This article was downloaded by: [Florida International University]On: 24 September 2013, At: 16:14Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Hydrological Sciences JournalPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/thsj20

Simulating the water budget of a Prairie Potholescomplex from LiDAR and hydrological models in NorthDakota, USAShengli Huanga, Claudia Youngb, Omar I. Abdul-Azizc, Devendra Dahald, Min Fenge &Shuguang Liuf

a ASRC Research and Technology Solutions (ARTS), Contractor to the US Geological Survey(USGS) Earth Resources Observation and Science (EROS) Center, 47914 252nd Street, SiouxFalls, SD 57198, USA; work performed under USGS contract G08PC91508b Earth Resources Technology (ERT), Inc., Contractor to the USGS EROS Center, Sioux Falls,SD 57198, USA; work performed under USGS contract G10PC00044c Department of Civil and Environmental Engineering, Florida International University, 10555W. Flagler Street, Miami, FL 33174, USAd Stinger Ghaffarian Technologies (SGT), Inc., Contractor to the USGS EROS Center, SiouxFalls, SD 57198, USA; work performed under USGS contract G10PC00044e State Key Laboratory of Resources and Environment Information System, Institute ofGeographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing100101, Chinaf USGS EROS Center, 47914 252nd 15 Street, Sioux Falls, South Dakota 57198, USAPublished online: 20 Sep 2013.

To cite this article: Shengli Huang, Claudia Young, Omar I. Abdul-Aziz, Devendra Dahal, Min Feng & Shuguang Liu ,Hydrological Sciences Journal (2013): Simulating the water budget of a Prairie Potholes complex from LiDAR and hydrologicalmodels in North Dakota, USA, Hydrological Sciences Journal, DOI: 10.1080/02626667.2013.831419

To link to this article: http://dx.doi.org/10.1080/02626667.2013.831419

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

1Hydrological Sciences Journal – Journal des Sciences Hydrologiques, 2013http://dx.doi.org/10.1080/02626667.2013.831419

Simulating the water budget of a Prairie Potholes complex from LiDARand hydrological models in North Dakota, USA

Shengli Huang1, Claudia Young2, Omar I. Abdul-Aziz3, Devendra Dahal4, Min Feng5

and Shuguang Liu6

1ASRC Research and Technology Solutions (ARTS), Contractor to the US Geological Survey (USGS) Earth Resources Observation andScience (EROS) Center, 47914 252nd Street, Sioux Falls, SD 57198, USA; work performed under USGS contract G08PC91508.2Earth Resources Technology (ERT), Inc., Contractor to the USGS EROS Center, Sioux Falls, SD 57198, USA; work performed underUSGS contract G10PC00044.3Department of Civil and Environmental Engineering, Florida International University, 10555 W. Flagler Street, Miami, FL 33174, USA4Stinger Ghaffarian Technologies (SGT), Inc., Contractor to the USGS EROS Center, Sioux Falls, SD 57198, USA; work performedunder USGS contract G10PC00044.5State Key Laboratory of Resources and Environment Information System, Institute of Geographic Sciences and Natural ResourcesResearch, Chinese Academy of Sciences, Beijing 100101, China6USGS EROS Center, 47914 252nd 15 Street, Sioux Falls, South Dakota 57198, USA

Received 21 September 2012; accepted 12 February 2013; open for discussion until 1 April 2014

Editor Z.W. Kundzewicz; Associate editor X. Chen

Citation Huang, S.L., Young, C., Abdul-Aziz, O.I., Dahal, D., Feng, M., and Liu, S.G., 2013. Simulating the water budget of a PrairiePotholes complex from LiDAR and hydrological models in North Dakota, USA. Hydrological Sciences Journal, 58 (7), 1–11.

Abstract Hydrological processes of the wetland complex in the Prairie Pothole Region (PPR) are difficult tomodel, partly due to a lack of wetland morphology data. We used Light Detection And Ranging (LiDAR) data setsto derive wetland features; we then modelled rainfall, snowfall, snowmelt, runoff, evaporation, the “fill-and-spill”mechanism, shallow groundwater loss, and the effect of wet and dry conditions. For large wetlands with a volumegreater than thousands of cubic metres (e.g. about 3000 m3), the modelled water volume agreed fairly well withobservations; however, it did not succeed for small wetlands (e.g. volume less than 450 m3). Despite the failurefor small wetlands, the modelled water area of the wetland complex coincided well with interpretation of aerialphotographs, showing a linear regression with R2 of around 0.80 and a mean average error of around 0.55 km2.The next step is to improve the water budget modelling for small wetlands.

Key words hydrological modelling; LiDAR; Prairie Pothole Region; water budget; wetland

Simulation du bilan hydrique d’un complexe de fondrières des prairies à partir de données LiDARet de modèles hydrologiques dans le Dakota du Nord (Etats-Unis d’Amérique)Résumé Les processus hydrologiques du complexe de zones humides de la région des fondrières des prairies sontdifficiles à modéliser, en partie à cause d’un manque de données de morphologie des zones humides. Nous avonsutilisé des jeux de données LiDAR (Light Detection And Ranging) pour calculer les caractéristiques des zoneshumides. Nous avons ensuite modélisé les pluies, les précipitations neigeuses, la fonte des neiges, le ruissellement,l’évaporation, le mécanisme de débordement par saturation, les pertes des eaux souterraines peu profondes, etl’effet des conditions humides et sèches. Pour les zones humides étendues ayant un volume supérieur à quelquesmilliers de mètres cubes (par exemple, environ 3000 m3), le volume d’eau modélisé s’accorde assez bien avec lesobservations, mais le modèle a été tenu en échec pour les zones humides de petite taille (par exemple, volumeinférieur à 450 m3). Malgré l’échec pour les petites zones humides, la zone d’eau modélisée du complexe dezones humides coïncidait bien avec l’interprétation des photographies aériennes, montrant une régression linéaireavec un R2 de l’ordre de 0,80 et une erreur moyenne autour de 0,55 km2. La prochaine étape est d’améliorer lamodélisation du bilan hydrique pour les zones humides de petite taille.

