journal of hydrology et al... · however, these water balance studies were computed using either...

Journal of Hydrology 501 (2013) 213–226

Contents lists available at ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier .com/ locate / jhydrol

Temporal and spatial variability of groundwater recharge on Jeju Island,Korea

0022-1694/$ - see front matter Published by Elsevier B.V.http://dx.doi.org/10.1016/j.jhydrol.2013.08.015

⇑ Corresponding author. Present address: Pacific Islands Water Science Center,U.S. Geological Survey, 677 Ala Moana Blvd., Suite 415, Honolulu, HI 96813, USA.Tel.: +1 808 587 2402.

E-mail address: [email protected] (A. Mair).

Alan Mair a,b,⇑, Benjamin Hagedorn c, Suzanne Tillery d, Aly I. El-Kadi a,b, Stephen Westenbroek e,Kyoochul Ha f, Gi-Won Koh g

a Water Resources Research Center, University of Hawaii at Manoa, 2540 Dole Street, Holmes Hall 283, Honolulu, HI 96822, USAb Department of Geology and Geophysics, University of Hawaii at Manoa, 1680 East-West Road, POST 701, Honolulu, HI 96822, USAc Department of Geological Sciences, California State University Long Beach, CA 90840, USAd International Boundary & Water Commission, 4171 North Mesa, Suite C-100, El Paso, TX 79902, USAe Wisconsin Water Science Center, U.S. Geological Survey, 8505 Research Way, Middleton, WI 53562, USAf Groundwater Department, Geologic Environment, Korea Institute for Geoscience and Mineral Resources, 124 Gwahang-no, Yuseong-gu, Daejeon 305-350, Republic of Koreag Jeju Special Self-Governing Province Development Corporation, 1717-35 Namjo-ro, Jocheon-eup, Jeju City 695-960, Republic of Korea

a r t i c l e i n f o

Article history:Received 2 May 2013Received in revised form 26 July 2013Accepted 10 August 2013Available online 19 August 2013This manuscript was handled byKonstantine P. Georgakakos, Editor-in-Chief,with the assistance of Ellen Wohl, AssociateEditor

Keywords:Groundwater rechargeIsland of JejuSoil water balanceSWB model

s u m m a r y

Estimates of groundwater recharge spatial and temporal variability are essential inputs to groundwaterflow models that are used to test groundwater availability under different management and climate con-ditions. In this study, a soil water balance analysis was conducted to estimate groundwater recharge onthe island of Jeju, Korea, for baseline, drought, and climate-land use change scenarios. The Soil WaterBalance (SWB) computer code was used to compute groundwater recharge and other water balancecomponents at a daily time step using a 100 m grid cell size for an 18-year baseline scenario(1992–2009). A 10-year drought scenario was selected from historical precipitation trends(1961–2009), while the climate-land use change scenario was developed using late 21st century climateprojections and a change in urban land use. Mean annual recharge under the baseline, drought, and cli-mate-land use scenarios was estimated at 884, 591, and 788 mm, respectively. Under the baseline sce-nario, mean annual recharge was within the range of previous estimates (825–959 mm) and onlyslightly lower than the mean of 902 mm. As a fraction of mean annual rainfall, mean annual rechargewas computed as only 42% and less than previous estimates of 44–48%. The maximum historical reportedannual pumping rate of 241 � 106 m3 equates to 15% of baseline recharge, which is within the range of14–16% computed from earlier studies. The model does not include a mechanism to account for addi-tional sources of groundwater recharge, such as fog drip, irrigation, and artificial recharge, and may alsooverestimate evapotranspiration losses. Consequently, the results presented in this study represent aconservative estimate of total recharge.

Published by Elsevier B.V.

1. Introduction

Freshwater on oceanic islands is derived exclusively from pre-cipitation, and groundwater aquifers replenished by infiltratingrainfall are a primary source of freshwater. As examples, ground-water aquifers supply 80% of drinking water on the island of Guam(western Pacific Ocean), 92% of the freshwater on the island of Jeju(northwest Pacific Ocean), 98% of freshwater in the Azoresarchipelago (north Atlantic Ocean), 99% of drinking water in theHawaiian Islands (central Pacific Ocean), and virtually all of thefreshwater on the islands of Saipan (western Pacific Ocean) and

Tutuila (south Pacific Ocean) (Gingerich and Oki, 2000; Cruz,2003; Kim et al., 2003; Tribble, 2008). Given their isolation andlimited storage capacity, groundwater resources on oceanic islandsare particularly susceptible to overpumping (e.g., upconing), saltwater intrusion (e.g., sea level changes, recharge decline), contam-ination from agricultural and urban development, and climatevariability. An evaluation of the influence of these factors ongroundwater availability is critical for sustainable water resourcemanagement. Inter-annual and multi-decadal climate variabilityhas negatively impacted groundwater availability on several is-lands resulting in declining water levels and stream baseflow,and the thinning of basal freshwater lenses (Oki, 2004; Chu andChen, 2005; Presley, 2005; Bailey et al., 2009). Groundwater re-charge is a critical input for tools, such as groundwater flow mod-els, that investigate freshwater availability and quality in criticalisland aquifers under different demand, climate, and land cover

http://crossmark.crossref.org/dialog/?doi=10.1016/j.jhydrol.2013.08.015&domain=pdf

http://dx.doi.org/10.1016/j.jhydrol.2013.08.015

mailto:[email protected]


http://www.sciencedirect.com/science/journal/00221694

http://www.elsevier.com/locate/jhydrol

214 A. Mair et al. / Journal of Hydrology 501 (2013) 213–226

scenarios (Gingerich and Voss, 2005; Oki, 2005; Todd Engineers,2005; Gingerich, 2008; Gingerich and Jenson, 2010). Hence, spa-tially distributed estimates of recharge are essential for assessinggroundwater availability under changing land use, land cover,and climate.

Freshwater is a vital natural resource on the island of Jeju,Republic of Korea, with significant economic, agricultural, aes-thetic, ecologic, and cultural importance. Jeju has experienced tre-mendous growth over the past 30 years and now supports aresidential population of 583,000 and 8.7 million visitors annually(Jeju Special Self-Governing Province, 2013). As a consequence ofthe rapid growth and increased groundwater withdrawals, saltwa-ter intrusion has become a persistent problem particularly in theeastern coastal portion of the island (Kim et al., 2003; Kim et al.,2006). In addition, increased fertilizer applications since the late1970s and human waste have caused widespread nitrate ground-water contamination (Spalding et al., 2001; Koh et al., 2005; Kohet al.,2006a; Koh et al., 2009; Koh et al., 2012b). Finally, the fre-quency and intensity of extreme rainfall events across Korea,including Jeju, have increased significantly and show intensifica-tion of the annual monsoon (Choi et al., 2008; Jung et al., 2011;Park et al., 2011). These upward trends are projected to persistand continue well into the 21st century (Min et al., 2004; Booet al., 2006; Im et al., 2011). Collectively, these changes in demand,land use, and climate have serious implications for freshwateravailability on Jeju. Hence, robust estimates of groundwater re-charge are critical for assessing the ability of the island’s aquifersto meet current and future freshwater needs.

