Small artificial impoundments have big implications for hydrology and freshwater biodiversity

Robert Morden, Water, Agriculture and Environment program, The University of Melbourne, robert.morden@unimelb.edu.au Avril C Horne, Water, Agriculture and Environment program, The University of Melbourne, avril.horne@unimelb.edu.au Nick Bond, Centre for Freshwater Ecosystems, La Trobe University, Wodonga, Victoria, Australia; N.Bond@latrobe.edu.au Rory Nathan, Water, Agriculture and Environment program, The University of Melbourne, rory.nathan@unimelb.edu.au Julian D. Olden, School of Aquatic and Fishery Sciences, University of Washington, Seattle, Washington, United States, olden@uw.edu

F o r R e v i e w O n l y more waterways when compared to major dams alone. Hydrological impacts are concentrated in 23 smaller streams (catchment area < 100 km 2 ), raising concerns that the often diverse and highly 24 endemic biota found in these systems may be under threat. Adjusting existing biodiversity planning 25 and management approaches to address the cumulative effects of many small and widely 26 distributed artificial impoundments presents a rapidly emerging challenge for ecologically 27 sustainable water management.

73
The scope of SAI impacts are challenging to characterize at a continental or global scale due to a lack 74 of data regarding their number and locations in many regions (Januchowski-Hartley et al. 2020). 75 Consequently, they are often ignored in investigations into the effects of flow alteration on 76 freshwater ecosystems, with research and policy attention instead focusing on large in-channel 77 structures and major extractions. In doing so, such studies make an implicit assumption that the 78 biggest ecological impacts arise from the largest individual extractions or impoundments, rather 79 than considering the totality of hydrological stresses in operation, including those associated with 80 the cumulative effects of SAIs.

81
This paper examines the relative role of SAIs and larger on-stream dams in causing hydrological 82 stress throughout a catchment, and the challenges associated with the management, and supporting 83 policy, of SAIs into the future. Impoundments of all types can affect upstream and downstream 84 biodiversity through multiple pathways, for example by altering habitat conditions ( waterway connectivity (Leitão et al. 2018; Barbarossa et al. 2020); here, we focus on the threat to 87 downstream biodiversity using a hydrological measure of the degree of impoundment. We look to 88 Australia and the United States to demonstrate how we continue to underestimate the risk posed to 89 global biodiversity from hydrological alteration, particularly in headwater streams, by continuing to 90 ignore the widespread, growing number and cumulative impact of SAIs.

93
The wide range of different terms for waterbodies distributed throughout catchments is a common 94 source of confusion (Biggs et al. 2017). Small natural impoundments are usually called 'ponds' or 95 'lakes', whereas small artificial impoundments are called 'farm ponds', 'farm storages', 'small 96 storages', 'tanks', 'stock ponds', or 'mill ponds' and are usually constructed with a low earthen bank 97 across a watercourse or landscape depression.

98
Local differences may also exist -in Australia small artificial impoundments are usually called 'farm 99 dams' (Nathan and Lowe 2012), but other terms such as 'floodplain storage', 'catchment dam' or 100 'runoff dam' are sometimes used to help identify the primary source of the water. In Europe, the 101 term 'small waterbodies' appears to be a more common label when referring to a wide range of 102 features such as storages, mill ponds, and ditches (Biggs et al. 2017).