Mots clefs modélisation hydrologique; LiDAR; région des fondrières des prairies; bilan hydrique; zones humides

This work was authored as part of the Contributor’s official duties as an Employee of the United States Government and is therefore a work of the United StatesGovernment. In accordance with 17 U.S.C. 105, no copyright protection is available for such works under U.S. Law.

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

2 Shengli Huang et al.

1 INTRODUCTION

1.1 PPR Wetlands

There are millions of depressional wetlands in thePrairie Pothole Region (PPR) of North America,extending from north-central Iowa to central Alberta(Fenneman 1931). Many wetlands are small andshallow, with depths generally less than one metre(Sethre et al. 2005, Zhang et al. 2009), and theyare underlain by glacial till of low permeability(Winter and Rosenberry 1995). The water balanceof these wetlands is influenced by many factors,including snow redistribution, precipitation, evap-otranspiration, runoff, groundwater exchange andantecedent water storage (Johnson et al. 2005; Fangand Pomeroy 2008, van der Kamp and Hayashi2009). Under these conditions, prairie pothole wet-lands have labile water surfaces determined by thewater balance between inflows and loss. In general,the major water inflows are snowmelt and summerprecipitation, and the greatest water loss is evapo-ration (Winter et al. 2001). In spring, the in situmelt of accumulated snow contributes extensive andrapidly melting water to the PPR wetlands; therefore,soon after snowmelt, the wetlands generally reachtheir maximum annual extent (Stewart and Kantrud1971, Winter and Rosenberry 1995, LaBaugh et al.1998; Parkhurst et al. 1998, Carroll et al. 2005,Johnson et al. 2010). In summer, open-water evap-oration is a significant water loss: average annualevaporation is nearly twice as much as that from pre-cipitation (Kohler et al. 1959). Deep groundwatermovement is very slow in the PPR wetlands; however,shoreline-related shallow groundwater loss accountsfor a considerable amount of the water loss in smalldepressions (Eisenlohr 1966, Millar 1971).

The PPR wetland landscape complexes are eco-logically and economically important for providingecosystem goods and services (Swanson et al. 2003,Voldseth et al. 2007, Johnson et al. 2010). Forprairie wetlands management, it is important to modelthe hydrological processes and understand how theamount of surface water stored in wetland com-plexes changes over time (LaBaugh et al. 1998).To address this issue, numerous studies have investi-gated hydrological processes and examined the waterbalance of prairie wetlands (e.g. Millar 1971, Poianiand Johnson 1993, Woo and Rowsell 1993, Winterand Rosenberry 1995, 1998, LaBaugh et al. 1996,1998, Rosenberry and Winter 1997, Hayashi et al.1998, van der Kamp and Hayashi 1998, 2009, Suet al. 2000, Spence 2007, Fang et al. 2010, Niemuth

et al. 2010, Liu and Schwartz 2011). However, a land-scape hydrological modelling that includes multiplewetlands is desired for two reasons.

First, a wetland complex rather than one or sev-eral ponds is desired. Over the landscape perspec-tive of the PPR, there are several different types oftemporary, seasonal, semi-permanent and lake wet-lands with different dynamics. For example, tempo-rary wetlands are the most environmentally variablewithin the wetland complex (Johnson et al. 2004),and it was revealed that smaller ponds, when com-pared with large ponds, have much higher recessionrates due to greater infiltration loss through shal-low groundwater flow (van der Kamp and Hayashi2009). Diverse complexes of wetlands contributehigh spatial and temporal environmental heterogene-ity, productivity and biodiversity to these glaciatedprairie landscapes (Johnson et al. 2010); therefore,all the wetlands differing in landscape type and sizeshould be considered together for evaluating the manyecosystem services (Leibowitz 2003, Johnson et al.2005, 2010, Niemuth et al. 2010). However, exist-ing studies of wetland complexes that have includeddetailed hydrological processes represent only partof the diverse hydrogeological setting of the prairie(Winter and Rosenberry 1995, LaBaugh et al. 1998,van der Kamp and Hayashi 1998). Semi-permanentwetlands were the only water regime considered inseveral models (Poiani and Johnson 1991, Poiani et al.1996, Johnson et al. 2005); therefore, the signifi-cant water loss from small wetlands through shal-low groundwater was not considered. For example,different versions (1.0, 2.0, 3.2) of WETSIM weredeveloped for hydrology and vegetation dynamics fora single-basin and semi-permanent wetlands (Poianiand Johnson 1991, 1993, Poiani et al. 1995, 1996,Carroll et al. 2005, Voldseth et al. 2007). Other wet-land types, such as temporary wetlands, were notwell addressed (e.g. Poiani and Johnson 1993, Larson1995).

Second, landscape wetland connectivity needsconsideration. Surface-water connections can occurbetween prairie potholes (Rosenberry and Winter1997, Winter and Rosenberry 1998, Johnson et al.2004). The extent of water connections is insuffi-ciently documented; thus, the effect of channel out-flow on prairie wetland ecology has been scarcelyexplored (Leibowitz and Vining 2003, Johnson et al.2004). To properly map the recurrence interval of dif-ferent connectivities would require precise elevationdata due to the gentle slopes and variable directionsof flow in wetland landscapes (Leibowitz and Vining

