Distribution and significance of small, artificial water … and significance of small, artificial water bodies across the United States landscape S.V. Smith *, W.H. Renwick , J.D

The Science of the Total Environment 299(2002) 21–36

0048-9697/02/$ - see front matter� 2002 Elsevier Science B.V. All rights reserved.PII: S0048-9697Ž02.00222-X

Distribution and significance of small, artificial water bodiesacross the United States landscape

S.V. Smith *, W.H. Renwick , J.D. Bartley , R.W. Buddemeiera, b c c

Department of Oceanography, University of Hawaii, Honolulu, HI 96822, USAa

Department of Geography, Miami University, Oxford, OH 45056, USAb

Kansas Geological Survey, University of Kansas, Lawrence, KS 66047, USAc

Received 18 December 2001; accepted 10 May 2002

Abstract

At least 2.6 million small, artificial water bodies dot the landscape of the conterminous United States; most are inthe eastern half of the country. These features account for approximately 20% of the standing water area across theUnited States, and their impact on hydrology, sedimentology, geochemistry, and ecology is apparently large inproportion to their area. These features locally elevate evaporation, divert and delay downstream water flow, andmodify groundwater interactions. They apparently intercept about as much eroded soil as larger, better-documentedreservoirs. Estimated vertical accretion rates are much higher, hence, inferred sedimentary chemical reactions must bedifferent in the small features than in larger ones. Finally, these features substantially alter the characteristics ofaquatic habitats across the landscape.� 2002 Elsevier Science B.V. All rights reserved.

Keywords: Artificial water bodies; Sediment accumulation; Hydrology; Conterminous United States

1. Introduction

The extent and importance of large, artificialwater catchment reservoirs across the landscapeare increasingly appreciated. Graf(1999) used theNational Inventory of Dams(NID; Table 1) toconclude that;75 000 artificial dams across theUnited States impound an amount of water approx-imately equivalent to 1 year’s run-off from thecontinent. He identified dams as ‘«significant

*Corresponding author. Present address: CICESE, Departa-mento de Ecologia, Ensenada, Baja California, Mexico; USmailing address: P.O. Box 434844, San Diego, CA 92143,USA. Tel.: q52-646-174-5050; fax:q52-646-175-0545.

E-mail address: [email protected](S.V. Smith).

features of every river and watershed of thenation.’ These features significantly slow the ratesof transport of water and contained dissolved andparticulate materials from land to the sea; elevatewater loss to evaporation; alter rates, pathwaysand locations of chemical reactions in freshwater;and disrupt freshwater aquatic habitats by frag-menting water flow to the ocean(e.g. Dynesiusand Nilsson, 1994; Graf, 1999; Vorosmarty and¨ ¨Sahagian, 2000; St. Louis et al., 2000).

A particular effect of reservoirs is the enhancedtrapping of sediments carried by rivers towards theocean(Trimble and Bube, 1990). This trapping isdramatically illustrated by Meade et al.(1990) intheir analysis of sediment transport by United

22S.V.

Smith

etal.

/T

heScience

ofthe

TotalE

nvironment

299(2002)

21–36

Table 1Databases, www addresses, and data characteristics for water body datasets used in this analysis

Database Source URL Data characteristics, comments(abbreviation)

USGS National Land Cover image http:yylandcover.usgs.govy Gridded 30-m pixels for individual states of conterminous Unitedfiles (modified NLCD) nationallandcoveryhtml States; data nominally for ‘leaf-off’ period, 1992. Processing

discussed in detail in text.;2.6=10 discrete water bodies.6

US Census Bureau inland water http:yywww.esri.comydatayonliney Water bodies and double-line stream polygons on USGS 1:100,000bodies, ‘Topologically Integrated tigeryindex.html quadrangle maps(seeGeographic Encoding and http:yywww.census.govyftpypubygeoywwwyGARMyCh15GARM.pdf).Referencing’(as ArcView shape Data collated from county data.;75,000 discrete water bodies,files available from Environmental including reservoirs, lakes, and streams.Systems Research Institute(ESRI)(TIGER)

1998 National Inventory of Dams, http:yycrunch.tec.army.milynidy Tabulated geographic coordinates(points) for ;75,000 artificial damsmaintained by US Army Corps of webpagesynid.cfm across United States. Database designed for flood hazard assessment.Engineers text file(NID) Sizes of water bodies and catchments given. From this we used

;43,000 water bodies.Water features data layer from http:yynationalatlas.govy Polygons collated by USGS from hydrography layer of 1:2 000 000

National Atlas data clearinghouse index.html digital line graphs(DLGs). ;5000 discrete water bodies(excludingshape file(NA) rivers) for conterminous United States. Most are listed as ‘lakes’ but

most are apparently artificial reservoirs.US Geological Survey digital line http:yyedcwww.cr.usgs.govydocy Hydrography layer(polygons) of 336 US Geological Survey

graphs(DLG) edchomeyndcdbyndcdb.html 1:24 000 DLGs(7.5 minute quadrangles) were analyzed. Processingdetails similar to modified NLCD, as discussed in text. Extrapolationto conterminous United States yields;9 000 000 discrete waterbodies.

23S.V. Smith et al. / The Science of the Total Environment 299 (2002) 21–36

States rivers. In the case of the Colorado River,for example, impoundments have reduced sedi-ment delivery to the Gulf of California by;100-fold. Downstream effects can also be dramatic(Williams and Wolman, 1984). At the mouth ofthe Colorado, tidal scour and erosion have over-taken delta construction in the virtual absence ofsediment supply(Carriquiry and Sanchez, 1999).

