extremely important as it is human-influenced systems that we are most reliant upon for water supply.

## Data Accessibility Statement

The data used in this paper will be made available after an embargo period to allow further publication using this dataset.

### 6. Appendix 1: Case studies and naturalisation methods

This section presents further details on the case study catchments, introduces the flow naturalisation approach and describes the type of naturalisation method used to generate the naturalised time series analysed in this paper.

#### 6.1 Flow naturalisation

Flow naturalisation involves considering the extent of human influences within a catchment, the scale of their impact on flows and the purpose for which the flow data are required. Artificial losses from a river system are often abstractions for public water supply, irrigation, agriculture and industry. Common human-induced gains are from sewage and industrial effluent returns, irrigation return flows and inter-basin transfers. The purpose of flow naturalisation, the parameters relevant to the study in question, and, crucially, the data available are important when determining the most appropriate naturalisation method.

Two commonly-used methods are naturalisation by decomposition and rainfall-runoff modelling (Brandt et al., 2017). Naturalisation by decomposition involves breaking the observed flow record down into its components (EA, 2001). The natural flow is deduced by quantifying all the artificial components in the observed flow record using the following calculation:

Natural flow = Gauged river flow + sum of all upstream abstractions – sum of all upstream discharges and return flows

This method is dependent on the availability of good quality, complete data for both the observed flows and artificial influences. Records of abstraction/discharge licence volumes and the purpose of the abstraction/discharge are examples of key information required for this process.

Rainfall-runoff modelling can be used to generate naturalised flows if the available data on artificial influences are inadequate for a decomposition approach (Porter and McMahon, 1971). Observed data from a period of undisturbed (natural) river flow is instead required with which to calibrate the model.

Data used in this study was naturalised prior to acquisition. The most suitable technique employed to produce these naturalised time series was therefore determined according to the type and availability of data and its applicability to either the naturalisation by decomposition or rainfall-runoff modelling approach.

#### 6.2 Svitata, Czech Republic

This catchment is located in eastern Czech Republic, underlain by sandstone aquifers suitable for drinking water extraction. Groundwater abstraction for drinking water therefore constitutes the main human influence and has increased considerably since 1975 (Tallaksen and Van Lanen, 2004). This significantly affects the catchment’s flow regime. According to the Köppen-Geiger classification system (Kotek et al., 2006) the climate type is Dfb, with warm summers, a humid continental climate and no significant precipitation difference between seasons. Despite the relatively short time series, the period of record used here coincided with at least one period of severe European hydrological drought (i.e. 1975–1976; EDC, 2019a) and therefore still captures a good range of climatic variability ranging from non-drought and drought conditions.

The BILAN lumped conceptual rainfall-runoff model (Kašpárek, 1998) was used by Van Loon and Van Lanen (2015) with the concepts of the observation-modelling framework (Van Loon and Van Lanen, 2013) to provide the naturalised data used here. The model solves the catchment-average water balance on a monthly timescale, calibrated using the observed discharge from the undisturbed period. Precipitation, air temperature and relative humidity data were used as input. Outputs were found to agree reasonably well with observations (Tallaksen and Van Lanen, 2004).

#### 6.3 Bilina, Czech Republic

Located in Krušne mountains, north-west Czech Republic, the development of large-scale mining activities in the catchment meant that the Bilina’s natural discharge was insufficient to meet the demands of growing industries and drinking water supply (Tallaksen and Van Lanen, 2004). Consequently, the main human influence on the Bilina since 1960 has been the augmentation with water transported from a nearby river basin. The dominant climate type is Cfb, indicating a mild temperate oceanic climate.

The BILAN model was used by Van Loon and Van Lanen (2015) with the concepts of the observation-modelling framework (Van Loon and Van Lanen, 2013) to also provide the naturalised data for this catchment.

#### 6.4 Ariège, France

The Ariège catchment is a mountainous sub-basin of the larger Garonne River basin in the Pyrenees. The river flow regime is influenced by a series of high-elevation reservoirs, relying mostly on snowmelt for recharge, whose main function is the production of hydro-electricity (Hendrickx and Sauquet, 2013). Water transfers (operating in both directions) between the Ebro River basin, located on the southern side of the Pyrenees in Spain, and the headwaters of the Ariège basin constitute a further human influence on the catchment. The climate classification of the basin is Cfb, meaning it experiences a mild temperate oceanic climate. Despite the relatively short record length, the time series for this case study coincided with periods of severe European hydrological drought (i.e. 2003; EDC, 2019b) and therefore captures a good range of climatic variability ranging from non-drought and drought conditions.