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

Water budget simulation of a wetland complex 3

2003). To address this challenge, Johnson et al.(2010) used 10 wetlands and 40 groundwater wells toparameterize the WETLANDSCAPE (WLS) modelfor wetland complex simulation in South Dakota.The WLS model allowed for overflow connectivitybetween basins, but the basin morphometry and over-flow connectivity among wetlands were determinedfrom a detailed ground survey using a total stationsurvey system, which is difficult to apply over a broadarea. Liu and Schwartz (2011) used a pothole com-plex hydrological model (PCHM) and coupled thepure power laws approach to fully account for the spa-tial and temporal distribution of surface waters. ThePCHM was developed mainly to simulate water area–frequency relationships, and the pure power laws areweakly coupled to the dynamic character of the waterbodies; therefore, the PCHM could not produce waterbodies that disaggregate during dry periods or coa-lesce during wet periods (Liu and Schwartz 2011).Niemuth et al. (2010) described spatial and tempo-ral patterns of wetness and variation in wetness, anddetermined how the inter-annual dynamics of wet-lands were related to the water regimes; however,their work precluded inferences about how individ-ual wetlands vary as a function of water regime, anddid not reveal spatial variance at the scale of wet-land complexes. Fang et al. (2010) used the ColdRegions Hydrological Modelling platform (CRHM)for an externally drained basin with an area of approx-imately 400 km2. The basin was divided into five“representative basins” (RBs) with each RB fur-ther divided into seven hydrological response units(HRUs). Their model performed very well in pre-dicting snowpack, soil moisture and streamflow whenevaluated against field observations; however, wetland

surface water area change was not targeted in thatstudy.

The objective of this paper was to model thewater area dynamics of a landscape wetland com-plex at daily time steps with “fill-and-spill” surfacewater connections and shallow groundwater takeninto account. To do this, we selected a 196-km2 studyarea to model the daily water balance of all the indi-vidual wetlands. The PPR wetlands are distributedin relatively flat topography and DEMs with resolu-tions of up to 10 m are usually inadequate to capturethe detailed relief of the region (Huang et al. 2011a).Thus, using a conventional low-quality DEM wouldnot be accurate enough to produce a prairie channelnetwork (Fang et al. 2010). The benefits of apply-ing a Light Detection And Ranging (LiDAR) remotesensing approach in a prairie hydrology study arepromising; a LiDAR-derived DEM was highly rec-ommended for investigating prairie hydrology (Fanget al. 2010). Our study was based on the 0.5-m DEMand the associated products derived from a LiDARsensor, as reported by Huang et al. (2011a).

2 STUDY AREA

The study area is a 6.4 km × 30.6 km (196 km2) blockin Stutsman County, North Dakota, USA (Fig. 1),which was reported by Huang et al. (2011a, 2011b).The study area ranges from 435.7 to 596.3 m in ele-vation and is covered by corn (14.0%), small grains(7.6%), soybeans (24.9%), sunflowers (1.1%), farm-steads (0.9%), grassland (27.0%), hayland (7.2%),roads (1.7%), trees (1.7%) and wetlands (13.8%).The study area is characterized by a dynamic con-tinental climate with long, cold, dry winters and

Fig. 1 The study area (LiDAR digital elevation as background) is located in the Prairie Pothole Region (PPR) of NorthDakota, USA. Aerial photographs were collected and processed for an area marked by the rectangle (see Huang et al.2011b). Permission from Elsevier.

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

4 Shengli Huang et al.

short, mild, variably wet summers (Kantrud et al.1989). The average annual temperature is 7◦C andthe average monthly temperature ranges from –13◦Cin January to 20◦C in July (Winter and Rosenberry1995). Average annual precipitation is 440 mm andthe average monthly precipitation ranges from 12 mmin December to 91 mm in June (Carroll et al.2005). In the marked rectangle of Fig. 1, six high-resolution aerial photographs and one LiDAR dataset were interpreted for wetland open water, result-ing in the water areas of 3.58, 1.96, 5.96, 5.78,5.77, 5.71 and 4.60 km2 for 18 June 1984, 28 July1990, 25 September 1997, 2 July 2004, 15 July 2005,13 July 2006 and 17 May 2008, respectively (seeHuang et al. 2011b).

3 DATA SETS

Daily weather data of 1980–2008 were compre-hensively compiled from DAYMET (http://www.daymet.org/), including: maximum temperature, min-imum temperature, mean temperature, precipitation,vapour pressure deficit, solar radiation and windspeed information. The soil database of SSURGOwas downloaded from the Natural ResourcesConservation Service (http://soildatamart.nrcs.usda.gov/). Land cover data from NLCD 2001, whichdescribes the land surface condition at 30-m reso-lution, were downloaded from http://www.mrlc.gov/

LiDAR data at 0.5-m resolution were processed(Huang et al. 2011a). The wetland catchments andtheir spilling points were delineated within the studyarea (Fig. 2). The connectivity among the wetlandbasins was modelled, and the maximum water distri-bution and storage capacity of individual basins wascomputed (Huang et al. 2011a). The water depths ofseveral wetlands were measured in summer at weeklyintervals. In Fig. 2, five wetlands (W1–W5) aremarked. The basin morphometries of these wetlandswere surveyed (see Huang et al. 2011a). By com-bining the water depth observations from 1980 to2005 and the wetland shape, we calculated their watervolume.

4 METHODOLOGY

4.1 Overview of wetland water balance

In this study, for a wetland of interest, the term “water-shed area” (WA) refers to the total constant land areathat contributes surface water to this wetland and theterm “ponded area” (PA) refers to the inundated area

Fig. 2 Wetland delineation from LiDAR DEM. The spill-ing points (red dots) are identified as the minimum eleva-tion along the catchment boundaries (yellow lines). Surfaceflow directions are depicted as white arrows, and the bi-directional arrows indicate that the basins merge whenthe wetlands are full. The blue polygons are water bodiesdetected from 17 May 2008 LiDAR data sets (see Huanget al. 2011a). Those five wetlands marked by numbers 1–5(i.e. W1 to W5) were measured in water depth and basinmorphometry.

of the wetland. The value of PA varies according tothe water balance: when the water level is beyond thespilling point, the wetland will fill and spill, thus PAwill reach its maximum area (PAmax) and the wet-land will reach its maximum capacity (V max). ThePA ranges from 0 (when the wetland is dry) to PAmax

(when the wetland is full). The parameters WA, PAmax

and V max were all modelled from LiDAR data (seeHuang et al. 2011a). The upland is the differencebetween WA and PA (Fig. 3).