The focus of these analyses has been on rela-tively large water bodies. These features apparentlydominate the area and volume of fresh waterstorage. In contrast, the role of ‘small water bodies’(loosely defined to have surface areas smaller thanapprox. 10 m) has been largely overlooked, in4 2

spite of their probable significance to sedimentand sedimentary carbon deposition(Mulhollandand Elwood, 1982; Ritchie, 1989; Dean and Gor-ham, 1998; Stallard, 1998; Smith et al., 2001).Specific sediment yield(sediment export from acatchment per unit of catchment area) and therelated variable, sediment delivery ratio(ratio ofsediment delivered to a catchment outlet to sedi-ment eroded within the basin) tend to decreasewith increasing basin area(e.g. Walling, 1983).Milliman and Syvitski(1992) observed that riverbasin sediment yield to the ocean decreases asbasin size increases. An explicit application of thisto catchments and reservoirs within the contermi-nous United States is given by Renwick(1996).

Here we estimate the distribution of number andarea of small water bodies across the conterminousUnited States. The majority of water bodies in thestudy appear to be artificial rather than natural, sothe results reflect the significance of anthropogenicalteration of the landscape as well as of the waterbodies in themselves. We also provide quantitativeexamples of the importance of these features.

2. Data sets and data analysis

Several data sets were used in this analysis(Table 1). Modifications of the original data aredescribed briefly in the table and further elaboratedbelow. The data were processed using ArcView3.2 (http:yywww.esri.comysoftwareyarcviewyindex.html). We used three available inventoriesof large water bodies: the National Atlas(NA),the Census Bureau’s TIGER data, and the National

Inventory of Dams(NID). The NA hydrographylayer is mapped at a scale of 1:2 000 000, andincludes;5000 discrete water bodies(excludingrivers) for the conterminous United States. Mostare listed as ‘lakes,’ but most are apparentlyartificial reservoirs. The TIGER dataset(1:100 000) includes lakes, reservoirs and rivers,but they are not distinguished as such. Approxi-mately 75 000 discrete water bodies are mapped.The NID, a database designed for flood hazardassessment, includes tabulated geographic coordi-nates(points) for ;75 000 artificial dams acrossthe United States. Shapes are not mapped, butareas and volumes of water bodies and catchmentareas are given for most features. We eliminatedfeatures that are obviously multiple dams on thesame water body; features for which catchmentsize or impoundment size was not available orcould not be estimated; features for which thegeographic coordinates were in error; and featuresin the States of Alaska and Hawaii. This reducedthe number of features to;43 000.

None of these data sets provides comprehensiveinformation on small water bodies. In general theyprovide a fairly complete inventory of features ofat least several hundred thousand square meters;they miss virtually all features smaller than;105

m . While we do not establish an absolute size2

boundary defining ‘small’ vs. ‘large’ water bodies,;10 m is a useful working boundary between4 2

the small and large water bodies.The primary information used for the compre-

hensive evaluation of small water bodies is amodified version of the US Geological SurveyNational Land Cover Data(modified NLCD). Thenationwide land-use, land-cover dataset consists ofgridded 30-m pixels for individual states of con-terminous United States; the data are nominallyfor the ‘leaf-off’ period of 1992. All water bodiesin the dataset were identified and converted topolygons of contiguous pixels. Touching polygonswere joined and treated as individual water bodies.

In order to enumerate discrete water bodies notbeing counted in other inventories, a 1-km bufferwas constructed around the large water bodies andrivers identified in the National Atlas(NA) data-base, and water features inside this buffer weresubtracted from the data. This buffer allowed for

24 S.V. Smith et al. / The Science of the Total Environment 299 (2002) 21–36

slight mismatches in the mapped locations offeatures. Further inspection identified linear fea-tures that were clearly streams; these features weredeleted. Finally, water bodies were deleted from a5-km buffer around the coastline and large riversin the ESRI ArcView data layers for the UnitedStates. This step eliminated large riverine featuressuch as floodplain swales that appear as impound-ed water bodies but are functionally parts of theriver systems, as well as some coastal wetlands.These data processing steps should lead to aconservative(low) estimate of the total number,distribution, and area of small water bodies.

One limitation of this procedure is that someextensive wetland areas such as the Everglades inFlorida are resolved as many small individualwater bodies, rather than as single features. Thisleads to an over-estimate of the number of waterbodies, hence an underestimate of average waterbody area, in these regions. For example,;41 000 individual water bodies are resolved in theFlorida Everglades by our technique. Overall, thiseffect seems fairly small. In total, approximately3% of water bodies in the modified NLCD dataare located in areas mapped as ‘swamp or marsh’in the NA data.

We also examined portions of the 1:24 000(7.5minute) US Geological Survey Digital Line Graph(DLG) coverage. While it would be desirable toundertake this analysis for the entire conterminousUnited States, most of the;55 000 quadranglesare not available in this form. We downloaded andanalyzed the hydrography layer from a sample of336 quadrangles distributed through the study area.The distribution of available 7.5 minute DLGs isnot random, but apparently reflects local priorities.Our sample does include quadrangles from eachof the 48 conterminous United States, and webelieve is reasonably representative of the overallstudy area. The DLG data were masked using thesame procedures as used for the modified NLCDdata, in order to eliminate riverine and coastalfeatures and to make these two datasetscomparable.

The remaining features in the modified NLCDand DLG layers were mapped as numbers andareas of discrete water body features in each ofthe ;2100 US Geological Survey eight-digit

hydrologic cataloging units(HUC-8) across theconterminous United States (http:yywater.usgs.govyGISyhuc.html) (Seaber et al.,1987). These modified datasets represent the smallwater bodies only, and are referred to hereafter asthe modified NLCD and DLG data used in thisanalysis.