The naturalisation by decomposition process removed the effects of the hydropower reservoir storage by taking account of reservoir levels and released discharges and adding them to the observed flow values (Vidal, 2017). The reconstructed time series was then lagged to adjust for routing times between water bodies.

The Upper-Guadiana catchment is a headwater catchment of the Guadiana River, located in central Spain. It has a Mediterranean, semi-arid climate with very warm summers and mild winters (classes Csa, Csb and Bsk) (Van Loon and Van Lanen, 2013). The catchment has experienced severe multi-year meteorological droughts during the 1980–90s, however the presence of water stores (aquifer systems and wetlands) often attenuates these anomalies after periods of high precipitation, preventing further propagation through the hydrological system (Van Loon and Van Lanen, 2013). When hydrological drought does develop, events tend to be very long due to the combination of a semi-arid climate and slow response time to precipitation. Agriculture (mainly vineyards) is the dominant land use in the catchment, with human influence on the hydrological regime of the Upper-Guadiana being abstraction for irrigation and artificial drainage. Borehole abstractions increased by 244% between 1974–1988, causing dramatic drawdown in water tables.

Van Loon and Van Lanen (2013) developed the observation-modelling framework to quantify human influence in this catchment. For this, the HBV hydrological model was used to generate the naturalised data utilised in this research.

#### 6.6 Thames, UK

The Thames catchment in south-east England is one of the driest regions in the country, experiencing a mild temperate oceanic climate (class Cfb). The gauging station at Kingston has one of the longest flow records in the UK. The basin overlies major chalk and oolite aquifers, which help sustain baseflow. This is a mixed influence catchment with several human activities influencing the hydrological regime simultaneously. Regulation from surface water, groundwater and reservoirs affect runoff, which is reduced by major abstractions for public water supply (PWS), industry and agriculture (CEH, 2019). Some effluent returns increase flows but overall this is substantially less than the effect of abstractions. Furthermore, the region has a high population and urban extent, which has increased since 1900.

Unlike the aforementioned case studies, no undisturbed period of record was available for the Thames catchment with which to calibrate a hydraulic model and calculate a naturalised time series. However, as records of abstraction and/or discharge licences are kept (and have historically been kept) in the UK, this allows naturalised flows to instead be calculated with a naturalisation by decomposition method. The major abstractions and discharges are monitored and reported by the parties involved, the volumes of which are used to adjust the gauged daily flow data on an average daily and monthly basis (Hammet, 2017).

#### 6.7 Lee, UK

The gauging site for this catchment is located in the middle reaches of the River Lee, south-east England, downstream of the confluence with the River Stort. It is a pervious chalk basin, consisting of predominantly rural headwaters with significant urban growth in the lower valley (CEH, 2019). It has a mild temperate oceanic climate (Cfb). Flows are reduced by surface water abstraction for PWS, industry and agriculture, and marginally increased by effluent returns. Considerable groundwater abstractions are exported from catchment (CEH, 2019). Like the Thames, naturalised flows are calculated with a naturalisation by decomposition method due to the availability of abstraction and/or discharge licence information and the absence of an undisturbed period of record.

### 7. Appendix 2: Drought and drought termination results

Table 7–1

Drought and drought termination metrics results from naturalised data.

Case studyDominant human activityNatural drought metricsNatural drought termination metrics

Drought frequency (Dfreq)Mean drought duration (Ddur) (months)Mean drought duration (Ddef) (mm)Mean maximum intensity (MI) (mm)Mean termination duration (Ddur) (months)Mean termination duration (Drate) (mm/months)

Svitata, Czech Republic Abstraction 17 0.9 0.5 1.4 0.4
Bilina, Czech Republic Water transfer 17 4.2 10.9 1.7 2.8 0.8
Ariège, France Hydropower dam 20 1.8 15.5 9.9 1.4 7.3
Upper-Guadiana, Spain Abstraction 16 3.6 0.6 0.2 2.7 0.09
Thames, UK Dam, abstraction, urbanisation 93 3.1 9.9 3.8 2.0 2.5
Lee, UK Dam, abstraction, urbanisation 81 3.4 6.1 1.9 2.3 1.1
Case studyDominant human activityNatural drought metricsNatural drought termination metrics