The conceptual cross-section of the wetlands atlandscape level is shown in Fig. 4. In our study,the water storage in a wetland of interest wasmodelled by:

Vt = Vt−1 + VCt (1a)

VCt = Pt · PAt + SRt(WA − PAt)

+ SIVt − ETt · PAt − SOVt − SGWt

+ αWA [(Pt − ETt) − (Pm − ETm)]

(1b)

where t is the day of simulation; and Vt is the waterstorage, and depends on the preceding water storageVt−1 and the daily volume change, VCt. The incom-ing water volume into the wetland is the result of:

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

Water budget simulation of a wetland complex 5

Fig. 3 Conceptual scheme of watershed area (WA), pondedarea (PA), maximum ponded area (PAmax) and spillingpoint (dot).

direct precipitation, Pt × PAt; surface runoff fromupland, SRt(WAt – PAt), including both snowmeltand rainstorm runoff; and the surface water flowfrom contributing wetlands through the spilling point,SIVt. The outgoing water volume from the wetland ofinterest includes: open-water evaporation, ETt × PAt;surface outflow through the spilling point, SOVt; andshallow groundwater loss, SGWt (van der Kamp andHayashi 2009). Deep and inter-wetland groundwaterflow has minor effects on the wetland water bal-ance under a normal moisture regime (van der Kampand Hayashi 2009); however, dry and wet conditionshave an important effect on the water budget (Winterand Rosenberry 1995, Rosenberry and Winter 1997).Therefore, in equations (1), αWA[(Pt – ETt) – (Pm

– ETm)] was used to adjust the dry and wet con-ditions on the water budget, where Pm and ETm

are daily mean precipitation and ET averaged from

1980–2005 climate data sets, respectively, while α

is a calibration parameter with a value of 0.42 (seeSection 4.5).

The model was driven by daily weather data.In each day, volume (V ) was converted to pondedarea (PA) based on the strong statistical relationshipbetween volume and area in a topographic depres-sion, as reported by Huang et al. (2011a). The resultis passed to the second day for iteration.

4.2 Precipitation (P), runoff (SR) andopen-water evaporation (ET)

The form of precipitation as rainfall or snowfall ismodelled based on the daily mean temperature. If theaverage daily temperature is below 0◦C, the fallingprecipitation is assumed to be snowfall (Carrollet al. 2005). Snowfall accumulates on the ground assnowpack in terms of snow water equivalent (SWE).Snowpack melting was modelled with the degree-daymethod as Mt = 0.274Tt, where Mt is the snowmeltdepth (m) on day t, Tt is the mean daily temperature(◦C) and 0.274 is the approximate degree-day ratio(cm ◦C−1 d−1) (Rango and Martinec 1995). WhenTt < 0◦C, Mt = 0 (i.e. no snowmelt). We assumed95% snowmelt runoff flowed into ponds because ofthe frozen soil (Woo and Winter 1993, Larson 1995,Winter and Rosenberry 1995, Hayashi et al. 1998).

If the average daily temperature is above 0◦C,the falling precipitation is assumed to be rainfall.The rainfall runoff was modelled with the US SoilConservation Service’s infiltration model, based onthe curve number (CN) approach (SCS 1972):

Rt ={ (Pt − 0.2St)2

Pt + 0.8StPt > 0.2St

0 Pt ≤ 0.2St

(2)

where Rt is the surface runoff (in m) per unit uplandarea on day t; and St is the soil abstraction factor (m)that is calculated as:

Fig. 4 Conceptual cross-section of wetlands at landscape level. Incoming and outgoing components are depicted.

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

6 Shengli Huang et al.

Table 1 Curve number computation from soil and landcover (according to NLCD 2001)..

Land cover class Hydrological soil group

A B C D

Open Water 100 100 100 100Grassland/Herbaceous 49 69 79 84Pasture hay 49 69 79 84Cultivated crops 67 78 85 89

St = 0.254

(100

CNt · β− 1

)(3)

where CNt is the curve number (CN) on day t; and β isa coefficient calibrated as 1.04 (see Section 4.5). TheCN under normal conditions was obtained from thecombination of SSURGO hydrological soil group andmajor NLCD 2001 land cover (Table 1). Antecedentmoisture condition (AMC) indicates basin wetnessand availability of soil moisture storage prior to a rain-fall event and can have a significant effect on runoffvolume. A total rainfall in the preceding 5-day period(R5, in mm) was used to adjust the CN (Chow et al.2002). When R5 < 12.7 in the dormant season (i.e.1 October–31 March), or for R5 < 35.6 in the grow-ing season (i.e. 1 April–30 September), the CN wasadjusted using the following equation:

CNnew = 4.2CN

10 − (0.058CN)(4)

For R5 > 27.9 in the dormant season, or R5 > 53.3 inthe growing season, the CN was adjusted using:

CNnew = 23CN

10 − (0.13CN)(5)

The CN approach described above incorporated bothspatially explicit SSURGO soil data and land coverdata (Table 1), as well as temporal AMC (equations(4) and (5)), so the CN values for different wetlandsused in the study were changing over space and time.

During the winter, water is assumed not to evap-orate from the lake ice until the mean air temperatureof the preceding 10 days in spring is greater than3◦C (Johnson et al. 2004, 2010). During the sum-mer, we modelled the standard grass-based Penman-Monteith potential evapotranspiration (PET) (Allenet al. 1998) as the open-water evaporation, with the

same assumption of Carroll et al. (2005) and Poianiet al. (1996) that the open water ET is equal to PET.

4.3 Groundwater loss (SGW)

The deep groundwater flow in low-conductivity tillis very slow and has negligible effects on water bal-ance; shoreline seepage losses are a proportionatelysmaller component of the water balance for a largewetland (Millar 1971, van der Kamp and Hayashi1998). However, for relatively small wetlands, infiltra-tion (i.e. lateral flow of shallow groundwater betweenwetland ponds and the riparian zone) accounts for themajority of water loss (Hayashi et al. 1998, van derKamp and Hayashi 2009). This kind of seepage fromthe pond is a major factor in the wetland water balanceand needs careful consideration (Su et al. 2000).