Groundwater recharge can be estimated using a variety of tech-niques including the soil–water balance, surface water, unsatu-rated zone, and saturated zone methods (Scanlon et al., 2002).Measurements needed to make these estimations are often diffi-cult, and there are varying levels of uncertainty and different spa-tio-temporal scales associated with each method. The soil-waterbalance method is considered a reliable recharge estimation tech-nique for broad, regional scale groundwater management purposesbecause it is not limited by any assumptions related to the mech-anisms that control the individual components. Due to its flexibil-ity, the soil–water balance method can be used to evaluate theeffects of climate and/or land use change across a wide range ofspace and time scales. The method can also produce time seriesof spatio-temporal recharge and is thus very useful for assessingtemporal trends. Hence, the method has been extensively usedon Pacific islands to estimate groundwater recharge (Giambelluca,1983; Izuka et al., 2005; Engott and Vana, 2007; Izuka et al., 2007;Engott, 2011; Johnson, 2012).

The water balance method follows a mass-balance procedurethat accounts for water entering, leaving, and being stored withina soil–plant control volume (Thornthwaite and Mather, 1955;Thornthwaite and Mather, 1957; Scanlon et al., 2002). Water thatinfiltrates below the soil–plant control volume (i.e., root zone) isoften called potential recharge to distinguish it from water thatreaches the actual water table or actual recharge (Rushton andWard, 1979). The distinction between potential and actual re-charge becomes important when the unsaturated zone is thick be-cause the time of travel to reach volcanic aquifers on oceanicislands can be on the order of years or decades (Voss and Wood,1993; Koh et al., 2012a). Oki (2008) found that the accounting or-der of recharge and evapotranspiration (ET) in soil–water balancemodels can result in large uncertainty in recharge estimates ifthe soil moisture storage capacity is small and the water balanceis computed using monthly time intervals, particularly in aridand semi-arid regions. Averaging water balance components overa longer time step, such as a month, tends to dampen out extremeprecipitation events that may be most responsible for contributingrecharge (Scanlon et al., 2002). Therefore, uncertainty in recharge

estimates computed using a water balance method can be mini-mized using the shortest computation interval that the data allow.

Over the past 15 years, several studies have estimated ground-water recharge on Jeju using a water balance approach as part ofefforts to assess freshwater availability (Hahn et al., 1997; KoreaWater Resources Corporation (KOWACO), 2003b; Won, 2004;Koh et al., 2006b). These studies report that mean recharge as afraction of rainfall varies from 44% to 48%. However, these waterbalance studies were computed using either monthly time stepsor limited precipitation datasets. Hagedorn et al. (2011) used watertable fluctuation and geochemical tracer methods to compute re-charge rates of 9–39% on Jeju, which suggests that previous waterbalance studies may have overestimated recharge. However, theirstudy may have biased low elevation-low rainfall areas, therebyexcluding high-rainfall portions of the island with potentiallymuch greater recharge rates. Thus, there is a need to test alterna-tive methods to further improve and refine the island’s total re-charge estimate.

A variety of soil water balance codes have been developed toestimate potential groundwater recharge (Giambelluca, 1983; Sch-roeder et al., 1994; Finch, 2001; Batelaan and De Smedt, 2007;Dripps and Bradbury, 2007; Flint and Flint, 2007; USGS, 2008;Westenbroek et al., 2010; Johnson, 2012). Most of these codesuse proprietary software, are implemented in a proprietary lan-guage, or are complicated to operate. The freely available SoilWater Balance (SWB) code was recently developed to estimatethe spatial and temporal distribution of natural groundwater re-charge in temperate-humid climates at a daily time step and at auser-specified grid resolution (Dripps and Bradbury, 2007;Westenbroek et al., 2010). A key advantage of SWB is the abilityto calculate recharge using commonly available geographic infor-mation system (GIS) data layers in combination with tabular cli-mate data. The SWB code has been successfully applied totemperate-humid climate areas in northern Wisconsin and aroundLake Michigan (Dripps and Bradbury, 2007; 2010; Feinstein et al.,2010; Westenbroek et al., 2010) but it has yet to be applied to oce-anic islands with similar climate.

In this study, the SWB code is applied to Jeju to estimate thespatial and temporal distribution of groundwater recharge andother water balance components including direct runoff, evapo-transpiration, and precipitation. The information is critical forassessing the sustainability of the island’s water resources undervarious climate and land-use conditions. The main specific objec-tives of the study are to (1) produce spatially and temporally dis-tributed estimates of groundwater recharge under current(baseline) conditions, hypothetical drought, and climate-land usechange scenarios, and (2) test the suitability of applying the SWBcode to oceanic islands stressing model and data limitations. Un-like previous applications of SWB, the model domain in this studyis comprised of mountainous terrain overlain with many highlypermeable lava flows and thick unsaturated zones. Hence, thestudy described herein offers insights into SWB’s use and limita-tions for applications in similar environments.

2. Study area

Jeju is located 85 km south of the Korean peninsula (33�N,126�E) and encompasses an area of 1828 km2 (Fig. 1) (KOWACO,2003b). The island is comprised of a dormant shield volcano withone central mountain peak, Mt. Halla, rising to an elevation of1950 m above sea level (masl). The climate varies from cool-tem-perate in winter to humid-monsoon in summer with mean air tem-peratures in the coastal areas ranging from 5 �C in January to 24 �Cin August. Mean annual rainfall across the island is 2082 mm andvaries from 1100 mm along the western coastline to over

Fig. 1. Map of Jeju location (inset), surface hydrologic units, rain and stream gauges, stream gauge drainage areas, and mean annual rainfall isohyets.

A. Mair et al. / Journal of Hydrology 501 (2013) 213–226 215

4000 mm near Mt. Halla (Mair et al., 2013). Roughly 67% of meanannual rainfall occurs from May to September, with the highestmonthly rainfall totals occurring in July and August (200–800 mm/month). Despite its maritime location, snow is notuncommon at higher elevations during the cool winter months.Coniferous/broadleaved forests and grassland blanket the interiorwhile croplands and orchards cover the coastal areas (Fig. 2a)(Chung, 2007; Korea Institute for Geoscience and Mineral Re-sources (KIGAM), 2010f). Evergreen broadleaved forest occupiesthe coastal plain up to an elevation of 600 masl, whereas deciduousbroadleaved forest occupies the zone from 600 to 1400 masl. Sub-alpine coniferous forest occurs above 1400 masl.