3. Results: distribution and abundance of waterbodies

3.1. Overall data characteristics

Results are summarized in Fig. 1 and Tables 2and 3. The modified NLCD dataset contains;2.6=10 water bodies. While the nominal res-6

olution of this coverage is 30=30 m pixels(i.e.900 m ), the smallest features are resolved by2

ArcView as triangular shapes with a calculatedarea of ;600 m . Almost half (43%) of the2

features are-10 m (i.e. essentially the limit of3 2

resolution); 90% are-10 m .4 2

The DLG data do not provide comprehensivecoverage across the entire United States, but theydo provide higher resolution of features than themodified NLCD. Based on the size distributionsof the water features, we estimate that the nominallower size limit on these features is;5 m (25m ). There is substantial variation in the lower2

size limits of mapped features from one DLG toanother, not surprising inasmuch as the originalmaps were prepared over a 40-year time span(1950s to 1990s). Extrapolating from the quadran-gles examined, there could be as many as 9=106

water bodies)25 m in area across the conter-2

minous United States.The US Census Bureau TIGER data provide the

most comprehensive and detailed digital coverageof water bodies other than the modified NLCDdata. This dataset includes approximately 75 000water bodies in the conterminous United States.However, in marked contrast with the modifiedNLCD and DLG databases, fewer than 1% of theTIGER features are-10 m in area, and only3 2

;6% are-10 m .4 2

The National Inventory of Dams(NID) dataset,as we modified it, has approximately 43 000 fea-tures in the conterminous United States; features


Fig. 1. Comparisons of numbers and areas of water bodies among NLCD, NA, TIGER and NID data sets.(a) Total numbers ofwater bodies identified in the modified NLCDqNA data (bars), compared with cumulative percentages by number for each dataset. No data set other than the modified NLCD accounts for a significant fraction of the total number of water bodies.(b) Areasof TIGER water bodies(bars), compared with cumulative percentages by area for each data set. The modified NLCD(uncountedin the other surveys) accounts for approximately 20% of the total water area.

Table 2Number, total area, mean area, and minimum reported area of water bodies in various data sets

Data set Number of Total surface Mean water Minimum waterwater bodies area body area body area(thousands) (thousand km)2 (thousand m)2 (m )2

Mod. NLCD 2600 21 8 600NID 43 62 140 80NA 5 89 1700 120 000TIGERa 75a 107 143 –b

DLG 9000 – – 25

Includes streams.a

Many of the smallest features are slivers; 94% of the features in the TIGER data are)10 m in area.b 4 2


Table 3Number and area of water bodies in the modified NLCD data set by two-digit hydrologic region

HUC-2 Name HUC-2 Water body Percent Number of Waterarea area water water bodies bodies(thousand km)2 (km )2 (thousands) per km2

01 New England 160 1240 0.78 66 0.4102 Mid Atlantic 278 1048 0.38 101 0.3603 South Atlantic-Gulf 711 3011 0.42 448 0.6304 Great Lakes 305 1802 0.59 143 0.4705 Ohio River 423 690 0.16 130 0.3106 Tennessee River 106 95 0.09 21 0.2007 Upper Mississippi 493 2672 0.54 184 0.3708 Lower Mississippi 265 1445 0.55 232 0.8809 Souris-Red-Rainy 151 1113 0.74 54 0.3610 Missouri River 1324 2893 0.22 459 0.3511 Arkansas-White-Red 643 1910 0.30 368 0.5712 Texas-Gulf 471 1368 0.29 214 0.4513 Rio Grande 342 104 0.03 9 0.0314 Upper Colorado 288 142 0.05 11 0.0415 Lower Colorado 375 80 0.02 5 0.0116 Great Basin 355 82 0.02 7 0.0217 Pacific Northwest 714 672 0.09 81 0.1118 California 420 373 0.09 27 0.06

Total 7824 20 740 0.27 2605 0.33

Fig. 2. Map of impoundment density(number km ), mapped by HUC-8 cataloging unit. The outline box is the 32–41 degreey2

transect with data summarized in Fig. 3.

are listed in that database specifically because theyrepresent potential flood hazards. While some fea-tures as small as approximately 80 m are included2

in the NID, only approximately 21% are-104

m in area.2

The NA database, which has the lowest resolu-


tion of the databases examined but which repre-sents the most widely available database of waterbodies across the United States, has only;5000features in the conterminous United States, thesmallest of which is;120 000 m .2

The pattern that emerges from this comparisonis that small water bodies are, numerically, over-whelmingly dominant across the conterminousUnited States; the available datasets differ widelyin representing those features. The modified NLCDdataset records 35 and 60 times as many featuresas the TIGER and NID databases, respectively,and ;500 times as many features as the NAdatabase. The DLG dataset apparently recordsmore than three times as many features as themodified NLCD. The estimated average waterbody density across the United States area of7.8=10 km , based on the satellite-derived mod-6 2

ified NLCD data, is 0.33 water bodies km . Ify2

these features were evenly distributed across theUS landscape, this would be equivalent to anaverage net catchment area(excluding areas trib-utary to upstream impoundments) of 3 km . If the2

sampled DLG map data are representative, thedensity may be)1.0 water bodies km (averagey2

catchment-1 km ).2

It has long been recognized that the lengths ofcoastlines and other complex geographic bounda-ries are dependent upon the scale at which thecalculations are made(e.g. Mandelbrot, 1967). Itcan be demonstrated that estimates of area are lesssusceptible to measurement scale than estimates ofboundary length. The topological analogy is evi-dent in the case of water body number, but notarea. A gross difference in the estimated numberof small water bodies across the landscape appar-ently makes relatively little difference in the esti-mated area.

3.2. Spatial patterns of water body distribution

3.2.1. Number of water bodiesFig. 2 is a map of ‘impoundment density’ across

the conterminous United States, based on themodified NLCD data. Densities range from-0.03water bodies km in much of the south-west toy2

)1 in the mid-west. In the more familiar notation

of catchment areas, these would range from aver-age catchment areas of-1 km in the mid-west2

to )30 km in the south-west.2

The highest concentration of water bodies is inagricultural regions, especially eastern portions ofthe Great Plains and the lower Mississippi Valley.On a state-by-state basis Texas leads the list, withapproximately 10% of the total. Other restrictedareas of high density occur in the glacial terrainof northern Minnesota, Michigan and New Eng-land, and in the wetlands of southern Louisiana,southern Georgia, and Florida. If the data areexamined by a two-digit hydrologic region(HUC-2; Table 3), impoundment density varies from;0.9 water bodies km in Region 08(Lowery2

Mississippi) to 0.01 km in Region 15(Lowery2

Colorado).There are clear east-to-west gradations in these

features. In order to gain insight into the distribu-tion, a 9-degree(latitude) by 34-degree(longi-tude) transect is presented for much of theconterminous United States(Fig. 3a). This transecttrims hydrological complications associated withextensive wetlands near the coasts and in thenorthernmost portion of the country. Except forthe DLG data, which are insufficient for suchdetailed analysis, the transect data are presented as1-degree longitudinal averages(each longitudinalstrip being;90 000 km in area). The more sparse2

DLG data are expressed as 2-degree averages. Thistransect incorporates approximately 40% of thearea of the conterminous United States.