The horizontal flow of groundwater from the wet-land to the upland is mainly due to upland evapo-transpiration (Winter and Rosenberry 1995, Hayashiet al. 1998, Su et al. 2000, van der Kamp and Hayashi2009). Phreatophytic vegetation absorbs and tran-spires water and thus creates depressions in the watertable around small wetlands during the growing sea-son, which increases the hydrostatic gradient and rateof seepage outflow. Greater and more rapid warmingof the shallow water accelerates the rate of evapo-transpiration (Millar 1971). Eventually the infiltratedwater flows laterally through a zone of high hydraulicconductivity, due to fractures in the till, and is used upby plants on the upland (Hayashi et al. 1998).

The flow of groundwater from the ponded waterto the margins is a highly transient phenomenon thatchanges over hours, days, seasons and years, follo-wing the trend of evaporation rate (Winter andRosenberry 1995, Hayashi et al. 1998, van der Kampand Hayashi 1998). A strong linear correlation existsbetween the rate of water level recession causedby shallow groundwater and the ratio of the pondshoreline length (m) to the pond area (m2) (Millar1971, van der Kamp and Hayashi 2009); therefore,we parameterized the shallow groundwater loss asfollows:

SGWt = ωRSrefETt

ETref· Ratio · PAt (6)

where ω reflects local shallow groundwater loss (cal-ibrated as 0.90; see Section 4.5); RSref (= 0.04 m2/d)is the reference recession slope; and ETref (=4.7 mm/d) is the average daily evaporation from largelakes in July. The two values of RSref and ETref were

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

Water budget simulation of a wetland complex 7

obtained from the studies of Millar (1971) and van derKamp and Hayashi (2009), conducted in Saskatoon,Saskatchewan. The Ratio of the pond shoreline length(m) to pond area (m2) was calculated for each indi-vidual wetland from the corresponding NWI polygonin units of 1/m. In some cases, NWI data are com-posed of polygons showing wetland zones (i.e. severalpolygons represent one wetland). In this case, wemerged these geographically adjacent zones into onepolygon, following the protocol described by Johnsonand Higgins (1997). If the zones are not adjacentgeographically, average Ratio was used. ParametersETt and PAt are the daily open-water evaporationand ponded area, respectively, as described previ-ously. Millar (1971) and van der Kamp and Hayashi(2009) showed that the temporal pattern in the rate ofshoreline-related water loss coincided with the trendin lake evaporation values, and the spatial variation ofrecession slopes of their three sites also agreed withthe evaporation difference; therefore, the ETt/ETref inequation (6) was used to capture and adjust the spatialand temporal variation of the recession slope, RSref.According to this equation, when the wetland is verylarge, the ratio is close to 0, resulting in very lowgroundwater loss. This agrees with the findings thatshallow groundwater is a minor factor for a large wet-land (Millar 1971, van der Kamp and Hayashi 1998).When a wetland is very small, the ratio increases andthis results in higher groundwater loss. For example,when the ratio is 0.1 and 0.2, the depth of groundwaterloss in July at Saskatoon (in this case ETt/ETref

=1) would be 4 and 8 mm/d. This agrees with theobservations summarized by Millar (1971) and vander Kamp and Hayashi (2009).

4.4 Surface water routing (SIV and SOV)

In early spring, snowmelt water runs off into depres-sions and may be routed from higher to lower-lyingwetlands by surface fill-and-spill mechanisms, partic-ularly in wet years (Hayashi et al. 1998, Gray et al.2001, van der Kamp et al. 2003, van der Kamp andHayashi 2009). The spilling point, maximum volumeand routing direction of each wetland were modelledby Huang et al. (2011a). Therefore, for a wetland ofinterest, we have the following information:

(a) which wetlands can contribute over-spilled waterto this wetland;

(b) where the extra water exits through the outlet, ifthe wetland of interest is full; and

(c) if and how this wetland basin merges with othergeographically adjacent basins and creates alarger basin (Figs 2 and 4).

In our study, the volume of surface water that flowsto other wetlands was modelled as the extra waterbeyond the maximum capacity at a daily time step.

4.5 Model calibration

The model parameters α, β and ω were calibratedusing the observed water volume of W1 from 1980 to1995. During calibration, we set the initial water con-dition of 1980 and drove our model using the dailyweather data. We set individual parameter values,made a series of simulations, generated wetland waterareas and then computed the correlation betweenthe observed volume and the modelled volume. Thisprocess was repeated until the minimum root meansquare error was achieved, and a set of accept-able parameters was obtained: α = 0.42, β = 1.04 andω = 0.90.

5 RESULTS

The comparisons between observed and modelledwater volume for the five wetlands are shown inFig. 5. From 1980 to 1992, the measured depth ofwetland W1 is low with a small peak in 1987 (around1 m) and a minimum depth in 1992 (almost 0 m); from1992 to 1999, the water depth of W1 increases signif-icantly from dry to 2.5 m; and since 1999, wetlanddepth remains greater than 1.6 m but shows a gen-eral decreasing trend. Based on the water depth andthe basin shape, the converted water volume shows asimilar pattern: From 1980 to 1992, the water volumeis low with a small peak in 1987 (around 20 000 m3)and a minimum volume in 1992 (almost dry); from1992 to 1999, the water volume increases signifi-cantly from dry to 110 000 m3; since 1999, its volumeremains >45 000 m3, but shows a general decreas-ing trend. When compared with observed volume, themodelled W1 volume coincides well in both trendand magnitude. This agreement can also be found foranother large wetland W2; however, the coincidencecannot be observed for small wetlands W3 and W4,which are less than 450 m3 in size. For wetland W5,with a moderate size of around 3000 m3, the agree-ment is better than for small wetlands, but poorerthan for large ones. All these comparisons indicatethat our modelling performed well for those wetlandsof sizes greater than thousands of cubic metres, but

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

8 Shengli Huang et al.

Fig. 5 Comparison of the observed and modelled water volume for wetlands W1–W5 (see Fig. 2 for locations).