Soils on Jeju are relatively thin and are comprised mostly of siltand volcanic ash with a saturated hydraulic conductivity rangingfrom 0.02 to 0.55 m/d (Kang et al., 2010). Much of the soils inthe high rainfall areas around Mt. Halla and the eastern portionof the island are classified as having high infiltration capacity,while soils with lower infiltration capacity are mainly confinedto areas in the drier western portion of the island (Fig. 2b) (KIGAM,2010g). Few perennial streams exist on the island and occur mainlyon the southern and northern flanks of Mt. Halla (Won et al., 2006).Pleistocene to Holocene-aged volcanic rocks characterized by highpermeability and storage capacity underlie the soils and comprisethe main volcanic aquifers of the island (Hahn et al., 1997; Kimet al., 2003; Won et al., 2005; Koh et al., 2006a; Won et al.,2006). Groundwater occurs in unconfined high level, parabasal,and basal aquifers with hydraulic heads that vary from 180 maslin the high level interior aquifers to about 2 masl in the coastalparabasal and basal aquifers.

3. Methodology

A soil water balance analysis was conducted for Jeju to estimatethe spatial and temporal distribution of natural groundwaterrecharge using the SWB computer code (Westenbroek et al.,2010). The SWB code is a deterministic, quasi three-dimensionalphysically based model that uses readily available soil, land cover,topographic, and climatic data to estimate potential groundwater

recharge at a daily time step on a user-defined grid cell-by-grid cellbasis (Dripps and Bradbury, 2007; Dripps and Bradbury, 2010). Re-charge output from the SWB code can be aggregated into monthlyor annual values for importing to a groundwater flow model. TheSWB code also produces estimates of other water balance compo-nents including rainfall, snow, ET, and direct runoff.

SWB uses a modified Thornthwaite-Mather soil water balanceapproach (Thornthwaite and Mather, 1955; Thornthwaite andMather, 1957) to calculate natural groundwater recharge as theresidual of the water balance using the following mass balanceequation:

R ¼ P � I þ SNmelt þ DRin � DRout � ETsm � DS ð1Þ

where R, P, I, SNmelt, DRin, DRout, ETsm, and DS are recharge, gross pre-cipitation, interception, snowmelt, direct runoff into the grid cellfrom upslope grid cells, direct runoff out of the grid cell, soil mois-ture ET, and change in soil moisture, respectively. The term for soilmoisture ET, ETsm, is used to account for soil moisture evaporativelosses including evaporation from soil and plant transpiration. Thus,total ET, ETtot, may be computed as interception, I, plus soil moistureET, ETsm (ETtot = I + ETsm). Cloud water interception (or fog drip),agricultural irrigation, and artificial recharge are not explicitly in-cluded in the model and data for these additional inputs are not of-ten available. Land cover class and hydrologic soil groupinformation data were not available for some coastal areas. Hence,the island boundary had to be reduced slightly to obtain a modeldomain with a common extent for each of the model input grids.As a result, the model domain encompassed an area of 1817 km2,which is 11 km2 or 0.6% less than the actual area of 1828 km2 (com-pare Figs. 1 and 2). We used a grid cell size of 100 m in this study.

Gross precipitation, P, was distributed using a regularized splinealgorithm and daily measurements from a 52-gauge network(Fig. 1) for the 18-year period from January 1, 1992 to December31, 2009 (6575 days) (KIGAM, 2010c). A spline algorithm wasselected to interpolate daily rainfall instead of geostatistical ap-proaches, such as kriging, to eliminate the need for daily variogrammodeling. Interception, I, was estimated using a bucket modelapproach in which a daily initial interception storage capacity

Fig. 2. (a) Land use/land cover map (KIGAM, 2010a), (b) hydrologic soil group (KIGAM, 2010b), (c) map of mean annual recharge during baseline period 1992–2009.


must be satisfied before precipitation can reach the soil surface asnet precipitation, Pnet, where Pnet = P � I. Net precipitation, Pnet, wasconverted into snowfall when the daily average temperature (Tave)

minus one-third the difference between the daily maximum tem-perature (Tmax) and daily minimum temperature (Tmin) was 60 �C.Snowfall that accumulates on the ground surface was converted


into snowmelt at a rate of 1.5 mm/�C d when Tmax > 0 �C (Westenb-roek et al., 2010). To the authors’ knowledge, no direct measure-ments of interception losses on Jeju have been reported in theliterature. Therefore, seasonal target interception loss rates forthe different types of land cover found on Jeju were selected fromother studies in mainland Korea, Japan, and other parts of theworld (Table 1). The model allows for seasonally-variable intercep-tion losses by dividing each year into a growing season when inter-ception storage capacities are high and a non-growing seasonwhen interception storage capacities are low.

The growing season length (GSL) was used to define one startand end date for the growing season for the entire island. TheGSL was estimated using two separate methods based on dailymeasurements of Tave and Tmin. First, the GSL was computed asthe period between the first span of at least six days after January1 with Tave > 5 �C, and the first span after July 1 of six days withTave < 5 �C (WCRP, 2010). Next, the GSL was also computed as thenumber of days between the last and first occurrence of Tmin = 0 �Cduring each year. Measurements of Tave and Tmin were recordedfrom 1961–2009 at Jeju City and Seogwipo City (Fig. 1) (KIGAM,2010a). We adjusted Tave and Tmin to coincide with the transitionalboundary between evergreen and deciduous forest, at approxi-mately 600 m (KIGAM, 2010f), using seasonally-variable environ-mental lapse rates (ELRs) of Tave and Tmin. The GSL computed at600 m was used to approximate the start and end of the growingseason for the entire island. Seasonally-variable ELRs for Tave, Tmin,and Tmax were computed using linear regression and measure-ments from a network of 19 gauges from January 1, 2006 to

Table 1Interception storage capacities and loss rates.

Land cover type Growing seasona

Daily interceptionstorage capacityc (mm/d)

SWB modelinterception lossd

Imodel (%)

TargetinterceptionItarget (%)

Agriculturalfacility

1.0 5 5

Agricultural land/cropland

3.4 15 15o

Deciduousbroadleavedforest

7.8 20 20h,i,j,k

Evergreenbroadleavedforest

6.2 20 20h,i,j,k

Facility land 1.1 5 5Grass 2.5 10 10m,n

Open land areas 1.0 5 5Orchard 3.9 15 15o

Roads 0.0 0 0Subalpine

coniferousforest

7.0 15 15f,g,h,l

Urban/residential 0.2 1.1 1p

a Growing season from April 1–November 30.b Non-growing season from December 1–March 31.c Calibrated quantity of maximum gross daily precipitation that is intercepted and evd Net interception loss predicted by SWB model expressed as fraction of gross precipie Interception losses approximated from literature review and expressed as fraction of Kim and Woo (1988),g Min and Woo (1995).h Lee et al. (1997).i Silva and Okumura (1996).j Deguchi et al. (2006).k Toba and Ohta (2005).l Shimizu et al. (2003).

m Corbett and Crouse (1968).n Thurow et al. (1987).o Calheiros de Miranda and Butler (1986).p Xiao et al. (1998).