103
In this paper we adopt the term 'small artificial impoundments' or 'SAIs' as it appears the most 104 precise and least ambiguous terminology. SAIs included in our analysis ranged over 400-fold in size 105 from as little as 250 m 2 up to more than 100,000 m 2 . In our case study, SAIs are typically constructed 106 for agricultural and livestock purposes, with a smaller number managed for hydropower, recreation, 107 aquaculture, or potable supply. Some examples of SAIs from around the world highlighting their 108 diversity of size and construction techniques are shown in Figure 1. 120 To understand the role of SAIs in contributing to hydrological stress throughout a catchment, the 121 DoR concept was applied to two case studies, the Murray Darling River Basin, Australia, and the 122 Arkansas River Basin, United States. These basins were selected as exemplars of the longstanding 123 challenges facing global rivers subjected to SAIs. The Murray Darling basin is the largest river basin in 124 Australia covering more than one million square kilometres, supplying drinking water to more than 125 three million people and generating roughly 40% of Australia's total agricultural production. The 126 Arkansas River basin, the second longest tributary of the Mississippi River, encompasses close to a 127 half million square kilometres, and supports substantial irrigated agricultural production. Similarly, in the Arkansas River basin, 3.5% of reaches by length are impacted by major on-stream 141 dams (Figure 2c), but when SAIs are included this proportion nearly triples to 9.7% (Figure 2d). SAIs 142 only represent 0.03% of total storage capacity, yet they increase the relative length of impacted 143 waterways by 280%.   In each scenario, streamflow from a single gauge location (above shows Mt Ida Creek, Victoria, Australia, gauge 406226, catchment area 174 km 2 ) was used as a hypothetical 'natural' flow, and the hydrological impact of impoundments was applied to this. The single large dam was set to capacity of 20% of mean annual flow (DoR = 20%) with an upstream watershed area 50% of the gauged catchment. The multiple small dams were set to capacity of 2500 m 3 each, with the same aggregate capacity and watershed area as the single large dam.
152x272mm (300 x 300 DPI) The Degree of Regulation (DoR) index is calculated for each reach in the network based on the 5 cumulative upstream storage relative to the cumulative average annual discharge, with units of 6 'years'. A DoR value of 0.5 therefore implies the total upstream storage volume is equivalent to 50% 7 the mean annual runoff, while a DoR of 3 implies 3 times (ie. 300%) the mean annual flow can be 8 captured or held in storages. Whilst having locally observed flow data is optimal for quantifying the 9 many different facets of flow alteration, DoR is still a strong surrogate at broader spatial scales 10 ( 13 This index requires input data to characterize the impoundment locations and capacities, the river 14 network, and the streamflow through the river network. Data sources for these key inputs are listed 15 in WebTable S1 for each of the case study catchments. 16

Impoundment information
17 Not all waterbodies were included in calculations. In both case study catchments, natural 18 waterbodies were excluded wherever they could be identified. Helpfully, the NHDPlus dataset 19 (Moore et al. 2019) includes a field "FCODE" which clearly identifies many types of waterbodies. This 20 field was used to specifically include only those features which were identified as a "reservoir" 21 (FCODE=43600), "reservoir for storing water" (FCODE=43613 to 43621), or "lake/pond" 22 (FCODE=39000 to 39012). Other features were excluded as either natural waterbodies, or artificial 23 waterbodies with no connection to natural drainage (eg. sewerage pondage, tailings, etc.).

24
In the Murray Darling basin, some large impoundments were excluded if they were known to be off-25 stream storages because their primary source of water is extraction from another storage or 26 waterway rather than runoff from their immediate upstream watershed. Also, SAIs were excluded 27 across large parts of the basin where the average slope of the surrounding terrain was 0.25% (1 in 28 400) or flatter. In such areas, surface runoff is very unlikely to reach a waterway in natural 29 circumstances, so small impoundments here are assumed to have no direct hydrological impact on a 30 waterway.

31
In the Arkansas River basin continuous areas with average slope flatter than 0.25% do exist, but they 32 are sufficiently small that filtering of SAIs was not considered necessary.

33
The slope threshold of 0.25% was selected based on two criteria: statistics and scenario information for each modelled location is presented in WebTable S2.

123
Using the STEDI software, extraction from each storage is also modelled. In all cases, the long term 124 average annual extraction was set equal to 50% of the dam capacity, with daily pattern of extraction 125 based on a rolling 2 week average of net evapotranspiration (Morton's actual evapotranspiration 126 minus rainfall). This was adopted as an approximation of water demands for irrigation.

127
Inflow for each modelled storage was based on the total natural flow for the catchment, adjusted 128 based on the simple ratio of total catchment area to the storage's upstream watershed. In other 129 words, flow was assumed to be generated uniformly across the catchment.

130
For each site and each scenario, the impact of storages was calculated on a daily basis for the period 131 from January 1980 to December 2014. WebFigure S3 compares the annual impacts on streamflows 132 for the single dam and multiple dam scenarios, as well as the impact on low flows.