Drought frequency (Dfreq)Mean drought duration (Ddur) (months)Mean drought duration (Ddef) (mm)Mean maximum intensity (MI) (mm)Mean termination duration (Ddur) (months)Mean termination duration (Drate) (mm/months)

Svitata, Czech Republic Abstraction 17 0.9 0.5 1.4 0.4
Bilina, Czech Republic Water transfer 17 4.2 10.9 1.7 2.8 0.8
Ariège, France Hydropower dam 20 1.8 15.5 9.9 1.4 7.3
Upper-Guadiana, Spain Abstraction 16 3.6 0.6 0.2 2.7 0.09
Thames, UK Dam, abstraction, urbanisation 93 3.1 9.9 3.8 2.0 2.5
Lee, UK Dam, abstraction, urbanisation 81 3.4 6.1 1.9 2.3 1.1
Table 7–2

Drought and drought termination metrics results from human-influenced data.

Case studyDominant human activityHuman-influenced drought metricsHuman-influenced drought termination metrics

Drought frequency (Dfreq)Mean drought duration (Ddur) (months)Mean drought duration (Ddef) (mm)Mean maximum intensity (MI) (mm)Mean termination duration (Ddur) (months)Mean termination duration (Drate) (mm/months)

Svitata, Czech Republic Abstraction 13 10.2 53.8 5.6 4.7 1.7
Bilina, Czech Republic Water transfer 1.2 1.6 1.6 1.6
Ariège, France Hydropower dam 12 2.3 45.9 28.5 1.8 16.8
Upper-Guadiana, Spain Abstraction 25 7.0 1.7 0.4 3.1 0.12
Thames, UK Dam, abstraction, urbanisation 152 4.2 18.2 4.7 2.5 2.6
Lee, UK Dam, abstraction, urbanisation 119 4.8 13.9 2.9 2.6 1.7
Case studyDominant human activityHuman-influenced drought metricsHuman-influenced drought termination metrics

Drought frequency (Dfreq)Mean drought duration (Ddur) (months)Mean drought duration (Ddef) (mm)Mean maximum intensity (MI) (mm)Mean termination duration (Ddur) (months)Mean termination duration (Drate) (mm/months)

Svitata, Czech Republic Abstraction 13 10.2 53.8 5.6 4.7 1.7
Bilina, Czech Republic Water transfer 1.2 1.6 1.6 1.6
Ariège, France Hydropower dam 12 2.3 45.9 28.5 1.8 16.8
Upper-Guadiana, Spain Abstraction 25 7.0 1.7 0.4 3.1 0.12
Thames, UK Dam, abstraction, urbanisation 152 4.2 18.2 4.7 2.5 2.6
Lee, UK Dam, abstraction, urbanisation 119 4.8 13.9 2.9 2.6 1.7

## Acknowledgments

This paper was developed within the framework of the Panta Rhei Research Initiative of the International Association of Hydrological Sciences (IAHS), as part of the Working Group ‘Drought in the Anthropocene’. It supports the work of the UNESCO-IHP VIII FRIEND-Water programme. We thank Ladislav Kašpárek (WRI, Czech Republic) for providing data and model results for the Bilina and Svitata, Jean-Philippe Vidal (IRSTEA, France) for data for the Ariège, the institutes UCLM, CSIC, and AEMET (Spain) for data for the Upper-Guadiana, and the National River Flow Archive (CEH, UK) for data for the Thames and Lee. We specifically like to mention Stephen Turner, Katie Muchan and Cath Sefton (CEH, UK) for providing the naturalised Lee data. Finally, we would like to extend our gratitude to the anonymous individuals for offering their thoughtful and insightful feedback during the review process.

## Funding information

AVL and SR were financially supported by the NWO project ‘Adding the human dimension to drought’ (reference number: 2004/08338/ALW). DW received financial support from NERC CENTA (reference number: NE/lL002493/1).

## Competing interests

The authors have no competing interests to declare.