December 31, 2009 and extending from sea level to 1671 m(Fig. 1) (KIGAM, 2010b).

Direct runoff, DRout, was estimated using the curve numbermethod for the 12 land cover classes and four hydrologic soilgroups mapped on Jeju (USDA-NRCS, 2003; KIGAM, 2010f; KIGAM,2010g). DRout was routed to downslope cells as DRin and allowed tocontribute to infiltration so that all runoff either infiltrated or wasrouted out of the model domain on the same day it originated.Other oceanic island water balance studies have used observedmonthly mean runoff-to-rainfall ratios to estimate DRout (Izukaet al., 2005; Engott and Vana, 2007; Engott, 2011; Johnson, 2012)Thus, the ability of SWB to route DRout to downslope cells repre-sents a potential enhancement over approaches that rely on empir-ical runoff-to-rainfall relationships. Maximum infiltration ratesassigned to each hydrologic soil group were used by SWB to specifya maximum daily recharge rate, R, and were estimated from therange of saturated hydraulic conductivity measurements reportedby Kang et al. (2010) (Table 2).

Potential ET (ETpot) was computed at a daily time step using theHargreaves–Samani method (Hargreaves and Samani, 1985). Themethod is one of five methods currently used by SWB to computeETpot. However, the Hargreaves–Samani method is the only methodused by SWB that is capable of producing spatially-variable esti-mates of ETpot. To make this computation, SWB requires spatiallyvarying input grids of Tmin and Tmax for each daily time step. Spa-tially-variable grids of Tmin and Tmax were estimated for the entireisland using seasonally-variable ELRs, measurements of Tmin andTmax at Jeju City, and a digital elevation model (DEM) of Jeju

Non-growing seasonb

losseDaily interceptionstorage capacityc (mm/d)

SWB modelinterception lossd

Imodel (%)

Targetinterception losse

Itarget (%)

0.2 2.7 2.5

0.0 0 0

1.9 15 15j,k,l

2.4 20 20k

0.2 2.7 2.51.0 10.1 10m,n

0.2 2.6 2.50.9 10.1 100.0 0 02.5 15.1 15g,l

0.1 1.2 116

aporated back to atmosphere.tation.f gross precipitation.

Table 2Maximum infiltration rates.

Hydrologic soil group Max infiltration ratea (m/d)

A 0.28B 0.20C 0.13D 0.05

a Estimated as roughly 80th, 60th, 40th, and 20th quintiles for soil groups A–D,respectively, from Kang et al. (2010).


(Fig. 1) (KIGAM, 2010a; KIGAM, 2010d). ETsm was computed at eachgrid cell as follows: (1) when Pnet � ETpot P 0, then ETsm = ETpot, (2)when Pnet � ETpot < 0, then ETsm equals only the amount of waterthat can be extracted from the soil as computed using the soilmoisture retention tables of Thornthwaite and Mather (1957)and modified by Westenbroek et al. (2010). Estimates of maximumsoil moisture storage capacity needed to use the soil moistureretention tables were computed as the product of the availablewater soil capacity times root zone depth (Table 3). Available watersoil capacity and root zone depth for different types of vegetationwere selected from Thornthwaite and Mather (1957) as modifiedby Westenbroek et al. (2010) (see Table 10 in Westenbroek et al.(2010)) using the following hydrologic soil group-soil type pair-ings: (1) group A – fine sand, (2) group B – fine sandy loam, (3)group C – silt loam, and (4) group D – clay loam. ArcMap™ soft-ware was used to conduct the rainfall interpolation, prepare thedaily input grids of Tmin, Tmax, and P, generate all of the other inputgrids, and post-process model output (ESRI, 2010).

3.1. Model calibration

For the baseline scenario, model calibration was conducted forinterception and direct runoff. Calibration for interception wasused to ensure the user-specified daily interception storage capac-ities for each land cover class matched the seasonal target inter-ception loss rates (Table 1). The daily interception storagecapacities were adjusted iteratively to match the seasonal targetinterception loss rates for each type of land cover. First, the modelwas repeatedly run (i.e., trial and error approach) using differentdaily interception storage capacities until we achieved reasonableagreement between the model-predicted interception loss rates,Imodel, and the target interception loss rates, Itarget (|Imodel � Itarget| -6 0.2%). However, storage capacities for roads, rivers, reservoirs,and other open water bodies were set to 0.0 mm/d under all con-ditions and excluded from adjustment during model calibration.Tree and plant phenology may respond to drought or climatechange conditions, which could induce changes in canopy storagecapacities. However, for simplicity, we assumed that changes intree and plant phenology (e.g., lower or higher leaf or plant area in-dex) under drought or climate change conditions translate intominimal changes in interception storage capacities. Hence, theinterception storage capacities used in the final run of this firstphase of calibration were used in all subsequent model runs.

During the second phase of model calibration, the model wasrun repeatedly using different curve numbers until we achievedreasonable agreement between direct runoff predicted by SWB,DRmodel, and measured direct runoff at four (4) perennial streamgauge locations, DRobserved (|(DRmodel � DRobserved)/DRobserved| -� 100% 6 1%) (Fig. 1). The SWB model is not capable of predictinggroundwater discharge to streams or springs, so there was no at-tempt to match the baseflow in any of the island’s streams orsprings. Therefore, the streamflow datasets were pre-processedby KIGAM to remove the baseflow component of streamflow (KI-GAM, 2010e). The four gauges had variable operating periods asfollows: (1) 2006–2009 (gauges 1 and 3), 2007–2009 (gauge 2),

and 2009 only (gauge 4). Because of the limited amount of concur-rent monitoring data, we chose to use the combined cumulativeflow from 2006 to 2009 at all four gauges for model calibration.Curve numbers for all land cover classes, except roads and openwater, were uniformly adjusted iteratively from their original val-ues until reasonable agreement was reached between the com-bined total observed direct runoff flow and the predicted totaldirect runoff from the four stream sites (Table 4). Following thecalibration process, the model was run for a baseline scenario timeperiod from January 1, 1992 to December 31, 2009.