Although differing in magnitude, the databasesshow much higher numbers of water bodies in theeastern half of the United States than in the west.Most of the east-to-west decrease occurs between958 and 1038 W, approximately the width ofKansas and other states in that north-south tier. Tothe east of 958 W, the modified NLCD and DLGdata give comparable estimates of impoundmentdensity(;1 km ). The DLG densities are slight-y2

ly higher, consistent with their greater resolution(25 vs. 900 m). To the west, the estimates diverge,2

as the DLG estimate decreases to 0.1 km andy2

the modified NLCD estimate decreases to 0.01.The divergence probably largely arises because themodified NLCD estimates represent estimates of


Fig. 3. Summary of water body numbers in 32–41 degree transect(Fig. 2). (a) Average number of water bodies per km in each2

longitudinal strip.(b) Percent of total area covered by water as identified from each database.(c) Water balance(precipitationminus potential evapotranspiration). (d) Large water body sediment yields(NID) compared to two estimates of small water bodysediment yields in the same longitudinal strip.

actual water as identified from satellite images,while the DLG coverage includes topographic lowsthat may contain water only on an ephemeralbasis. If this explanation is generally correct, thenanalysis of the DLG data overestimates the water

body abundance. The NID and TIGER data are inclose agreement with one another on water bodynumbers, but well below the modified NLCD. TheNA data give yet lower numbers and are notshown. Based on the TIGER and modified NLCD


numbers, the number of small water bodiesexceeds the number of large water bodies by afactor of;70 along the transect.

3.2.2. Area of water bodiesWhen water surface areas are examined(Table

2; Fig. 3b), the results are quite different. Largewater bodies account for most of the water areaacross the transect. The NID represents artificialreservoirs(or lakes that have been significantlymanipulated artificially), while TIGER representsany water body large enough to have been mappedat a scale of 1:100 000. The agreement in totalarea between these two data sets—one that isentirely artificial and the other that is natural plusartificial—is persuasive evidence that unalterednatural lakes other than the Great Lakes accountfor only a small percentage of water area acrossthe conterminous United States. The NA data(notshown) record a roughly similar area as NID andTIGER, despite having far fewer features. Aninteresting characteristic of the NA data is thatmost (;90%) of its features are identified as‘lakes,’ apparently because of their proper names(Lake Powell, Franklin D. Roosevelt Lake, etc.)even though the NID—TIGER comparison wouldsuggest that most of these ‘lakes’ are artificial.The only significant large natural lake on thetransect is the Great Salt Lake, which is a promi-nent spike in the TIGER area(but not in the NID)at 1128 W. The NID and TIGER data each showthe transect area to be approximately 1.0% water.For comparison, the modified NLCD small waterbodies have an area of approximately 0.2% of thetransect, or;17% of the total water area.

4. Discussion: importance of small water bodies

4.1. Origin of small water bodies

The dataset we use in this analysis does notallow explicit identification of small water bodiesas natural or artificial. Natural lakes and pondscan be found throughout the conterminous UnitedStates. The majority of natural lakes are almostcertainly of glacial origin. Numerous lakes are alsofound in karst landscapes such as central Florida.

Lakes of aeolian origin occur in many areas,including the high plains of Texas and New Mex-ico and the southeastern coastal plain. Oxbowsand swales occur on the floodplains of large rivers,although most of these have been eliminated fromour dataset. Lakes of tectonic origins(most ofthem dry or ephemeral) are common in the westernUnited States.

Nonetheless, it is evident from the geographicdistribution of water bodies that most of the smallwater bodies we have identified are of humanorigin. One argument is based on land use. Acrossagricultural areas, thousands of water bodies dotthe landscape where there were apparently virtuallyno such features a century or less ago. The averagedensity of water bodies in Oklahoma(0.88 km )y2

is nearly double that of Minnesota(0.46 km ),y2

although on a percent area basis Minnesota hasmore than double the water of Oklahoma.

Various citations (Holland, 1971; OklahomaWater Resources Board, 1990) and numerous doc-uments found on the World Wide Web provideestimates of the numbers of artificial farm pondsin various States(AL, GA, IA, IN, KS, MO, OK,TN, TX). For these nine states, the ‘published’total number of constructed farm ponds is approx-imately 1.9 million; our estimate of water bodiesfor these states is 1.1 million(i.e. 42% of thefeatures mapped in the modified NLCD data).With two prominent and puzzling exceptions(MO:500 000 published vs. 137 000 NLCD; TN:190 000 published vs. 35 000 NLCD), the agree-ment between the published number of farm pondsand our estimate for total small water bodies isgood(1.2 million published vs. 0.9 million NLCD)(r s0.45 including MO, TN;r s0.87 excluding2 2

those states). Various web publications give thestatewide addition of new ponds as 1–3% per year.Due to differing and unstated criteria for thepublished estimates of farm pond numbers, as wellas differing and unstated dates for when thoseestimates were made, some of this evidence mightbe considered anecdotal; however, the agreementis adequate to conclude that most of the smallwater bodies we have recorded are artificial, andthat their abundance is increasing dramatically. As


Fig. 4. Examples of water bodies in selected USGS 1:24 000 quadrangles. Comparison of DLG, modified NLCD, and TIGER waterbodies.(a) Wasco, CA.(b) Bend, SD.(c) Basehor, KS.(d) Russ, MO.(e) Crossville, TN.(f) Peacham, VT.


such, these features constitute a fundamentalhuman transformation of the landscape.