## Author contributions

• Contributed to conception and design: JM, SR, AVL, SP, and DW

• Contributed to acquisition of data: SP, SR, and AVL

• Contributed to analysis and interpretation of data: JM, SR, AVL

• Drafted and/or revised the article: JM, with input of SR, SP, DW, and AVL

• Approved the submitted version for publication: JM, SR, SP, DW, and AVL

1
Alaoui
,
A
,
Rogger
,
M
,
Peth
,
S
and
Blöschl
,
G
.
2018
.
‘Does soil compaction increase floods? A review.’
Journal of Hydrology
557
:
631
642
. DOI:
2
Alcamo
,
J
,
Doll
,
P
,
Henrichs
,
T
,
Kaspar
,
F
,
Lehner
,
B
,
Rosch
,
T
and
Siebert
,
S
.
2003
.
‘Global estimation of water withdrawals and availability under current and business as usual conditions’
.
Hydrologival Sciences Journal
48
:
339
48
. DOI:
3
Antofie
,
T
,
Naumann
,
G
,
Spinoni
,
J
and
Vogt
,
J
.
2014
.
‘Estimating the water needed to end or ameliorate the drought in the Carpathian region’
.
Hydrology and Earth System Sciences
11
:
1493
1527
. DOI:
4
Apurv
,
T
,
Sivapalan
,
M
and
Cai
,
X
.
2017
.
‘Understanding the Role of Climate Characteristics in Drought Propagation’
.
Water Resources Research
53
:
9304
9329
. DOI:
5
Ashraf
,
FB
,
Haghighi
,
AT
,
Riml
,
J
,
Alfredsen
,
K
,
Koskela
,
JJ
,
Kløve
,
B
and
Marttila
,
H
.
2018
.
‘Changes in short term river flow regulation and hydropeaking in Nordic rivers’
.
Scientific Reports
8
:
17232
. DOI:
6
Banque Hydro French Database
.
2017
. Accessible at: www.hydro.eaufrance.fr (last accessed on: 15/05/2018).
7
Barker
,
LJ
,
Hannaford
,
J
,
Chiverton
,
A
and
Svensson
,
C
.
2016
.
‘From meteorological to hydrological drought using standardised indicators’
.
Hydrology and Earth System Sciences
20
:
2483
2505
. DOI:
8
Barnett
,
TP
,
Pierce
,
DW
,
Hidalgo
,
HG
,
Bonfils
,
C
,
Santer
,
BD
,
Das
,
T
,
Bala
,
G
,
Wood
,
AW
,
Nozawa
,
T
,
Mirin
,
AA
,
Cayan
,
DR
and
Dettinger
,
MD
.
2008
.
‘Human-Induced Changes in the Hydrology of the Western United States’
.
Sciencexpress
. DOI:
9
Bell
,
VA
,
Davies
,
HN
,
Kay
,
AL
,
Marsh
,
TJ
,
Brookshaw
,
A
and
Jenkins
,
A
.
2013
.
‘Developing a large-scale water-balance approach to seasonal forecasting: Application to the 2012 drought in Britain’
.
Hydrological Processes
27
:
3003
3012
. DOI:
10
Beven
,
K
.
1993
.
‘Prophecy, reality and uncertainty in distributed hydrological modelling’
.
16
(
1
):
41
51
. DOI:
11
Birkel
,
C
,
Soulsby
,
C
and
Tetzlaff
,
D
.
2011
.
‘Modelling catchment-scale water storage dynamics: reconciling dynamic storage with tracer-inferred passive storage’
.
Hydrological Processes
25
(
25
):
3924
3936
. DOI:
12
Brandt
,
MJ
,
Johnson
,
KM
,
Elphinston
,
AJ
and
Ratnayaka
,
DD
.
2017
. ‘Twort’s Water Supply’, 7th Edition.
London
:
IWA Publishing
.
13
Bravar
,
L
and
Kavvas
,
ML
.
1991
.
‘On the physics of droughts, II: Analysis and simulation of the interaction of atmospheric and hydrologic processes during droughts’
.
Journal of Hydrology
129
:
299
333
. DOI:
14
Bromley
,
J
,
Cruces
,
J
,
Acreman
,
M
,
Martınez
,
L
and
Llamas
,
MR
.
2001
.
‘Problems of sustainable groundwater management in an area of over-exploitation: The Upper Guadiana catchment, central Spain’