133
WebFigure S3 demonstrates that the annual volumetric impacts due to a single large storage is the 134 same order of magnitude as for multiple SAIs, although in most catchments the impacts of SAIs tend 135 to be higher. The effects on percent of low flow days are the same for both scenarios. The combined surface areas of all SAIs was greater than the surface area of a single storage even though they had 137 the same overall capacity, which is an expected consequence of the typical geometry of artificial 138 impoundments. Higher rates of evaporation resulted in longer filling times for SAIs, which is the 139 most likely reason why the impacts of multiple SAIs are often slightly higher with greater variability 140 than single large storages.  basin. Note that features were excluded if their capacity was greater than 300,000 m 2 , their surface 7 area was greater than 1x10 6 m 3 , or their average depth was less than 0.3m. Features were also 8 excluded if their surface area (acres) was published as an integer less than 10. impounding 50% of the overall catchment area, and on the right multiple 2500 m 3 storages with 7 aggregate DoR = 20% are impounding 50% of the overall catchment area.   We are pleased to re-submit our manuscript FEE20-0366 "Small artificial impoundments have big implications for hydrology and freshwater biodiversity" (revised title) for your consideration. We greatly appreciate the positive comments from all reviewers and welcome the opportunity to improve the manuscript based on their constructive feedback. The reviewer's suggestions were very helpful to highlight specific areas where the paper required additional clarification.
Our responses to the reviewer's comments are outlined below and have been highlighted with tracked changes in the revised manuscript alongside numerous other improvements. Please note that where we have indicated line numbers in the manuscript, we are referring to the clean version without tracked changes.
We believe that the manuscript has improved significantly by addressing the reviewer's comments and look forward to hearing from you in due course. 2. In Figure 2, please enlarge all of the smallest text to improve readability. This primarily includes the green numbers and the numbers (especially superscripts) in the keys.
3. Please enlarge Figure 3 to have a resolution of at least 300 dpi at a width of 4.5 inches. Please also ensure that the text is legible at this size. When resubmitting this figure, please supply as a jpeg or tif file. Figure 4, please sharpen the text and ensure that it is legible, as it is quite small. When resubmitting this figure, please supply as a jpeg or tif file.

For
5. Please rename "Box 1" in the main text as "Panel 1".
Thank you for these editorial comments. These issues have been resolved in the latest submission.