3.2. Drought scenario

Mair et al. (2013) investigated meteorological drought on Jejuduring 1961–2009 and identified a 10-year drought period during1963–1972, which contained three severe droughts each lastingfrom 22 to 26 months. We used this same 10-year period to assessthe impact of drought on island-wide recharge. We created grids ofdaily rainfall by scaling the mean monthly rainfall maps describedby Mair et al. (2013) by the daily rainfall measurements at the JejuCity and Seogwipo City gauges. The Jeju City gauge was used toscale rainfall across the northern watersheds from Hangyeong toGujwa, while the Seogwipo City gauge was used for the southernwatersheds from Daejeong to Seongsan (Fig. 1). The scaling factorfor each area was computed as the daily rainfall measurement di-vided by the mean monthly rainfall (e.g. factor for day 1 = January1, 1963 daily rainfall at Jeju City divided by mean monthly rainfallfor January at Jeju City gauge). The mean monthly rainfall mapswere then multiplied by the scaling factors determined from eachgauge (i.e. Jeju scaling factor applied to northern watersheds andSeogwipo scaling factor applied to southern watersheds). Grids ofTmax and Tmin were prepared using temperature monitoring datafrom January 1, 1963 to December 31, 1972 at Jeju City and themonthly ELRs. The scaled areas were then merged for input tothe SWB model. The model was then run for the 10-year simula-tion period while keeping all other input parameters the same asin the baseline scenario.

3.3. Climate – land use change scenario

A climate-land use change scenario was created to simulateprojected changes in surface temperature and land use on ground-water recharge on Jeju. Changes in temperature were simulatedusing a change factor approach (Hay et al., 2000; Diaz-Nieto andWilby, 2005). Studies of multi-decadal temperature trends onother high-relief islands, such as the Hawaiian Islands, indicatethat the rate of warming can vary with elevation (Giambellucaet al., 2008). However, the lack of long-term monitoring data athigher elevations on Jeju prohibits a similar analysis. Therefore,we assumed no future changes in the ELRs for Tave, Tmin, and Tmax.First, observed changes in the mean for Tave, Tmin, and Tmax duringthe baseline period (1992–2009) were computed relative to the1961–1990 mean at two long-term monitoring sites, Jeju Cityand Seogwipo City. Next, multi-model ensembles (MMEs) of sevencoupled atmosphere-ocean general circulation models (AOGCMs)of future increases in Tave under an A2 emission scenario as re-ported by Min et al. (2004) were used to compute projected in-creases in mean Tave near Jeju during 2070–2099 relative to the1961–1990 mean. An A2 emission scenario was selected becauseit represents a more extreme future climate condition for testingchanges in recharge. The annual change factor for Tave was thencomputed as the difference between the projected changes inmean Tave during 2070–2099 and the observed changes in meanTave during 1992–2009. A discussion of inter-model and MMEuncertainty is provided in Min et al. (2004).

Table 3Root zone depth and maximum soil moisture storage capacity from Westenbroek et al. (2010).

Land cover type Hydrologic soil group

A B C D

Root zonedepth (m)

Max soil moisturestorage capacity(mm)

Root zonedepth (m)


Root zonedepth (m)


Root zonedepth (m)


Agriculturalfacility

0.50 50 0.50 75 0.62 124 0.40 100

Agricultural land/cropland

0.50 50 0.50 75 0.62 124 0.40 100

Deciduousbroadleavedforest

2.50 250 2.00 300 2.00 400 1.60 400

Evergreenbroadleavedforest

2.50 250 2.00 300 2.00 400 1.60 400

Facility land 0.50 50 0.50 75 0.62 124 0.40 100Grass 1.00 100 1.00 150 1.25 250 1.00 250Open land areas 0.50 50 0.50 75 0.62 124 0.40 100Open water 0 0 0 0 0 0 0 0Orchard 1.50 150 1.67 251 1.50 300 1.00 250Rivers/reservoirs 0.50 50 0.50 75 0.62 124 0.40 100Roads 0.60 60 0.60 90 0.60 120 0.60 150Subalpine

coniferousforest

2.50 250 2.00 300 2.00 400 1.60 400

Urban/residential 0.60 60 0.60 90 0.60 120 0.60 150

Available water capacity: Group A = 100 mm/m; group B = 150 mm/m; group C = 200 mm/m; group D = 250 mm/m.

Table 4Adjusted curve numbers following baseline model calibration.

Land cover type Hydrologic soil group

A B C D

Agricultural facility 51 64 71 75Agricultural paddy/cropland 57 66 73 77Deciduous broadleaved forest 31 52 64 69Evergreen broadleaved forest 31 52 64 69Facility land 67 75 79 82Grass 43 60 69 73Open land 43 60 69 73Open water 100 100 100 100Orchard 49 65 75 79Rivers/reservoirs 43 60 69 73Roads 98 98 98 98Subalpine coniferous forest 31 52 64 69Urban/residential 89 92 94 95


Observed changes in mean Tmin and Tmax between the periods1992–2009 and 1961–1990 were used to scale the annual changefactor for mean Tave into projected changes for mean Tmin and Tmax

during 2070–2099. Min et al. (2004) reported that projected in-creases in Tave across east Asia are expected to vary seasonally withthe greatest change occurring during the winter months of Decem-ber, January, and February. Hence, we used their seasonally-vari-able projected increases in Tave to transform the annual changefactors into seasonally-adjusted change factors for Tmin and Tmax.The seasonally-adjusted change factors for Tmin and Tmax were thenapplied to the daily grids of Tmin and Tmax from our baseline datasetto produce a new 18-year dataset. The adjusted daily Tave, Tmin, andTmax measurements at Jeju City and Seogwipo City (adjusted to600 m elevation) were used to re-compute the GSL as describedearlier.

Min et al. (2004) reported that mean annual rainfall around Jejuduring 2070–2099 is expected to increase by 5% relative to the1961–1990 mean. They also noted that the seasonal rainfall cyclewill be further concentrated in the monsoon season. Mair et al.(2013) reported that mean annual rainfall at Jeju City and

Seogwipo City during 1992–2009 increased by an average of 6%relative to the 1961–1990 mean (3% at Jeju City and 8% at Seog-wipo City). They reported that monsoon rainfall contributed lessto annual rainfall in June and July but more in August and Septem-ber during 1992–2009 relative to the 1961–1990 mean. However,they also noted that the seasonal rainfall cycle during 1992–2009exhibits more concentrated rainfall during the annual monsoonthan projected by Min et al. (2004). Thus, the rainfall regime during1992–2009 already exhibits changes in excess of that predicted forthe late 21st century. Therefore, we made no adjustments to the18-year rainfall baseline dataset for the climate-land use changescenario.

Land use change was incorporated by converting the type de-scribed as facility land to urban/residential (Fig. 2a) and updatingthe affected parameterizations (Tables 1, 2 and 4). The reclassifica-tion eliminated the facility land land-use type and increased themore impervious urban/residential area in the model by 46%. Themodel was then run for an 18-year simulation period while keep-ing all other input parameters the same as the baseline scenario.