Fig. 6 Water area comparison between modelling and inter-pretation within the area marked by the rectangle in Fig. 1.

did not succeed for small wetlands. Nevertheless, thewater area comparison between aerial photographsand modelling at landscape level (Fig. 6) shows the

linear regression with R2 of around 0.80 and a meanaverage error of around 0.55 km2, indicating theacceptable accuracy of our water budget modelling.

6 DISCUSSION

The hydrological process and water levels in thePPR wetland are very complex and affected by manyfactors: wetland size, basin morphometry, soil type,land use, catchment size, topographic position andwetland drainage patterns, as well as precipitation(Kantrud et al. 1989, Euliss and Mushet 1996, vander Kamp et al. 1999). Rainfall, snowfall, snowmeltand rainstorm runoff, evapotranspiration, as well asfill-and-spill mechanisms, all influence water massbalance (van der Kamp and Hayashi 2009). Takingadvantage of the high-resolution LiDAR elevation

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

Water budget simulation of a wetland complex 9

data set, we modelled the entire wetland complexinstead of just an individual wetland so that the inter-dependence of wetlands of different sizes could beevaluated. Our wetland complex modelling at thelandscape level benefits the research and managementof many ecosystem services, such as flooding, waterquality and biodiversity.

Regarding flooding, the human-altered PPR wet-lands may contribute floodwater to downstreamrivers, stimulating interest in modelling water storagecapacities of wetlands and their surrounding catch-ments to facilitate flood mitigation efforts. A land-scape perspective is required to link the performanceof wetland functions with off-site flood effects, andthe ability to define the extent of hydrological con-nection with respect to different flood recurrenceintervals would be useful (Leibowitz 2003). Huanget al. (2011a) demonstrated a conceptual model forusing LiDAR data to improve the estimation of flood-water mitigation potential. Nevertheless, to improverealism, the water level at the beginning of floodmust be known. The study reported here can pro-vide a daily water budget prior to a flood occurrenceand thus meet this demand. Regarding water quality,both isolation and connection are important factorsthrough the transport of ions, nutrients, sedimentsand contaminants, but their effect on wetland water-quality function has received relatively little atten-tion (Leibowitz 2003, Leibowitz and Vining 2003,Whigham and Jordan 2003). Our study could provideinsight into the isolation and connectivity of wet-lands and thus enable a better understanding of waterquality.

Regarding biodiversity, PPR wetland complexessupport higher species richness compared with sin-gle, isolated wetlands of comparable total surface area(Naugle et al. 1999). The PPR wetlands, with longerwet–dry cycles, are hypothesized to have a higherfloral and faunal diversity, such as fish, waterfowl,amphibians and invertebrates (van der Valk 2005).The plant species composition, evenness and net pri-mary productivity of hydrologically-isolated wetlandswere significantly different from those in wetlandsconnected with temporary surface-water connectionsand/or soil-water pathways, which may be due tosalinity and inundation (Cook et al. 2007). Ourstudy could provide the required information on thehydrological isolation and interconnection for diver-sity study and management (Swanson et al. 2003, vander Kamp and Hayashi 2009).

Although our wetland complex modelling at thelandscape level rather than at individual wetland level

may benefit many ecosystem services, our currentmodelling has several deficiencies, some of whichcould lead to poorer model performance for smallwetlands. First, the water table was not tracked belowthe wetland bottom in our study; this may reducemodel performance, especially for small wetlands thathave more labile water change and dry out more fre-quently than large wetlands (Voldseth et al. 2007).Second, different soil and vegetation at the fringe ofstanding water and in the upland have different evapo-transpiration rates (LaBaugh et al. 1998). We mod-elled the wetland shallow groundwater loss throughthe ratio of shoreline to area, but we could not capturethe spatial variation of vegetation and soil; also, theratio of shoreline to the area of each individual wet-land was constant, so could not capture its dynamicsfollowing water change. Small wetlands suffer morefrom this deficiency, because they are more affectedby this shallow groundwater loss (Eisenlohr 1966,Millar 1971, van der Kamp and Hayashi 2009). Third,water was assumed not to evaporate from ice-coveredwetlands. Small wetlands differ from large wetlandsin the extended ice cover (Barica and Mathias 1979),which could contribute to limited model performancefor small wetlands. Fourth, in our hydrological mode-lling, the spilling point location, the inter-wetlandconnectivity, the maximum storage capacity (V max)and the relationship between volume (V ) and pondedarea (PA) were all applied to this model; however,these data were not perfect and bias did exist, as dis-cussed by Huang et al. (2011a). Fifth, snow distribu-tion is not spatially homogeneous due to grass trap orwind blowing (Voldseth et al. 2007, Fang et al. 2010);however, we assumed a spatially-homogeneous snowdistribution in our model. Finally, land cover and landuse changed between 1980 and 2009, resulting inthe change of curve number; however, we only usedstatic NLCD 2001 land cover classes in our study. Allthese deficiencies may lead to biases in our modelling,especially for small wetlands. Further study is neededto improve the estimation.

7 CONCLUSION

The hydrological processes for the landscape wetlandcomplex in the PPR are difficult to model, partly dueto the lack of wetland morphology. The use of LiDARdata is an ideal solution to this problem. Without theuse of high-resolution DEMs, the subtle landscapefeatures of the PPR wetlands in areas of extremelylow relief cannot be resolved. The LiDAR approachprovides highly accurate elevation data that capture

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

10 Shengli Huang et al.

detailed wetland morphology, enabling us to knowthe catchment area, spilling point, maximum pondedarea and maximum volume of each wetland, as well asthe channel network and where basins merge amongwetlands. This information is critical for hydrologicalmodelling of the complex wetland landscape (e.g.water spilling).