Comments to the Author
This paper is very well-written, and the authors do an excellent job of highlighting an issue that many ignore: the cumulative ecological impacts of small artificial impoundments across watersheds. I generally agree with the positive responses from the two Reviewers, and I support Reviewer 1's mention that some limitations need to be discussed. This leads to: My only major comment is that more methodological information is needed in the supplemental material to support the paper's findings. For example, in order to publish Figure 3 and WebFigure S2, additional details regarding the model are required. Some information is provided in WebPanelS1 but not nearly enough to support the study's findings. Please see my specific comments below regarding WebPanel S1. Also, please mention at least once in the main text some limitations to the findings (or "challenges" -however the authors wish to contextualize them) because they are, indeed, model outputs.
We very much appreciate this feedback. WebPanel S1 in the revised manuscript now includes additional details of the hydrological modelling and its limitations, and further discusses how small artificial impoundments (SAIs) were identified and characterised in each study area. Our changes are further discussed below in response to the reviewer's comments.
Somewhere in the paper, it would be worth mentioning that there are impacts beyond biodiversity, including water quality. It can be a quick mention, to be sure, since the paper is about biodiversity impacts.
We agree with this comment and have now added text near line 83 in the revised manuscript to illustrate that the implications of small impoundments extend well beyond just biodiversity.
Also, be careful to ensure the figures can stand independently. To do this, please edit the captions so that abbreviations used in the figures are spelled out in the captions. For example, in Webfigure S1, spell out DoR in the caption. This is an excellent suggestion. Changes have been made to almost all figures either improving the caption title or clearly defining abbreviations, in order to ensure that they are stand-alone and do not dependent on the main body of the text.
Webfigure S3 -This figure is a bit confusing. It's supposed to represent a gradient of impacts freshwater species, but the gradient is across the landscape -where they don't exist. Also, the caption doesn't seem to match the legend in the figure. I suggest only showing the gradient along the streams themselves and matching the caption and figure legend.
This is an excellent comment. On reflection, this figure could have been presented more clearly. The intention is to visually demonstrate that "threatened freshwater species are not uniformly distributed across each study area". Accordingly, the caption has been revised to include those exact words, and the legends changed accordingly. To reduce confusion, the figure no longer shows 'impacted' waterways, it simply displays major rivers to allow readers to understand the catchment physical layout.
The numbers of species are shown as a gradient across the entire landscape. This is a deliberate representation for the purposes of clarity. Each study area has hundreds of thousands of individual reaches and minor streams, which appear visually messy when plotted at this scale. This landscape representation was chosen to be smoother and clearer. The caption has been adjusted to better explain this representation.
Webpanel S1 Calculating DoR, lines 4-7 -Please be explicit here by providing units on the variables. If total volumetric (and please add "volumetric" in front of "capacity") capacity of impoundments is L3, and the long-term average annual streamflow is L3 T-1, then the index would be in T? The figures suggest it's a percentage, not a time unit.
Correct, the units of this index are indeed 'years'. We have clarified this point in WebPanel S1 line 5.
Line 14 -Provide the average slope in % for better interpretability and add why that slope was chosen.
We agree that this number has been presented in a somewhat arbitrary manner. Further discussion of the basis for this part of the analysis have been provided in WebPanel S1 lines 33 to 44.
Lines 17-20 -What were the specific criteria for "some large waterbodies" being excluded? Please detail that here.
We are very grateful for this comment, it has highlighted an oversight in our original wording when we were considering another river basin in the United States. In fact, no large dams were required to be excluded based on the National Inventory of Dams. The relevant text has been removed.
The revised manuscript now clearly states that large dams in the Arkansas River basin were included/excluded based on data provided in the NHDPlus dataset. Details of this process have been provided in WebPanel S1 lines 18 to 23.
Line 32: Is 1.91 the assumed depth, and if so, this is in m? Also, how was this new relationship developed, meaning using what data? Please be explicit.
We agree that this equation was perhaps confusing as originally presented. More information has been provided in WebPanel S1 lines 51 to 75, plus a new WebFigure S1. As well as describing the development of this equation and its limitations, we have noted that Line 40+ -More information is needed regarding STEDI, the model that supports this study. It sounds like a simple water budget model with only the parameters mentioned on lines 45-47? Is this correct? If so, please include that equation and any other governing equations of the model -and is meant specifically by terms such as "climate effects on the surface of the water body". Also, how was the model parameterized, and what objective functions (or loss functions) were used to evaluate the outputs? How well did the model perform? This is of key importance to Figure 3. What are the primary uncertainties and limitations in using this model? Please add details here and a sentence or two to the main manuscript.
We agree that STEDI is central to some of the study's key conclusions, yet our original description of the model was perhaps not sufficiently detailed. Further description of the model and its core algorithm has now been fully articulated in WebPanel S1 lines 94 to 111. This also includes a clear statement of the key limitations of the model and the potential implications with respect to the major conclusions of our study in WebPanel S1 lines 143 to 146.
Webtable S1 -Spell out SAI in the caption.
This change has been made.

Reviewer 1 comments to author
The authors focus on downstream effects, but fish migration and thus migration barriers are twoway concerns. Also, degree of regulation is a poor surrogate measure of flow regime alterations (peak & low flow magnitudes, frequencies, timing & duration) (Poff et al. 2007 PNAS 104:5732-5737). These issues should be discussed as limitations of the degree of regulation measure as an estimate of impact.
Reviewer 1 is entirely correct that SAIs may potentially have impacts on biodiversity both upstream and downstream. The revised manuscript includes a brief discussion of this point near line 83.
Whilst we agree that no single metric can adequately describe the diverse ways in which dams can impact on downstream hydrology (e.g. cf. hydropower vs irrigation release strategies), the degree of regulation (DoR) metric has become a well accepted means of characterising the potential impact on downstream hydrology that can be mapped across diverse systems. This issue has been briefly mentioned near line 113, and again in WebPanel S1 line 8. Lines 86. Also called "tanks", "stock ponds", "mill ponds" Lines 114. supports-not "supporting" All the above minor changes have been made.
Lines 115. Define reach here. Is it segments (distance between tributaries or major geomorphic change) or sites (area above dams) or something else?
A definition of 'reach' has been added on line 128, and in more detail in WebPanel S1 line 80.
Lines 120-129. Somewhere, the authors should emphasize that small (first -third order) streams represent most (~80%) of the river/stream length of any river basin (Colvin et al. 2019).
We agree that this is an important issue, and perhaps most relevant to the section where we discuss how SAI effects on biodiversity tend to be biased toward smaller streams. Additional text as suggested has been added near line 164.
Lines 158-160. Summarize the Arkansas basin numbers here as well.
Results for the Arkansas River have been included in the revised manuscript.
Lines 306. Add a figure title.
All the above minor changes have been made.
Lines 311. Distinguish major dams from SAIs in both basins.
The intent was that the legend for panels (a) and (b) were also applicable to (c) and (d).
Based on this comment, the figure has been modified so that each panel has a separate legend.