4. Results and discussion

4.1. ELR and GSL assessment

The magnitude of the monthly ELRs for Tave, Tmin, and Tmax aregreatest during the months of November, December, and January(�7.0 to �7.5 �C/1000 m) and lowest during the months of May,June, and July (�3.7 to�5.1 �C/1000 m) (Table 5). The seasonal var-iability of monthly ELRs indicates that usage of annual ELRs forestimating Tave, Tmin, and Tmax distributions on Jeju would be inad-equate for capturing temperature dynamics during a seasonal cy-cle. Therefore, we used the monthly ELRs to prepare the modelinput grids of Tmin and Tmax, and conduct the GSL analyses. The re-sults of the GSL analyses indicate that the growing season at 600 mbegan between days 83 and 96, and ended between days 319 and339 of each calendar year during 1992–2009 (Table 6). The meanGSL computed using the Seogwipo City station was 16 days longer

Table 5Environmental lapse rates (ELRs) and change factors for Tmin, Tmax, and Tave.

Month ELR (�C/1000 m) Change factor (�C)

Tmin Tmax Tave Tmin Tmax Tave

January �7.50 �7.00 �7.28 4.10 2.50 3.30February �7.39 �6.89 �7.17 4.10 2.50 3.30March �6.44 �6.00 �6.22 3.85 2.25 3.05April �6.06 �5.17 �5.61 3.60 2.00 2.80May �5.11 �3.94 �4.56 3.35 1.75 2.55June �4.72 �3.72 �4.22 3.10 1.50 2.30July �5.06 �4.72 �4.89 2.85 1.25 2.05August �6.22 �6.22 �6.22 2.85 1.25 2.05September �6.56 �5.78 �6.17 3.10 1.50 2.30October �7.11 �6.22 �6.67 2.60 1.00 1.80November �7.28 �7.17 �7.22 3.10 1.50 2.30December �7.28 �7.28 �7.28 4.10 2.50 3.30


than the Jeju City station during 1992–2009, which suggests somenon-uniform seasonal variation in temperature across the island.For this study, we selected the average start and end dates asday 90 and day 328, respectively, for the growing season. Duringa non-leap year, day 90 corresponds to March 31 and day 328 cor-responds to November 24. Model output can be produced by SWBat daily, monthly, and annual time steps. As a simplification formodel calibration, we used monthly model output to calibratemodeled interception loss rates with target interception loss rates(Table 1). The GSL was then modified slightly to align with startand end dates of a calendar month. Thus, we defined the growingseason as extending from April 1 to November 30 (days 91–334),and the non-growing season as extending from December 1 toMarch 31 for the baseline scenario. The GSL of 244 d(334 � 90 = 244) used for Jeju is 68 d longer than the mean GSLof 176 d on peninsular South Korea (Jeong et al., 2013), which isto be expected given Jeju’s location further south and warmer mar-itime climate.

For the climate–land use change scenario, the GSL analysesindicate the growing season at 600 m will begin around day 65and end on day 345 of each calendar year (Table 6). However,

Table 6Summary of growing season length at 600 m.

Period Method Mean Std. dev.

Tavea Tmin

b

Base station

Jeju Seogwipo Jeju Seogwipo

Start of growing season1961–2009 101 91 101 95 97 51961–1990 104 95 105 98 101 51992–2009 96 83 94 88 90 62070–2099 73 50 69 68 65 10

End of growing season1961–2009 330 336 314 317 324 111961–1990 329 333 310 314 321 111992–2009 332 339 319 323 328 92070–2099 347c 357c 337 337 345 10

Length of growing season1961–2009 230 245 213 223 227 141961–1990 225 238 205 215 221 141992–2009 236 257 224 235 238 132070–2099 274 307 269 270 280 18

a Start computed as first span after January 1 of at least 6 days with Tave > 5 �C;end computed as first span after July 1 of 6 days with Tave < 5 �C.

b Start computed as last day after January 1 where Tmin < 0 �C; end computed asfirst day after July 1 where Tmin < 0 �C.

c Years computed to have no end of growing season were set to 365 days foraveraging.

much greater variability in the growing season is predicted withthe GSL method based on Tave, including many years with no com-puted end to the growing season. Therefore, we chose the begin-ning and end date of the growing season predicted by the GSLmethod based on Tmin, which indicates a beginning and end dateof day 68 (March 9) and day 337 (December 4), respectively, or aGSL = 270 d. Thus, under our climate-land use change scenario,the GSL lengthens by 26 d (270 - 244 = 26) when compared tothe baseline scenario (1992–2009). and 55 d (270 - 215 = 55) whencompared to the 1961-1990 mean. Likewise, the mean change inthe GSL for the northeastern USA (2070-2099 minus 1961-1990),a region similarly positioned geographically relative to a large con-tinental land form, derived from multi-model simulations usingnine AOGCMs for an A2 emission scenario is estimated to be43 d (Hayhoe et al. 2007). Since 1961, the GSL at Jeju City and Seog-wipo City has increased at highly significant rates of 3.2 and 4.5 d/decade, respectively (Mair et al., 2013), which is consistent withthe observed mean increase in the GSL of 4.5 d/decade across pen-insular South Korea since 1982 (Jeong et al., 2013). As a result, themean GSL on Jeju (adjusted to 600 m) has already lengthened by19 d over the last 50 years (1992–2009 minus 1961–1990;Table 6) . Since the SWB code divides each year into a growingand non-growing season for assigning daily interception storagecapacities, a longer GSL effectively increases the potential forgreater annual interception losses.

4.2. Model calibration

Model calibration yielded daily interception storage capacitiesranging from 0.2 mm/d for urban/residential to 7.8 mm/d fordeciduous broadleaved forests for the growing season (Table 1).During the non-growing season, the storage ranged from 0.0 mm/d for agricultural land/cropland to 2.4 mm/d for evergreen broad-leaved forest. Through adjustment of daily interception storagecapacity values, the model achieved very good agreement betweenthe predicted interception losses and the target interception lossesfor each land cover type during the growing season and non-grow-ing season. Mean annual interception loss, I, was estimated at302 mm or 14% of gross precipitation, P, during the 18-year base-line period. The target interception losses were entirely based onstudies conducted at other locations. Thus, despite good agreementbetween model predictions and target values, the accuracy of theseestimated interception losses could not be verified with fieldobservations from Jeju.