Selected examples of water bodies in variousenvironments are shown in Fig. 4. In every casemany features appear in the DLG data that do notappear in the NLCD or TIGER coverages. Insemiarid regions such as California and SouthDakota (Fig. 4a,b), surface water as detected inthe NLCD layer is less extensive. The area shownin Fig. 4a is a flat, agricultural landscape in theSan Joaquin Valley. The water bodies appear to beholding ponds for irrigation water. Only a few ofthem show up in the modified NLCD coverage;perhaps some have filled in or were dry at thetime the satellite data were collected. ArcView’sconversion of single-pixel water bodies to trian-gular units is clearly visible. The mid-west isrepresented in Fig. 4c,d. In both of these, thelandscape is rolling and agricultural. Virtually allthe water bodies shown are small ponds construct-ed for soil conservation andyor agricultural watersupply purposes. Many of those in this image aretoo small to show up on the modified NLCDcoverage, or were dry or have subsequently filledin. The Gasconade River visible in the TIGERdata at the right-hand portion of Fig. 4c is anexample of a river that is included as a water bodyin the total listed in Table 2. Two small reservoirslarge enough to appear in TIGER data are visiblein Fig. 4e, an example from Tennessee, while Fig.4f is a glaciated landscape in northern Vermont.The NLCD layer identifies a few small, apparentlynatural, water bodies. The larger of these alsoappear in the TIGER coverage. Most of the waterbodies in Fig. 4f that appear on the DLGs but noton the modified NLCD layer are vegetatedwetlands.

4.2. Hydrological significance of small waterbodies

Much of the explanation for both the area andnumber distribution of water bodies can be derivedfrom simple hydrological considerations. A tran-sect of precipitation minus potential evapotranspir-ation (P-PET) (Fig. 3c) demonstrates that themid-continent peak in numbers and then the west-ward decline in both numbers and area occur along

the transition from positive to negative waterbalance. The large number and area of the waterbodies in the mid-continent in large part representlocal human attempts to compensate for natural‘loss’ of water. This management strategy maywork at the site of the impoundment; waterremains locally available. However, the larger-scaleaggregate effect of this local water trapping willbe to elevate evaporation, rather than allowing thiswater to flow downstream or percolate.

Evaporation is commonly parameterized as afunction of wind speed(u, mys), the water vaporpressure of the surface water and the overlying air(e , e ; expressed here in mbar), and a massw a

transfer coefficient(N) (Sverdrup et al., 1942;Hutchinson, 1957; Harbeck, 1962). It is useful toexpresse explicitly in terms of saturation vapora

pressure at the air temperature and the relativehumidity (h). Harbeck (1962) further observedthat N is a function of the areaA (m ) of the2

water body in question. Converting Harbeck’sformulations to metric units yields the following:

y0.05 Ž .Es0.30A u e yhe (1)w as

Specifically, the product 0.3(A) in Eq. (1)y0.05

equals the transfer coefficient,N. Evaporation goesto 0 as the quantity(e yhe ) goes to 0.e andw as o

e are a function of temperature; in the presenceas

of any wind, evaporation will occur if either atemperature differential or relative humidity below1 drivese )he .w as

Harbeck(1962) attributed the decrease of evap-oration as a function of size in large part tochanging surface roughness of the water body. Asecond consideration that may particularly elevateevaporation in small water bodies is higher sum-mer heating in shallow systems.

The smallest systems considered by Harbeckwere approximately 4000 m . The median size of2

the modified NLCD water bodies in this study is;1000 m , and the mean is approximately 80002

m . The mean of the water bodies in the TIGER2

and NID databases is approximately 140 000 m2

(Table 2, Fig. 1). N for these systems would rangefrom 0.21 (NLCD median) up to 0.16 (TIGERmean). Large reservoirs()10 m ) would have7 2

N-0.13. Based on Eq.(1) evaporation from small


Table 4Results of regression analyses of sediment yields derived fromDendy and Champion(1978), by two-digit hydrologic region

HUC-2 N R2 Slope(b) Intercept(a) P

01 8 0.39 y0.80 2.42 0.1002 47 0.01 y0.06 1.99 0.5503 55 0.12 y0.15 2.57 0.0104 47 0.06 y0.18 2.10 0.1005 157 0.17 y0.18 2.57 0.0006 41 0.05 y0.06 2.51 0.1607 109 0.04 y0.08 2.62 0.0508 38 0.59 y0.30 3.18 0.0009 18 0.02 y0.06 1.63 0.5710 290 0.23 y0.25 2.67 0.0011 218 0.02 y0.06 2.51 0.0312 115 0.02 y0.06 2.52 0.1313 51 0.10 y0.17 2.51 0.0214 35 0.24 y0.78 2.03 0.0015 74 0.02 y0.06 1.93 0.2716 21 0.01 y0.06 2.13 0.6617 101 0.09 y0.29 1.60 0.0018 196 0.05 y0.15 2.76 0.00

The coefficient,a, represents the average yield(m km3 y2

year ) for a catchment with a drainage area of 1 km . They1 2

exponent,b, describes the change in yield as a function ofcatchment size. In all of the HUC-2 regions the exponentb isnegative, implying that specific yield decreases as a functionof catchment area. Ifa is constant over any area, then decreas-ing yield with increasing area implies that an increasingamount of eroded sediment is retained outside of the receivingwater body within progressively larger catchments.

water bodies will be decidedly elevated above thatfrom large water bodies.