With the application of LiDAR in our models, thechange of surface water stored in a wetland complexover time can be modelled fairly well with the factorsof rainfall, snowfall, snowmelt open-water evapora-tion, surface runoff, shallow groundwater loss and theeffect of dry/wet conditions taken into account. Themodel performed fairly well for those wetlands withvolumes greater than thousands of cubic metres (e.g.about 3000 m3, as observed in W5), but did not suc-ceed for small wetlands (e.g. volume < 450 m3, asobserved in W3 and W4). Despite the failure in smallwetlands, water area at the landscape level could bemodelled at an acceptable accuracy. The estimationof the water budget for small wetlands needs to beimproved in the future.

Acknowledgements The authors are very grate-ful to Drs Yiping Wu, Ramesh Singh, NormanBliss and Mingshi Chen for internally reviewingthe manuscript, and for their advice on modeldesign and calibration. The authors also gratefullythank Professor Hayashi for the advice on shallowgroundwater modelling, and Thomas Adamson forrevising the English. Any use of trade, product, orfirm names is for descriptive purposes only and doesnot imply endorsement by the US Government.

Funding This work was supported by US Departmentof Agriculture’s Conservation Effects AssessmentProgram (CEAP)—Wetlands, the US GeologicalSurvey’s Geographic Analysis and Monitoring(GAM) and the Global Change Research Programs.

REFERENCES

Allen, R.G., et al., 1998. Crop evapotranspiration—Guidelines forcomputing crop water requirements. Rome, Italy: Food andAgriculture Organization of the United Nations, FAO Irrigationand Drainage Paper 56, ISBN 9251042195.

Barica, J. and Mathias, J.A., 1979. Oxygen depletion and winterkillrisk in small prairie lakes under extended ice cover. Journal ofthe Fisheries Research Board of Canada, 36 (8), 980–986.

Carroll, R.W., et al., 2005. Simulation of a semipermanent wetlandbasin in the Cottonwood Lake Area, East-Central North Dakota.Journal of Hydrological Engineering, 1 (10), 70–84.

Chow, V.T., Maidment, D.K., and Mays, L.W., 2002. Applied hydrol-ogy. New York: McGraw-Hill.

Cook, E.R., et al., 2007. North American drought: reconstructions,causes, and consequences. Earth-Science Reviews, 81 (1–2),93–134.

Eisenlohr, W.S. Jr., 1966. Water loss from a natural pond throughtranspiration by hydrophytes. Water Resources Research, 2 (3),443–453.

Euliss, N.H. Jr. and Mushet, D.M., 1996. Water-level fluctuation inwetlands as a function of landscape condition in the prairiepothole region. Wetlands, 16 (4), 587–593.

Fang, X. and Pomeroy, J.W., 2008. Drought impacts on Canadianprairie wetland snow hydrology. Hydrological Processes, 22(15), 2858–2873.

Fang, X. et al., 2010. Prediction of snowmelt derived streamflow in awetland dominated prairie basin. Hydrology and Earth SystemSciences, 14 (6), 991–1006.

Fenneman, N.M., 1931. Physiography of western United States. NewYork: McGraw-Hill.

Gray, D.M., et al., 2001. Estimating areal snowmelt infiltration intofrozen soils. Hydrological Processes, 15 (16), 3095–3111.

Hayashi, M., van der Kamp G., and Rudolph, D.L., 1998. Mass trans-fer processes between a prairie pothole and adjacent uplands, 1:Water balance. Journal of Hydrology, 207 (1–2), 42–55.

Huang, S., et al., 2011a. Demonstration of a conceptual model forusing LiDAR to improve the estimation of floodwater mitiga-tion potential of Prairie Pothole Region wetlands. Journal ofHydrology, 405 (3–4), 417–426.

Huang, S., et al., 2011b. Integration of Palmer Drought Severity Indexand remote sensing data to simulate wetland water surface from1910 to 2009 in Cottonwood Lake area, North Dakota. RemoteSensing Environment, 115 (12), 3377–3389.

Johnson, R.R. and Higgins, K.F., 1997. Wetland resources of easternSouth Dakota. Brookings: South Dakota State University.

Johnson, W.C., et al., 2004. Influence of weather extremes onthe water levels of glaciated prairie wetlands. Wetlands, 24,385–398.

Johnson, W.C., et al., 2005. Vulnerability of Northern Prairie wet-lands to climate change. BioScience, 55 (10), 863–872.

Johnson, W.C., et al., 2010. Prairie wetland complexes as landscapefunctional units in a changing climate. BioScience, 60 (2),128–140.

Kantrud, H.A., Krapu, G.L., and Swanson, G.A., 1989. Prairie basinwetlands of the Dakotas: a community profile. Washington, DC:US Fish and Wildlife Service Biological Report, 85 (7.28).

Kohler, M.A., Nordenson, T.J., and Baker, D.R., 1959. Evaporationmaps for the United States. US Weather Bureau TechnicalPaper, 37, 18.

LaBaugh, J.W., Winter, T.C., and Rosenberry, D.O., 1998. Hydrologicfunctions of prairie wetlands. Great Plains Research, 8, 17–37.

LaBaugh, J.W., et al., 1996. Changes in atmospheric circulationpatterns affect midcontinent wetlands sensitive to climate.Limnology Oceanography, 41 (5), 864–870.

Larson, D.L., 1995. Effects of climate on numbers of Northern Prairiewetlands. Climatic Change, 30 (2), 169–180.

Leibowitz, S.G., 2003. Isolated wetlands and their functions: anecological perspective. Wetlands, 23 (3), 517–531.

Leibowitz, S.G. and Vining, K.C., 2003. Temporal connectivity in aprairie pothole complex. Wetlands, 23 (1), 13–25.

Liu, G., and Schwartz, F.W., 2011. An integrated observational andmodel-based analysis of the hydrologic response of prairiepothole systems to variability in climate. Water ResourcesResearch, 47, W02504, doi:10.1029/2010WR009084.

Millar, J.B., 1971. Shoreline–area ratio as a factor in rate of waterloss from small sloughs. Journal of Hydrology, 14 (3–4),259–284.

Naugle, D.E., et al., 1999. Scale-dependent habitat use in threespecies of prairie wetland birds. Landscape Ecology, 14 (3),267–276.