During the second phase of calibration, the combined total sim-ulated direct runoff volume at the four stream gauges (109 mm)during 2006–2009, the direct runoff monitoring period, matchedvery well with the observed combined total direct runoff(108 mm) (Fig. 3). However, comparison of the mean annual flowsat each gauge indicates the model overestimates flows at gauges 1and 3, and underestimates flows at gauges 2 and 4. Inter-annualvariability shows large differences between simulated and ob-served flows, particularly at gauge 2 (see Table S1 in Supplemen-tary material). The large differences at gauge 2 appear related, inpart, to the under-representation of rainfall occurring in the drain-age area of the West Jeju watershed. The four stream gauges mea-sured runoff from drainages within the Middle Seogwipo, WestSeogwipo, West Jeju, and Hallim watersheds (Fig. 1). Mean annualrainfall in each drainage area during the four-year monitoring per-iod, 2006-2009, varied from 1396 mm within the drainage of gauge1 to 2654 mm within the drainage of gauge 3. Annual runoff-to-rainfall ratios were computed for each of the four gauged drainageareas by dividing the observed and simulated mean annual runoffby the mean annual rainfall in each drainage area (Table S1). Theobserved inter-annual runoff-to-rainfall ratios at these four gaugesranged from 0.9% in 2009 (gauge 1) to 12% in 2007 (gauge 1). The

Fig. 3. Observed and simulated direct runoff at four stream gauges. See Table S1 fordetailed breakdown.

Fig. 4. Annual soil water balance estimates from baseline, drought, and land use-climate change scenarios, and from earlier studies. See Table S2 for detailedbreakdown.


simulated ratios showed even greater variability and ranged from0.3% in 2009 (gauge 2) to 18% in 2007 (gauge 1). Soil water balancestudies on other volcanic islands have reported mean annual is-land-wide runoff-to-rainfall ratios ranging from 13% (Guam) to42% (Kauai) (Shade, 1995; Shade and Nichols, 1996; Engott andVana, 2007; Johnson, 2012). On the island of Hawaii, runoff-to-rainfall ratios as low as 1% have been reported in dry portions ofthe island (Engott, 2011). Thus, the runoff-to-rainfall ratios re-ported on Jeju are relatively low but mostly within the range of ra-tios reported on other volcanic islands. While these results suggestthat SWB was capable of capturing differences in long-term runoffgeneration between watersheds, the model was not capable of cap-turing the annual and sub-annual variability between simulatedand observed flows. Uncertainty surrounding the use of appropri-ate curve numbers may also be another contributing factor. Wa-tershed modeling at the watershed-scale using these flow datacould be used to help better understand runoff generation and fur-ther refine the curve numbers used in a water balance model suchas SWB.

4.3. Recharge

Mean annual precipitation during the baseline scenario (1992–2009) was computed as 2104 mm with losses of 1034 mm for totalET (49%), 185 mm for direct runoff (9%), and 884 mm for ground-water recharge (42%) (Fig. 4). The recharge spatial distribution gen-erally followed a pattern similar to rainfall with the greatestrecharge occurring in the wet southern flank of Mt. Halla, and lessrecharge in the drier western portion of the island (Fig. 2c). Of thethree recharge scenarios analyzed in this study, the baseline sce-nario produced the highest mean island recharge (884 mm), fol-lowed by the climate-land use change scenario (788 mm), andthen the drought scenario (591 mm) (Fig 5). When compared tothe baseline scenario, estimates of mean island recharge for thedrought scenario and the climate-land use change scenario werelower by 33% and 11%, respectively.

Previous water balance estimates of total island groundwaterrecharge have ranged from 825 to 959 mm with a mean of902 mm. Thus, the estimated mean recharge is well within therange of previous recharge estimates and differs from the mean va-lue of the four prior estimates by only 18 mm (2%). When com-pared to the baseline scenario mean of 884 mm, the fourprevious estimates range from 6.7% less recharge (Hahn et al.,1997) to 8.5% more recharge (KOWACO, 2003b). The lower Hahnet al. (1997) estimate of mean recharge is characterized by lowermean input rainfall (11% lower). In addition, their study reported

higher and lower fractions of rainfall being converted to direct run-off and ET, respectively, when compared to the baseline scenario.Similarly, the higher KOWACO (2003b) estimate is also character-ized by lower mean rainfall (6% lower) as well as higher (lower)fractions of rainfall being converted to direct runoff (ET). Mean an-nual recharge varies three-fold across the island from a low of384 mm in the western district to a high of 1204 mm in the south-ern district (see Table S2 in Supplementary material). Recharge inthe western district was lower than estimates from earlier studies(556–564 mm), while recharge in the southern district was higherthan previous estimates (836–1088 mm) (Hahn et al., 1997; KOW-ACO, 2003a; Koh et al., 2006b). Estimates of recharge in the north-ern and eastern districts of the island were within the range ofearlier estimates. Inter-annual recharge ranged from 419 to1514 mm and also closely followed inter-annual rainfall trends(Fig. 6). Similarly, inter-annual recharge as a fraction of rainfall var-ied from 32% to 51%. Recharge under deciduous broadleaved forestwas highest among the three dominant unirrigated land cover clas-ses for soil groups A, B, and C (Fig. 7). These results clearly indicatethat both land cover and soil type play significant roles in control-ling natural recharge rates on Jeju.

The precipitation dataset used in this study represents a moreextensive monitoring period and greater number of monitoringpoints when compared to the previous studies, particularly aroundMt. Halla, which explains the differences in mean rainfall. In addi-tion, the SWB model does not include inputs for irrigation, fog drip,or artificial recharge, which may be contributing significantly toactual recharge in selected areas on Jeju (Choi and Lee, 2012; Moonet al., 2012). Thus, the results under each of the three scenariosrepresent an estimate of natural recharge. Incorporating additionalsources of recharge into a soil water balance modeling analysiscould be used to further refine the soil water balance analysisand could increase overall groundwater recharge estimates forJeju. However, the extent to which these sources could enhance re-charge on Jeju is not known. Therefore, the results presented here-inrepresent a conservative estimate of total groundwater recharge.Actual recharge rates are likely higher than reported in this studybecause of additional inputs of irrigation, fog drip, and artificialrecharge.

Information on pumping rates in recent years is lacking. How-ever, Hahn et al. (1997) reported a total pumping rate of241 � 106 m3/yr in 1993. As a fraction of estimated recharge,the pumping rate reported in 1993 is 15% of baseline recharge,which is within the range computed from earlier studies (14–16%). Given the extensive growth over the last two decades, it

Fig. 5. Mean annual recharge under (a) drought scenario based on rainfall during 1963–1972, (b) climate change scenario based on projected changes in temperature andrainfall for late 21st century.

Fig. 6. Time series plot of annual recharge, ratio of annual rainfall to mean annual rainfall, and annual recharge to annual rainfall.


Fig. 7. Mean annual recharge for different combinations of non-irrigated dominantland cover and hydrologic soil groups.


is highly likely that recent annual pumping rates exceed the 1993rate. However, it is unclear whether increased pumping implies agreater fraction of recharge being removed from the island’s aqui-fers since actual recharge rates may be higher than estimated inthis study.