These small water bodies will induce otherhydrological effects as well. For example, perco-lation and groundwater recharge will be affected,in ways reflecting the local hydrology. While wehave not yet undertaken detailed continent-scaleanalysis of the locations of these water bodies inrelation to drainage networks, analysis in selectedlocalities suggests that they are located primarilyon small streams with small catchment areas. Forexample, 80 ponds inventoried in two studies insouthwestern Ohio have an average surface areaof 2800 m and average catchment areas of 0.132

km (L. Theis, unpublished data; Hayes-Bohanan,2

1989). The position of these water bodies inupland catchments with small drainage areas hasthe effect of increasing the residence time of waterin upstream areas relative to downstream areas. Inareas of relatively permeable subsurface materials,this may result in increased groundwater rechargein upland areas. A similar analysis of position hasnot been carried out in the transitional and water-deficit areas to the west of the(water-excess) Ohiolocation, so we do not know how generalizablethe observation will be.

4.3. Sedimentological significance of small waterbodies

Sediment yield provides an index of the relativeimportance of the small and large water bodies.Dendy and Champion(1978) tabulated the sedi-ment accumulation rates in;1600 impoundmentsacross the United States. The drainage areas forthe impoundments in their survey range from-0.02 km to)400 000 km . Approximately half2 2

of these impoundments had drainage areas-6km (Fig. 5 in Dendy and Champion, 1978), thus2

are similar in magnitude to catchment areas in themodified NLCD. We used these data to calculatesediment yield(m km year ft km year )3 y2 y1 y2 y1

as a function of catchment area(km ) for each of2

the 18 HUC-2 regions across the conterminousUnited States(Table 4). The equations are of theform:

bYieldsa=Area (2)

In simplified notation for this paper, we assigntotal erosion products(E ) into three categories:T

(1) products than can be assigned to inventoriedwater bodies with explicitly known catchmentareas(E ); (2) products that are trapped in theNID

heretofore un-inventoried small water bodies dis-cussed in this paper(E ); and products thatNLCD

are deposited in other alluvial and colluvial depos-its across the landscape(E ). Thus:O

E sE qE qE (3)T NID NLCD O

We analyze two components of these products,E and E . The third component,E , isNID NLCD O

characterized by difference(see also Smith et al.,2001).

Tabulated catchment areas of the individual NIDwater bodies are used to calculateE accordingNID

to the regression equations. From consideration ofwater area(above and Fig. 3b), these calculations


represent sediment accumulation in the majorityof water area across the conterminous UnitedStates. The sediment yield to NID reservoirs alongthe transect is 317"447 t km yeary2 y1

(mean"standard deviation of data for the 1-degreelongitudinal strips on the transect) (Fig. 2d).Scaled to the water surface areas of the reservoirs,the vertical accretion rate is 4"4 cm year .y1

The NID reservoir inventory includes only afew of the small water bodies recorded in themodified NLCD, because the larger NID waterbodies have been removed from the modifiedNLCD. That modified dataset has)50 times asmany features as NID along the transect. We cantherefore calculateE with essentially no dupli-NLCD

cation of water bodies withE . This use of theNID

regression equations is appropriate, because theyare based on the gross—not net—catchment areasupstream of the impoundments. The regressionequations are applied to the estimated mean catch-ment size associated with groups of impound-ments, the inverse of the modified NLCDimpoundment density. Initially, impoundment den-sity is based on the number of impoundments ineach of the HUC-8 cataloging units. This providesan estimate ofE as an addition toE forNLCD NID

each of the HUC-8 areas. This estimated yieldalong the transect(308"239 t km year ) isy2 y1

almost identical to the NID-based estimate. As thewater area of the modified NLCD impoundmentsis much smaller than the water-body area of theNID data, the estimated vertical accretion rate inthe small water bodies is faster(22"19 cmyear ).y1

This calculation based on the HUC-8 units isan overestimate ofE , because it assumes theNLCD

modified NLCD features are evenly distributedwithin these units. If, instead, the modified NLCDfeatures are heterogeneously distributed within theHUC-8 units, then areas with locally high waterbody density would have high local yield, but theentire cataloging unit(the summation of high-density areas and low-density areas) would becharacterized by lower total sediment accumula-tion. In order to evaluate the degree of this over-estimate, the modified NLCD was re-sampled tocalculate impoundment density within 7.5 minutequadrangles maps. Because there are approximate-

ly 55 000 7.5 minute quadrangles across the con-terminous United States, in comparison with 2100HUC-8 units, this represents)25-fold decrease inthe average aggregation area for calculatingimpoundment density. In effect, this analysisimposes the homogeneity constraint for calculatingimpoundment density onto much smaller areasthan the HUC-8 units. Along the transect,;21%of the quadrangles have no NLCD water bodies.This recalculation lowers estimated sediment accu-mulation into modified NLCD water bodies byonly approximately 10%(from 308 to 278"218 tkm year ). We conclude that explicit resolutiony2 y1

of catchment areas for NLCD feature would pro-duce only modest decreases in estimated yield. Inspecific local studies for which the small waterbodies are mapped onto the local topography, theyield equations could be applied explicitly to thecatchment areas of the modified NLCD features,but that is presently impractical for the entireUnited States.

The conclusion that there is similar sedimentyield from the landscape into the NID and modi-fied NLCD water bodies therefore appears robust.While the results might be different if the expo-nential slopes in Table 4 were varied dramatically,the most important point to emphasize here is thatnegative slopes force the conclusion that yielddecreases as a function of catchment size. Exami-nation of more intensive ‘gray literature data’ forspecific locations(Ohio and Kansas), as well asthe summary by Walling(1983) and the basin-scale analyses by Milliman and Syvitski(1992)emphasizes that this general negative log–log rela-tionship is widespread in the United States, if notnecessarily universal. We use the Dendy andChampion summary, because it is the most com-prehensive, readily available summary for theUnited States. The conclusion is that approximate-ly 600 t km year is being trapped along they2 y1

transect in large plus small water bodies. Approx-imately half is being trapped in small water bodiesnot counted in standard inventories. Contemporaryerosion across the United States is estimated to be;900 t km year (Smith et al., 2001). There-y2 y1

fore accumulation in small plus large water bodiesapparently accounts for approximately two-thirdsof total erosion products, with small and large


water bodies accounting for roughly equal amountsof accumulation. The remaining third would beassigned toE according to this continental-scaleO

analysis.