Reviewer 2 comments to author
This is a well written manuscript. Indeed, small dams and other man-made structures been largely ignored in riverine ecosystems.
We are grateful for this comment. The authors are keenly aware of the hydrological issues associated with small impoundments, and we are pleased that our enthusiasm seems to have been conveyed in this case.  Global and continental studies consistently show major dams as a dominant sourcesources of 19 hydrological stress affectingthreatening biodiversity in the world's major rivers, but the cumulative 20 impact of veryimpacts from small waterbodies on downstream biodiversityartificial impoundments 21 concentrated in headwater streams hashave rarely been acknowledged. Using the Murray Darling 22 River basin (Australia) and the Arkansas River basin (USA) as case studies, we examine the 23 hydrological impact of small artificial waterbodies. Their scaleimpoundments. The extent of 24 impacttheir influence is very significant, hydrologically affecting between altering hydrology in 280% 25 and -380% more waterways than when compared to major dams alone. Their 26 hydrologicalHydrological impacts are biased towardconcentrated in smaller streams with (catchment 27 areas less thanarea < 100 km 2 , which can harbour ), raising concerns that the often diverse and 28 highly diverse communities of aquaticendemic biota not found in larger catchments .these systems 29 may be under threat. Adjusting existing biodiversity planning and management approaches to deal 30 with the diffuse nature of these waterbodies will be address the cumulative effects of many small 31 and widely distributed artificial impoundments presents a keyrapidly emerging challenge for the 32 futureecologically sustainable water management.   downstream biodiversity using a hydrological measure of the degree of impoundment. We look to 96 Australia and the United States to demonstrate how we continue to underestimate the risk posed to 97 global biodiversity from hydrological alteration, particularly in headwater streams, by continuing to 98 ignore the widespread and, growing number and cumulative impact of SAIs.

101
The wide range of different terms for waterbodies distributed throughout catchments is a common 102 source of confusion (Biggs et al. 2017). Small natural impoundments are usually called 'ponds' or 103 'lakes', whereas small artificial impoundments are called 'farm ponds', 'farm storages', or 'small 104 storages', 'tanks', 'stock ponds', or 'mill ponds' and are usually constructed with a low earthen bank 105 across a watercourse or landscape depression.

106
Local differences may also exist -in Australia small artificial impoundments are usually called 'farm 107 dams'(Nathan and Lowe 2012), but other terms such as 'floodplain storage', 'catchment dam' or 108 'runoff dam' are sometimes used to help identify the primary source of the water. In Europe, the 109 term 'small waterbodies' appears to be a more common label when referring to a wide range of 110 features such as storages, mill ponds, and ditches (Biggs et al. 2017).

111
In this paper we adopt the term "small artificial impoundments" or "SAIs" as it appears the most 112 precise and least ambiguous terminology. SAIs included in our analysis ranged over 400-fold in size 113 from as little as 250 m 2 up to more than 100,000 m 2 . In our case study, SAIs are typically constructed 114 for agricultural and livestock purposes, with a smaller number managed for hydropower, recreation,

130
To understand the role of SAIs in contributing to hydrological stress throughout a catchment, the 131 DoR concept was applied to two case studies, the Murray Darling basinRiver Basin, Australia, and the 132 Arkansas River basinBasin, United States. These basins were selected as exemplars of the 133 longstanding challenges facing global rivers subjected to SAIs. The Murray Darling basin is the largest 134 river basin in Australia covering more than one million square kilometres, supplying drinking water 135 to more than three million people and generating roughly 40% of Australia's total agricultural 136 production. The Arkansas River basin, the second longest tributary of the Mississippi River, 137 encompasses close to a half million square kilometres, and supportingsupports substantial irrigated 138 agricultural production.