.
Int. J. Water Resour. Dev
.
17
(
3
):
379
396
. DOI:
15
CEH (Centre for Ecology and Hydrology)
.
2019
.
National River Flow Archive
. http://nrfa.ceh.ac.uk/ (last accessed on: 25/02/2019).
16
Dai
,
A
.
2011
.
‘Drought under global warming: A review’
.
Wiley Interdisciplinary Reviews: Climate Change
2
(
1
):
45
65
. DOI:
17
DEFRA (Department for Environment Food & Rural Affairs)
.
2013
. ‘Managing abstraction and the water environment’.
Cardiff, Wales
. Available at: https://consult.defra.gov.uk/water/abstractionreform/supporting_documents/abstractreformconsultmanage20131217.pdf (last accessed on: 27/12/2018).
18
Dettinger
,
MD
.
2013
.
‘Atmospheric rivers as drought busters on the US west coast’
.
Journal of Hydrometeorology
14
(
6
):
1721
1732
. DOI:
19
Du Plessis
,
JL
and
Jacobs
,
HE
.
2014
.
‘Model for estimating domestic outdoor water demand of properties in residential estates’
.
Procedia Engineering
89
:
967
974
. DOI:
20
EA (Environment Agency)
.
2001
.
‘Good practice in flow naturalisation by decomposition’
National Hydrology Group, Environment Agency, Version 2.0
.
21
EA (Environment Agency)
.
2013
.
‘Environmental Flow Indicator’
.
Factsheet
. Available at: http://webarchive.nationalarchives.gov.uk/20140328104910/http://cdn.environment-agency.gov.uk/LIT_7935_811630.pdf (last accessed on: 27/12/2018).
22
EDC (European Drought Centre)
.
2019a
.
‘Drought of 1976–1976: Central and Northern Europe’
. Available at: http://www.geo.uio.no/edc/droughtdb/edr/DroughtEvents/1976_Event.php (last accessed on: 16/05/2019).
23
EDC (European Drought Centre)
.
2019b
.
‘Drought of 2003: Europe’
. Available at: http://www.geo.uio.no/edc/droughtdb/edr/DroughtEvents/2003_Event.php (last accessed on: 16/05/2019).
24
Firoz
,
AMB
,
Nauditt
,
A
,
Fink
,
M
and
Ribbe
,
L
.
2018
.
‘Quantifying human impacts on hydrological drought using a combined modelling approach in a tropical river basin in central Vietnam’
.
Hydrology and Earth System Sciences
22
(
1
):
547
565
. DOI:
25
Fleig
,
AK
,
Tallaksen
,
LM
,
Hisdal
,
H
and
Demuth
,
S
.
2006
.
‘A global evaluation of streamflow drought characteristics’
.
Hydrology and Earth System Sciences
10
:
535
552
. DOI:
26
Flörke
,
M
,
Kynast
,
E
,
Bärlund
,
I
,
Eisner
,
S
,
Wimmer
,
F
and
Alcamo
,
J
.
2013
.
‘Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study’
.
Global Environmental Change
23
(
1
):
144
156
. DOI:
27
Gupta
,
HV
,
Sorooshian
,
S
and
Yapo
,
PO
.
1998
.
‘Toward improved calibration of hydrological models: Multiple and noncommensurable measures of information’
.
Water Resources Research
34
(
4
):
751
763
. DOI:
28
Hamilton
,
C
.
2016
.
‘Define the Anthropocene in terms of the whole Earth’
.
Nature
536
:
251
. DOI:
29
Hammett
,
J
.
2017
.
Personal correspondence via email with Environment Agency
30
Hannaford
,
J
,
Lloyd-Hughes
,
B
and
Keef
,
C
.
2011
.
‘Examining the large-scale spatial coherence of European drought using regional indicators of precipitation and streamflow deficit’
.
Hydrological Processes
25
:
1146
1162
. DOI:
31
Heim
,
RR
, Jr
and
Brewer
,
MJ
.
2012
.
‘The global drought monitor portal: The foundation for a global drought information system’