4.4. Geochemical and ecological significance ofsmall water bodies

Both the difference in sedimentation ratebetween the large and small water bodies(dis-cussed above), and the effects of small waterbodies in changing the residence times and distri-butions of water and erosion products(sediments)within a watershed, have further ramifications. Therapid build-up of the water-body sediments(pri-marily eroded topsoil) not only buries organicmatter deeply in topographic depressions(whichwill be less subject to exhumation and erosioneven when ponds are no longer present), but alsocauses a systematic shift in redox conditions inthe deeper sediment that substantially slows oxi-dation of organic matter(Schlesinger, 1997; pp.244, 253–254). The rapid sediment burial in thesmall water bodies(E ), in combination withNLCD

their typical proximity to human and agriculturalsources of nutrient loading, will result in eutroph-ication and high organic input causing suboxic oranoxic conditions and diagenetic reactions whichare very different than in larger water bodies withslower sedimentation(E ).NID

The effect of burial is reinforced by hydrologicconditions. Rice(2002) observed that, with regardto carbon sequestration, ‘Soil water content also isimportant. Optimal microbial activity occurs at ornear field capacity-the maximum amount of waterthat soil can hold against gravity. As soil becomeswaterlogged, decomposition slows and becomesless complete«Decomposition also slows as soilsdry.’

The effect of artificial ponds is to maintain therapidly accumulating sediments in a permanentlysaturated condition. Even when ponds are dredgedto extend their lifetimes, practices in the centralpart of the USA are to use the spoil to extend theearthen dam creating the pond, or to leave theexcavated sediments piled nearby(R. Sleezer,personal communication), which creates a drierenvironment than that of undisturbed soil. Either

fate retards the oxidation of organic matter.Decomposition of organic matter will be greatlyslowed and organic preservation will be enhancedunder the combined conditions found in small,artificial water bodies. In addition to carbon cycleeffects, small water bodies may also be expectedto have significant effects on local- and landscape-scale budgets and fluxes of nitrogen andphosphorus.

Ecological and biological effects of small waterbodies are also important. The closest naturalanalog to the density and spacing of pond in thecentral USA is probably the ‘prairie pothole’ envi-ronment, which Sorenson et al.(1998) found tobe both critical habitat and highly sensitive toclimate variation. While artificial ponds will notfill exactly the same roles for a variety of reasons,the existence of a high spatial density of permanentaquatic microenvironments across the landscapewill be important to the survival, migration, andrange extension of a wide variety of species, bothnatural and invasive.

Our findings pointing to the addition of millionsof artificial water bodies with a total area of tensof thousands of km can also be viewed in the2

context of the loss of natural wetlands; Mitsch andGosselink(1993) report that the 48 conterminousstates of the USA have lost approximately 500 000km of wetlands. Of the wetlands remaining, of2

over half are forested or salt-water wetlands, whichare in general not the types of water bodiesaddressed in this study(Dahl et al., 1991).

Although there is a large difference between thewetland areas lost and the artificial water bodiesadded, we suggest that there are two aspects ofthe comparison that call for further study:(1) thedisparity of numbers of water bodies lost andgained will not be as great as the difference inarea, and the result of human intervention may bea more uniform distribution of smaller but morereliably saturated aquatic microenvironments thanthe pre-existing natural situation; and(2) there areprofound qualitative differences between naturalwetlands and artificial ponds in terms of not onlysize, but cover(open water vs. marsh vegetation),seasonal and interannual variability, placement onthe landscape, and proximity and response toadjacent human land use. We expect that these


water bodies have significant environmentalimpacts on the biota. The presence of large num-bers of constructed water bodies in landscapes thatotherwise had few or no lakes or wetlands consti-tutes a qualitative change in the environment withmajor potential ecological consequences.

5. Conclusions

Human influence on hydrological and sedimen-tological processes across the landscape is wellrecognized at large scales, a point that has beenmade by various authors(e.g. Dynesius and Nils-son, 1994; Graf, 1999; Vorosmarty and Sahagian,¨ ¨2000; St. Louis et al., 2000). We have demonstrat-ed that this human influence is quantitativelyimportant well below the scales of the large,inventoried features. Local water balance and sed-imentation are affected out of proportion with thearea and volume of the small features, and geo-chemical and ecological impacts of small waterbodies differ both quantitatively and qualitativelyfrom the impacts of large water bodies.

These distributed effects are difficult to dealwith at large (regional) scales by using routingmodels that are targeted to pinpoint specific fea-tures at specific locations. As the features areephemeral on time scales of years to a very fewdecades, are too numerous to be explicitly inven-toried, and are changing(largely increasing) innumbers, it is not apparent that explicit routingmodels are the best way to quantify their impacts.Yet we have demonstrated that the featuresareimportant. It therefore follows that, models, inven-tories and other analyses at these larger scalesneed to incorporate the aggregate effects of thesefeatures across the landscape.

Among the consequences of small water bodiesare:

● elevated evaporation, decreased downstreamflow, and altered groundwater recharge;

● significant sediment trapping; more rapid verti-cal accretion and infilling than large waterbodies;

● stronger redox gradients than in large waterbodies, hence different diagenetic reactions(e.g.of organic material); because of rapid infilling,

more ephemeral storage(decades vs. centuries)than in large water bodies; and

● severely modified aquatic habitats that may atleast partially compensate for—but spatiallyredistribute—anthropogenically lost naturalwetlands.

Does the enumeration of the number of waterbodies really matter? One can suppose that small,ephemeral mud puddles can be legitimatelyignored in any reasoned inventory of water bodydistribution across the landscape. Their presenceacross the landscape can be viewed as brief, localtransition phenomena as water soaks into theground and enters some more persistent hydrolog-ical feature. Farm ponds and other water bodies atscales of hundreds to thousands of square meterscannot be so readily dismissed. These aggregatemodifications of the landscape represent highlyvisible features that persist over periods rangingfrom seasons or years to decades. As they fill withsediment, they tend to be replaced. And theirnumbers across many landscapes are increasing.Their presence appears to be significant; and theirdistribution, abundance, and function are relativelyeasy to assess.