139
The DoR was calculated for all reaches -defined as the segments between tributaries -in the river 140 network for both case study basins, in the first instance considering only major on-stream dams, and 141 then accounting for the additionpresence of SAIs. A threshold to identify impacted rivers is difficult 142 to estimate with any confidence. For comparative purposes, a DoR value of 16.7% has been adopted 143 based on a recent global study of the impact of large storages (Grill et al. 2019). See Supporting 144 Information for calculation methods.

145
Differences in estimates of degree of regulation are striking. In the Murray Darling River basinBasin, 146 when considering only major on-stream storages (Figure 2a) we find that around 10% of reaches by 147 length are flow impacted (Figure 2b). But when SAIs are included, the proportion of impacted 148 streams in the basin almost quadruples to 37%, with impacted streams represented across almost 149 the entire basin. SAIs only represent 7% of total storage capacity, yet their influence increases the 150 relative length of impacted waterways by 380% compared to the extent of impacts from large 151 storages. Similarly, in the Arkansas River basin, 3.5% of reaches by length are impacted by major on-152 stream dams (Figure 2c), but when SAIs are included this proportion nearly triples to 9.7% ( Figure  153 2d). SAIs only represent 0.03% of total storage capacity, yet they increase the relative length of 154 impacted waterways by 280%.

155
Climate is an important driver of the results reported here. Areas with mean annual rainfall higher 156 than approximately 1000 mm have sufficiently high rates of runoff that the DoR rarely exceeds 157 16.7% even with high levels of SAI development. Conversely, areas with less than around 400 mm 158 have such low runoff that even the presence of a small number of SAIs could produce aresults in 159 high estimates of DoR. However, these areas tend to have relatively low levels of SAI development, 160 most likely because a combination of low runoff and high evaporation make open water 161 impoundments impractical for most agricultural purposes.

162
Hydrological modelling was also used to showrevealed that the effects of SAIs on downstream flow 163 regimes are broadly similar to the effects of large dams. Figure 3 shows the results of the 164 hydrological modelling forUsing one Murray Darling River Basin site whereas an example, the effect 165 on downstream flow regime of a hypothetical large dam was compared to a large number of SAIs 166 with the same aggregate capacity and watershed. (Figure 3). The overall percentage reduction in 167 annual flow was somewhat higher for SAIs than for a single large storage, but the net effect on flow 168 exceedance and numbers of low flow days were very similar. Another four sites modelled in the 169 same way showed comparable results (see Supporting Information for modelling methods and  170 results for other catchments). In effect, if a large dam can be considered a source of flow regulation, 171 then collections of SAIs need tomust be consideredviewed as a form of 'distributed flow regulation'.

174
In both case study basins we found that SAIs primarily affect smaller and headwater streams, and 175 some instances these streams may have higher conservation priority because they support greater 176 numbers of threatened species than waterways affected by large dams alone. This is particularly 177 important, as first to third order streams make up to 80% of waterways in most basins (Colvin et al.

212
The dangers of cumulative impacts -. When many individual landowners construct new SAIs free 213 from any significant regulatory control, their individual impacts may be negligible but their 214 cumulative impacts can give rise to "the tyranny of small decisions" (Kahn 1966   percentiles. Note that in panels (b)  The Degree of Regulation (DoR) index is calculated for each reach in the network based on the 5 cumulative upstream storage relative to the cumulative average annual discharge, with units of 6 'years'. A DoR value of 0.5 therefore implies the total upstream storage volume is equivalent to 50% 7 the mean annual runoff, while a DoR of 3 implies 3 times (ie. 300%) the mean annual flow can be 8 captured or held in storages. Whilst having locally observed flow data is optimal for quantifying the 9 many different facets of flow alteration, DoR is still a strong surrogate at broader spatial scales 10 ( impoundments upstream of a given point in the river network is divided by the long term average 14 annual streamflow at the same given point. , which was the threshold adopted in the current study.