.
Earth Interactions
16
(
15
):
1
28
. DOI:
32
Hendrickx
,
F
and
Sauquet
,
E
.
2013
.
‘Impact of warming climate on water management for the Ariège River basin (France)’
.
Hydrological Sciences Journal
58
(
5
):
976
993
. DOI:
33
Heudorfer
,
B
and
Stahl
,
K
.
2016
.
‘Comparison of different threshold level methods for drought propagation analysis in Germany’
.
Hydrology Research
48
(
3
). DOI:
34
Hisdal
,
H
and
Tallaksen
,
LM
.
2000
. ‘Drought Event Definition’.
Technical Report No. 6: Assessment of the Regional Impact of Droughts in Europe, Department of Geophysics
.
Norway
:
University of Oslo
2000.
35
Hisdal
,
H
,
Tallaksen
,
LM
,
Clausen
,
B
,
Peters
,
E
and
Gustard
,
A
.
2004
. ‘Hydrological drought characteristics’. In:
Tallaksen
,
LM
and
Van Lanen
,
HAJ
(eds.),
Hydrological drought: processes and estimation methods for streamflow and groundwater
48
:
139
198
.
Elsevier
.
36
Ho
,
M
,
Parthasarathy
,
V
,
Etienne
,
E
,
Russo
,
TA
,
Devineni
,
N
and
Lall
,
U
.
2016
.
‘America’s water: Agricultural water demands and the response of groundwater’
.
Geophysical Research Letters
43
:
7546
7555
. DOI:
37
Karl
,
T
,
Quinlan
,
F
and
Ezell
,
DS
.
1987
.
‘Drought termination and amelioration: Its climatological probability’
.
Journal of Climate and Applied Meteorology
26
:
1198
1209
. DOI:
38
Kašpárek
,
L
.
1998
. ‘Regional study of the impacts of climate change on hydrological conditions in the Czech Republic’.
Report T.G., Masaryk Water Research Institute
,
Prague, Czech Republic
.
39
Kern
,
JD
,
Characklis
,
GW
,
Doyle
,
SB
and
Whisnant
,
RB
.
2012
.
‘Influence of deregulated electricity markets on hydropower generation and downstream flow regime’
.
Journal of Water Resources Planning and Management
138
(
4
):
342
355
. DOI:
40
Kirono
,
DGC
,
Kent
,
DM
,
Hennessy
,
KJ
and
Mpelasoka
,
F
.
2011
.
‘Characteristics of Australian droughts under enhanced greenhouse conditions: Results from 14 global climate models
.
Journal of Arid Environments
75
(
6
):
566
575
. DOI:
41
Kottek
,
M
,
Grieser
,
J
,
Beck
,
C
,
Rudolf
,
B
and
Rubel
,
F
.
2006
.
‘World Map of the Köppen-Geiger climate classification updated’
.
Meteorologische Zeitschrif
15
:
259
263
. DOI:
42
Laurence
,
WF
and
Williamson
,
GB
.
2001
.
‘Positive feedbacks among forest fragmentation, drought and climate change in the Amazon’
.
Conservation Biology
15
(
6
):
1529
1535
. DOI:
43
Liu
,
Y
,
Ren
,
L
,
Zhu
,
Y
,
Yang
,
X
,
Yuan
,
F
,
Jiang
,
S
and
Ma
,
M
.
2016
.
‘Evolution of hydrological drought in human disturbed areas: A case study in the Laohahe Catchment, northern China’
.
. Article ID: 5102568. DOI:
44
Loecke
,
TD
,
Burgin
,
AJ
,
Riveros-Iregui
,
DA
,
Ward
,
AS
,
Thomas
,
SA
,
Davis
,
CA
and St.
Clair
,
MA
.
2017
.
‘Weather whiplash in agricultural regions drives deterioration of water quality’
.
Biogeochemistry Letters
133
:
7
15
. DOI:
45
Marsh
,
T
and
Harvey
,
CL
.
2012
.
‘The Thames flood series: a lack of trend in flood magnitude and a decline in maximum levels’
.
Hydrology Research
43
(
3
):
203
214
. DOI:
46
McKee
,
TB
,
Doesken
,
NJ
and
Kleist
,
J
.
1993
.
‘The relationship of drought frequency and duration to time scales’