Acknowledgments

We thank Dennis Swaney, Fred Mackenzie, RichSleezer, Avram Primack, and Bob Parkinson forhelp with data analysis and interpretation, andongoing discussions of aspects of the significanceof small water bodies across the landscape. Thanksto Lindsay Theis for the citation of unpublisheddata. SOEST Contribution No. 5974. Some aspectsof this work have derived from products or activ-ities funded in part by US National Science Foun-dation Grant OCE 00-03970, by the Land–OceanInteractions in the Coastal Zone core project ofthe International Geosphere–Biosphere Pro-gramme, and by the Kansas Geological Survey.

References

Carriquiry JD, Sanchez A. Sedimentation in the ColoradoRiver delta and upper Gulf of California after nearly acentury of discharge loss. Mar Geol 1999;158:125–145.


Dahl TE, Johnson CE, Frayer WE. Wetlands, status and trendsin the conterminous United States, mid-1970s to mid-1980s:first update of the national wetlands status report. Washing-ton, DC: US Dept. of the Interior, Fish and Wildlife Service,1991. (22 pp).

Dean WE, Gorham E. Magnitude and significance of carbonburial in lakes, reservoirs, and peatlands. Geology1998;26:535–538.

Dendy FE, Champion WA. Sediment deposition in US reser-voirs: summary of data reported through 1975, Misc. Pub.1362. Washington, DC: US Dept. of Agriculture, 1978. (82pp).

Dynesius M, Nilsson C. Fragmentation and flow regulation ofriver systems in the northern third of the world. Science1994;266:753–762.

Graf WL. Dam nation: a geographic census of American damsand their large-scale hydrologic impacts. Water Res1999;35:1305–1311.

Harbeck GE. A practical field technique for measuring reser-voir evaporation utilizing mass-transfer theory, Prof. Paper272-EUS Geol. Surv, 1962. p. 101–105.

Hayes-Bohanan J. Source-area erosion rates in areas tributaryto Miami Whitewater Lake. MA thesis. Miami University,1989. (96 pp).

Holland DD. Sediment yields from small drainage areas inKansas. Bulletin 16. Kansas Water Resources Board, 1971.(26 pp).

Hutchinson GE. A treatise on limnology. Geography, physics,and chemistry;(Part 1) Geography and physics of lakes.New York: Wiley, 1957. (540 pp).

Mandelbrot B. How long is the coast of Britain? Statisticalself-similarity and fractional dimension. Science1967;156:636–638.

Meade RH, Yuzyk TR, Day TJ. Movement and storage ofsediment in rivers of the United States and Canada. In:Wolman MG, Riggs HC, editors. Surface Water Hydrology,Geology of North America. Boulder, CO: O-1. Geol. Soc.Am, 1990. p. 255–280.

Milliman JD, Syvitski JPM. Geomorphicytectonic control ofsediment discharge to the ocean: importance of small moun-tainous rivers. J. Geol. 1992;100:525–544.

Mitsch WJ, Gosselink JG. Wetlands. 2nd ed. New York: VanNostrand Reinhold, 1993. (722 pp)

Mulholland PJ, Elwood JW. The role of lake and reservoirsediments as sinks in the perturbed global carbon cycle.Tellus 1982;34:490–499.

Oklahoma Water Resources Board. Oklahoma Water Atlas,Publication 135. pp, 1990.

Renwick WH. Continental-scale reservoir sedimentation pat-terns in the United States. In: Walling DE, Webb BW,editors. Erosion and Sediment Yield: Global and RegionalPerspectives, vol. 236IAHS Publ, 1996. p. 513–522.

Rice CW. Storing carbon in soil: why and how? Geotimes2002;47:14–17.

Ritchie JC. Carbon content of sediments of small reservoirs.Water Resour Bull 1989;25:301–308.

Schlesinger WH. Biogeochemistry: an analysis of globalchange. San Diego: Academic Press, 1997. (588 pp).

Seaber PR, Kapinos FP, Knapp GL. Hydrologic Unit MapsWater-Supply Paper 2294. US Geol. Surv, 1987. (63 pp).

Smith SV, Renwick WH, Buddemeier RW, Crossland CJ.Budgets of soil erosion and deposition for sediments andsedimentary organic carbon across the conterminous UnitedStates. Global Biogeochem Cycles 2001;15:697–707.

Sorenson LG, Goldberg R, Root TL, Anderson MG. Potentialeffects of global warming on waterfowl populations breedingin the Northern Great Plains. Climatic Change 1998;40:343–369.

St. Louis VL, Kelly CA, Duchemin E, Rudd JWM, RosenbergDW. Reservoir surfaces as sources of greenhouse gases tothe atmosphere: a global estimate. BioScience 2000;50:766–775.

Stallard RF. Terrestrial sedimentation and the carbon cycle:coupling weathering and erosion to carbon burial. GlobalBiogeochem Cycles 1998;12:231–257.

Sverdrup HU, Johnson MW, Fleming RW. The oceans: theirphysics, chemistry, and general biology. Englewood Cliffs,NJ: Prentice-Hall, 1942. (1060 pp).

Trimble SW, Bube KP. Improved reservoir trap efficiencyprediction. Environ Prof 1990;12:255–272.

Vorosmarty CJ, Sahagian D. Anthropogenic disturbance of the¨ ¨terrestrial water cycle. BioScience 2000;50:753–765.

Walling DE. The sediment delivery problem. J Hyrdol1983;65:209–237.

Williams G, Wolman M. Downstream effects of dams onalluvial rivers, Prof. Paper, 1286. US Geol. Surv, 1984. (83pp).

Documents

Distribution and significance of small, artificial water … and significance of small, artificial water bodies across the United States landscape S.V. Smith *, W.H. Renwick , J.D