15
This index requires input data to characterize the impoundment locations and capacities, the river 16 network, and the streamflow through the river network. Data sources for these key inputs are listed 17 in WebTable S1 for each of the case study catchments. 18

Impoundment information
19 Not all waterbodies were included in calculations. In both case study catchments, natural 20 waterbodies were excluded wherever they could be identified. Helpfully, the NHDPlus dataset 21 (Moore et al. 2019) includes a field "FCODE" which clearly identifies many types of waterbodies. This 22 field was used to specifically include only those features which were identified as a "reservoir" 23 (FCODE=43600), "reservoir for storing water" (FCODE=43613 to 43621), or "lake/pond" 24 (FCODE=39000 to 39012). Other features were excluded as either natural waterbodies, or artificial 25 waterbodies with no connection to natural drainage (eg. sewerage pondage, tailings, etc.).

26
In the Murray Darling basin, some large waterbodiesimpoundments were excluded if they were 27 known to be off-stream storages because their primary source of water is extraction from another 28 storage or waterway rather than runoff from their immediate upstream watershed. Also, SAIs were 29 excluded across large parts of the basin where the average slope of the surrounding terrain was 30 0.25% (1 in 400) or flatter. In such areas, surface runoff is very unlikely to reach a waterway in 31 natural circumstances, so small impoundments here are assumed to have no direct hydrological 32 impact on a waterway.

33
In the Arkansas River basin continuous areas with average slope flatter than 0.25% do exist, but they 34 are sufficiently small that filtering of SAIs was not considered necessary.

35
The slope threshold of 0.25% was selected based on two criteria: waterway or a small fold in the landscape. This difference in construction technique 48 underscores obvious differences in hydrological connectivity. While there is no distinct 49 boundary between these two techniques, Inspection of detailed aerial imagery suggests that 50 a slope of 0.25% provides a reasonable lower bound of where the latter technique occurs.

51
A range of data sources was used to estimate the capacity of impoundments. In the Murray Darling 52 basin, capacities of major dams were assigned based on the published capacity in the Register of 53 Large

57
In the Arkansas River basin, capacities of major dams were assigned based on the published capacity 58 in the National Inventory of Dams (NID) (USACE 2019). For the majority of small waterbodies, 59 aimpoundments, volumetric capacities are not known. However, the NID does record the surface 60 area and capacity of some smaller impoundments in the study area. A new relationship between 61 surface area and capacity was developed based on the National Inventory of Dams using only dams 62 in the Arkansas River basin smaller than 300,000 m 2 or 1x10 6 m 3 , and where the average depth is 63 greater than 0.3m. Thisthis data. Some filtering of the NID was required to obtain a meaningful 64 relationship was then applied to all remaining SAIs.as follows:

65
 To ensure the relationship was applicable to smaller impoundments, only those with valid 66 surface area and capacity values smaller than 300,000 m 2 or 1x10 6 m 3 were included. 67  A small number of dams were found to have very shallow average depth, suggesting an 68 unusual structure such as a shallow flood control dam. Only those with average depth 69 greater than 0.3m were included. 70  The NID records surface areas in units of acres. In some cases, this value is sometimes 71 recorded as an integer, leading to significant rounding errors if the surface area is less than 72 10 acres. Dams with surface area recorded as an integer less than or equal to 10 acres were 73 excluded.

74
The surface area and volumetric capacity of all remaining features in the NID in the Arkansas River 75 basin are shown in WebFigure S1, leading to an empirical relationship as follows: WebFigure S1: Developing an empirical relationship between the surface area and volumetric 5 capacity of impoundments included in the National Inventory of Dams (NID) in the Arkansas River 6 basin. Note that features were excluded if their capacity was greater than 300,000 m 2 , their surface 7 area was greater than 1x10 6 m 3 , or their average depth was less than 0.3m. Features were also 8 excluded if their surface area (acres) was published as an integer less than 10. impounding 50% of the gaugedoverall catchment area, and on the right multiple 2500 m 3 storages 7 with aggregate DoR = 20% are impounding 50% of the gaugedoverall catchment area.