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13

Water budget simulation of a wetland complex 11

Niemuth, N.D., Wangler, B., and Reynolds, R.E., 2010. Spatial andtemporal variation in wet area of wetlands in the prairie potholeregion of North Dakota and South Dakota. Wetlands, 30 (6),1053–1064.

Parkhurst, R.S., et al., 1998. Evaporation from a small prairie wetlandin the Cottonwood Lake area, North Dakota: an energy budgetstudy. Wetlands, 18 (2), 272–287.

Poiani, K.A. and Johnson, W.C., 1991. Global warming and prairiewetlands: consequences for waterfowl habitat. BioScience, 41(9), 611–618.

Poiani, K.A., and Johnson, W.C., 1993. A spatial simulation model ofhydrology and vegetation dynamics in semi-permanent prairiewetlands. Ecological Application, 3 (2), 279–293.

Poiani, K.A., Johnson, W.C., and Kittel, T.G.F., 1995. Sensitivityof a prairie wetland to increased temperature and sea-sonal precipitation changes. Journal of the American WaterResources Association, 31 (2), 283–294, doi:10.1111/j.1752-1688.1995.tb03380.x.

Poiani, K.A., et al., 1996. Climate change and Northern Prairiewetlands: simulations of long-term dynamics. LimnologyOceanography, 41 (5), 871–881.

Rango, A. and Martinec, J., 1995. Revisiting the degree-day methodfor snowmelt computations. Journal of the American WaterResources Association, 31 (4), 657–669.

Rosenberry, D.O. and Winter, T.C., 1997. Dynamics of water-tablefluctuations in an upland between two prairie-pothole wetlandsin North Dakota. Journal of Hydrology, 191 (1–4), 266–289.

SCS (US Soil Conservation Service), 1972. SCS National engi-neering handbook, Section 4. Hydrology. Washington, DC:Government Printing Office.

Sethre, P.R., Rundquist, B.C., and Todhunter, P.E., 2005. Remotedetection of prairie pothole ponds in the Devils Lake Basin,North Dakota. GIScience & Remote Sensing, 42 (4), 277–296.

Spence, C., 2007. On the relation between dynamic stor-age and runoff: a discussion on thresholds, efficiency,and function. Water Resources Research, 43, W12416,doi:10.1029/2006WR005645.

Stewart, R.E. and Kantrud, H.A., 1971. Classification of naturalponds and lakes in the glaciated prairie region. Washington,DC: US Government Printing Office, US Fish and WildlifeService Resource Publication 92.

Su, M., Stolte, W.J., and van der Kamp, G., 2000. Modelling Canadianprairie wetland hydrology using a semi-distributed streamflowmodel. Hydrological Processes, 14 (14), 2405–2422.

Swanson G.A., et al., 2003. Dynamics of a prairie pothole wet-land complex: implication for wetland management. In: T.C.Winter, ed. Hydrological, chemical, biological characteristicsof a prairie pothole wetland complex under highly variable

climate conditions: the Cottonwood Lake Area, east-centralNorth Dakota. US Geological Survey, Professional Paper 1675,55–94.

van der Kamp, G. and Hayashi, M., 1998. The groundwater rechargefunction of small wetlands in the semi-arid Northern Prairies.Great Plains Research, 8, 39–56.

van der Kamp, G. and Hayashi, M., 2009. Groundwater–wetlandecosystem interaction in the semiarid glaciated plains of NorthAmerica. Hydrogeology Journal, 17 (1), 203–214.

van der Kamp, G., Stolte, J., and Clark, R.G., 1999. Drying out ofsmall prairie wetlands after conversion of their catchments fromcultivation to permanent brome grass. Hydrological SciencesJournal, 44 (3), 387–397.

van der Kamp, G.W., Hayashi, M., and Gallén, D., 2003. Comparingthe hydrology of grassed and cultivated catchments in thesemi-arid Canadian prairies. Hydrological Processes, 17 (3),559–575.

van der Valk, A.G., 2005. Water-level fluctuations in North Americanprairie wetlands. Hydrobiologia, 539 (1), 171–188.

Voldseth R.A., et al., 2007. Model estimation of land-use effectson water levels of Northern Prairie wetlands. EcologicalApplication, 17 (2), 527–540.

Whigham, D.F. and Jordan, T.E., 2003. Isolated wetlands and waterquality. Wetlands, 23 (3), 541–549.

Winter, T.C. and Rosenberry, D.O., 1995. The interaction of groundwater with prairie pothole wetlands in the Cottonwood Lakearea, east-central North Dakota, 1979–1990. Wetlands, 15 (3),193–211.

Winter, T.C. and Rosenberry, D.O., 1998. Hydrology of PrairiePothole wetlands during drought and deluge: a 17-year studyof the Cottonwood Lake wetland complex in North Dakota inthe perspective of longer term measured and proxy hydrologicalrecords. Climatic Change, 40 (2), 189–209.

Winter, T.C., et al., 2001. Water source to four US wetlands:Implications for wetland management. Wetlands, 21 (4),462–473.

Woo, M.K. and Rowsell, R.D., 1993. Hydrology of a prairie slough.Journal of Hydrology, 146, 175–207.

Woo, M.K. and Winter, T.C., 1993. The role of permafrost and sea-sonal frost in the hydrology of northern wetlands in NorthAmerica. Journal of Hydrology, 141 (1–4), 5–31.

Wright, C.K., 2010. Spatiotemporal dynamics of prairie wetland net-works: power-law scaling and implications for conservationplanning. Ecology, 91 (7), 1924–1930.

Zhang, B., Schwartz, F.W., and Liu, G., 2009. Systematics inthe size structure of prairie pothole lakes through droughtand deluge. Water Resources Research, 45, W044210, 12.doi:10.1029/2008WR006878.

Dow

nloa

ded

by [

Flor

ida

Inte

rnat

iona

l Uni

vers

ity]

at 1

6:14

24

Sept

embe

r 20

13