. In:
Proceedings of the 8th conference on applied climatology, Anaheim, American Meteorological Society
,
179
184
.
47
Mishra
,
AK
and
Singh
,
VP
.
2010
.
‘A review of drought concepts’
.
Journal of Hydrology
391
:
202
216
. DOI:
48
Mo
,
KC
.
2011
.
‘Drought onset and recovery over the United States’
.
Journal of Geophysical Research
116
:
D20106
. DOI:
49
Návar
,
J
and
Lizárraga-Mendiola
,
L
.
2016
.
‘Temporal Patterns in River Flows in Basins in Northern Mexico’
.
Technology and Water Sciences
7
(
1
):
35
44
.
50
Nkemdirim
,
L
and
Weber
,
L
.
1999
.
‘Comparison between the droughts of the 1930s and the 1980s in the southern prairies of Canada’
.
Journal of Climate
12
:
2434
2450
. DOI:
51
Novo
,
P
,
Dumont
,
A
,
Willaarts
,
BA
and
Lopez-Gunn
,
E
.
2015
.
‘More cash and jobs per illegal drop? The legal and illegal water footprint of the Western Mancha Aquifer (Spain)’
.
Environmental Science & Policy
51
:
256
266
. DOI:
52
Palmer
,
WC
.
1965
.
‘Meteorological drought’
.
US Department of Commerce, Weather Bureau Research Paper. 45
,
58
.
53
Pan
,
M
,
Yuan
,
X
and
Wood
,
EF
.
2013
.
‘A probabilistic framework for assessing drought recovery’
.
Geophysical Research Letters
40
:
1
6
. DOI:
54
Parry
,
S
,
Prudhomme
,
C
,
Wilby
,
RL
and
Wood
,
PJ
.
2016a
.
‘Drought termination: concept and characterisation’
.
Processes in Physical Geography
40
(
6
):
743
767
. DOI:
55
Parry
,
S
,
Wilby
,
RL
,
Prudhomme
,
C
and
Wood
,
PJ
.
2016b
.
‘A systematic assessment of drought termination in the United Kingdom’
.
Hydrology and Earth System Sciences
20
:
4265
4281
. DOI:
56
Parry
,
S
,
Wilby
,
R
,
Prudhomme
,
C
,
Wood
,
P
and
McKenzie
,
A
.
2018
.
‘Demonstrating the utility of a drought termination framework: prospects for groundwater level recovery in England and Wales in 2018 or beyond’
.
Environmental Research Letters
13
(
6
):
064040
. DOI:
57
Porter
,
JW
and
McMahon
,
TA
.
1971
.
‘A model for the simulation of streamflow data from climatic records’
.
Journal of Hydrology
13
:
297
324
. DOI:
58
Rangecroft
,
S
,
Van Loon
,
AF
,
Maureira
,
H
,
Verbist
,
K
and
Hannah
,
DM
.
2019
.
‘An observation-based method to quantify the human influence on hydrological drought: upstream–downstream comparison’
.
Hydrological Sciences Journal
. in press. DOI:
59
Rolls
,
RJ
,
Leigh
,
C
and
Sheldon
,
F
.
2012
.
‘Mechanistic effects of low-flow hydrology on riverine ecosystems: ecological principles and consequences of alteration’
.
Freshwater Science
31
(
4
):
1163
1186
. DOI:
60
Samaniego
,
L
,
Thober
,
S
,
Kumar
,
R
,
Wanders
,
N
,
Rakovec
,
O
,
Pan
,
M
,
Zink
,
M
,
Sheffield
,
J
,
Wood
,
EF
and
Marx
,
A
.
2018
.
‘Anthropogenic warming exacerbates European soil moisture droughts’
.
Nature Climate Change
8
:
421
426
. DOI:
61
Scanlon
,
BR
,
Reedy
,
RC
,
Faunt
,
CC
,
Pool
,
D
and
Uhlman
,
K
.
2016
.
‘Enhancing drought resilience with conjunctive use and managed aquifer recharge in California and Arizona’
.
Environmental Research Letters
. DOI:
62
Soulsby
,
C
,
Gibbons
,
CN
and
Robins
,
T
.
1999
.
‘Inter-Basin Water Transfers and Drought Management in the Kielder/Derwent System’
.
Water and Environment Journal
13
(
3
):
213
223
. DOI:
63
Steffen
,
W
,
,
W
,
Deutsch
,
L
,
Gaffney
,
O
and
Ludwig
,
C
.
2015
.
‘The trajectory of the Anthropocene: The Great Acceleration’
.
The Anthropocene Review
2
(
1
):
81
98
. DOI:
64
Tallaksen
,
LM
and
Van Lanen
,
HAJ
.
2004
. ‘Hydrological drought: processes and estimation methods for streamow and groundwater’. In:
Developments in Water Science
48
.
Amsterdam, the Netherlands
:
Elsevier Science B.V.
65
Tijdeman
,
E
,
Hannaford
,
J
and
Stahl
,
K
.
2018
.
‘Human influences on streamflow drought characteristics in England and Wales’
.
Hydrology and Earth System Sciences
22
:
1051
1064
. DOI:
66
Van Loon
,
AF
.
2015
.
‘Hydrological drought explained’
.
Wiley Interdisciplinary Reviews: Water
2
:
359
392
. DOI:
67
Van Loon
,
AF
,
Gleeson
,
T
,
Clark
,
J
,
Van Dijk
,
AIJM
,
Stahl
,
K
,
Hannaford
,
J
, Di
Baldassarre
,
G
,
Teuling
,
AJ
,
Tallaksen
,
LM
,
Uijlenhoet
,
R
,
Hannah
,
DM
,
Sheffield
,
J
,
Svoboda
,
M
,
Verbeiren
,
B
,
Wagener
,
T
,
Rangecroft
,
S
,
Wanders
,
N
and
Van Lanen
,
HAJ
.
2016
.
‘Drought in the Anthropocene’
.
Nature Geoscience
9
:
89
91
. DOI:
68
Van Loon
,
AF
and
Van Lanen
,
HAJ
.
2013
.
‘Making the distinction between water scarcity and drought using an observation-modelling framework’
.
Water Resources Research
49
:
1483
1502
. DOI:
69
Verdon-Kidd
,
DC
,
Scanlon
,
BR
,
Ren
,
T
and
Fernando
,
N
.
2017
.
‘A comparative study of historical droughts over Texas, USA and Murray-Darling Basin, Australia: Factors influencing initialization and cessation’
.
Global Planetary Change
149
:
123
138
. DOI:
70
Vidal
,
J
.
2017
.
Personal correspondence via email, from Hydrology-Hydraulics Research Unit, France
71
,
Y
,
van Beek
,
LPH
,
Wanders
,
N
and
Bierkens
,
MFP
.
2013
.
‘Human water consumption intensifies hydrological drought worldwide’
.
Environmental Research Letters
8
. DOI:
72
Walker
,
WE
,
Harremoes
,
P
,
Rotmans
,
J
,
van der Sluijs
,
JP
,
van Asselt
,
M
,
Janssen
,
P
and
Krayer von Krauss
,
MP
.
2003
.
‘Defining uncertainty: A conceptual basis for uncertainty management in model-based decision support’
.
Integrated Assessment
4
:
5
17
. DOI:
73
Wanders
,
N
,
,
Y
and
Van Lanen
,
HAJ
.
2015
.
‘Global hydrological droughts in the 21st century under a changing hydrological regime’
.
Earth System Dynamics
6
(
1
):
1
15
. DOI:
74
Xu
,
J
,
Ma
,
N
and
Lv
,
C
.
2016
.
‘Dynamic equilibrium strategy for drought emergency temporary water transfer and allocation management’
.
Journal of Hydrology
539
:
700
722
. DOI:
75
Yevjevich
,
V
.
1967
. ‘An objective approach to definitions and investigations of continental hydrologic droughts’.
:
.
76
Yuan
,
X
and
Wood
,
EF
.
2013
.
‘Multimodel seasonal forecasting of global drought onset’
.
Geophysical Research Letters
40
(
18
):
4900
905
. DOI:
77
Zhang
,
JZ
,
Ding
,
ZW
and
Luo
,
MT
2017
.
‘Risk analysis of water scarcity in artificial woodlands of semi-arid and arid China’
.
Land Use Policy
63
:
324
330
. DOI:
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License (CC-BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. See http://creativecommons.org/licenses/by/4.0/.