4.4. Rainfall

The spline maps of daily rainfall during 1992–2009 (i.e., thegridded files of rainfall input to the SWB model) produced a meanannual island rainfall of 2104 mm (Fig. 4). Mean annual snowfallamounted to only 20 mm (or 1%) of mean annual island rainfall,which suggest that freshwater derived from snowmelt is a smallcomponent in the overall water budget of the island. In a separatestudy that only investigated rainfall spatial variability on Jeju (i.e.,no water balance analysis), Mair et al. (2013) estimated a slightlylower mean annual rainfall of 2082 mm using ordinary kriging(OK) (Goovaerts, 1997) and annual average gauge rainfall fromthe same set of gauges over the same time period (Fig. 1). Thehigher mean rainfall in this study (+22 mm or 1%) may be due,in part, to the slightly smaller island area used in this study thatexcludes the drier perimeter areas of Jeju (11 km2). The estimateof mean island rainfall from this study and Mair et al. (2013) are6–12% greater than estimates used in previous water balancestudies (Fig. 4, Table S2). These higher estimates of mean annualrainfall are mainly due to the addition of observations from tworain gauges located in the high elevation and high rainfall areasnear Mt. Halla. Despite similar estimates of mean island rainfallby the spline and OK algorithms, larger interpolation differencesare evident in the spatial distribution at selected locations. Thedifferences between the two mean annual rainfall maps are dueto the interpolation approach (spline vs. OK) and the time seriesof rainfall at each gauge (daily data for spline vs. annual data forOK). Because of much higher daily intra-island rainfall variability,the mean annual map produced by Mair et al. (2013) using an-nual data and the OK approach are likely a better representationof the spatial distribution of mean annual rainfall. For example,the spline algorithm produces less mean annual rainfall in theupper Aewol, Hallim, and West Jeju watersheds, and greater aver-age rainfall in the upper Jocheon and East Seogwipo watershedswhen compared with OK. Hence, the underestimation of rainfallin the upper West Jeju watershed may be a cause of the model’sunderestimation of observed direct runoff at the Oedocheon(gauge 2) site.

4.5. Direct runoff

The streamflow measurements used to calibrate the model inthis study imply that much less surface runoff, as a fraction of rain-fall, occurs on Jeju Island than previously estimated (Fig. 4). Mea-surements at the four stream gauge locations were not used inprevious studies on Jeju and, therefore, represent an improvementin quantifying the amount of rainfall converted to direct runoff.However, the lack of measurements in the eastern half of the islandprohibited an assessment of model performance in that area of theisland. Incorporation of additional measurements of direct runofffrom other locations on the island, such as the eastern watersheds,would help to further refine water balance estimates and confirmwhether previous studies have overestimated direct runoff.

4.6. Evapotranspiration

Interception loss, I, and soil moisture ET, ETsw, were estimated at304 and 731 mm, respectively, during the baseline period. Thus,mean annual ETtot totaled 1035 mm or 49% of gross precipitation,and implies substantially higher ET on Jeju than estimated by ear-lier studies: 34% by KOWACO (2003b) and Won et al. (2006), and37% by Hahn et al. (1997). The previous studies did not explicitlyaccount for interception losses, which can be significant from for-est canopies and vegetative surfaces. Indeed, forest interceptionlosses in maritime climates can be much greater than those arisingfrom transpiration due to the utilization of advected energy (Cal-der, 1998). Total ET losses under different types of forest and veg-etation on Jeju are not reported in the literature; however, ETpot

and reference ET (ETref), and components of the energy/water bud-get have been studied on Jeju (Rim, 2004; Cassardo et al., 2009;Rim, 2009). Rim (2004) estimated mean annual ETpot as 890 mmand 946 mm using the Penman method over a 10-year period from1991 to 2000 at two coastal locations, Jeju City and Gosan, respec-tively. Rim (2009) later reported annual ETref ranging from 910 to>1127 mm (upper boundary not specified) at coastal locations ofJeju City, Seogwipo City, and Seongsanpo in four separate years(1975, 1985, 1995, and 2004), and showed that the greatest ETref

in South Korea occurs along the southern coast of the Korean pen-insula and on Jeju. In this study, mean annual ETtot was estimatedat selected rain gauge sites by averaging the simulated ETtot amonggrid cells within a 500 m radius of the gauge location. Mean annualETtot totaled 645, 796, 1002, and 765 mm at four coastal gaugelocations including Jeju City, Gosan, Seongsanpo, and SeogwipoCity, respectively (Fig. 1). These limited data suggest that mean an-nual ETtot was occurring at about 72% and 84% of the mean annualETpot rates computed for Jeju City and Gosan, respectively. Cassardoet al. (2009) estimated components of the energy and water bal-ance for all of South Korea during the summer months of June, July,and August in 2005. They found that ET is the dominant term inboth energy and water budgets during the monsoon season andsometimes exceeds net radiation between storm events. They ob-served that mountainous areas of South Korea with high rainfall,including Jeju, have very high ET. They also noted that Jeju hashigher net radiation and sensible heat flux when compared to pen-insular South Korea. Thus, relative to most of peninsular South Kor-ea, Jeju is likely to experience higher ET due to its maritimeposition, forested mountainous interior, and concurrent seasonali-ty of high rainfall coupled with high net radiation.

ETtot has been characterized in other parts of Korea and Japan(Shimizu et al., 2003; Kosugi and Katsuyama, 2007; Ha and Ko,2009). Ha and Ko (2009) used the eddy covariance method to esti-mate a ETtot loss rate (i.e., loss rate expressed as a function of grossprecipitation) of 53% over a three-month period from May toAugust, 2008 in the Seocheon area, South Korea. Kosugi andKatsuyama (2007) estimated ETtot in a Japanese cypress forest in


southern Japan using a water balance approach and the eddycovariance method. They reported a mean ETtot loss rate of 46%using a water balance approach over a 33-year period during1972–2004 and a mean ETtot loss rate of 48% using the eddy covari-ance method during 2001-2003. Shimizu et al. (2003) reported aloss rate of 39–40% during 1996–1998 in a mixed stand of conifer-ous and broadleaved forest in southern Japan. These studies sug-gest that the mean ETtot loss rate of 49% computed by the SWBmodel is within the range of loss rates reported at other forest sitesin the region. However, the interception rates used in this study arebased solely on literature reviews and may not reflect interceptionlosses unique to Jeju. In addition, the SWB code does not reduceETpot by the amount of precipitation lost to interception. Thus,there is the possibility that the model’s estimate of ETtot may ex-ceed ETpot (i.e., ETtot = I + ETsw > ETpot), which implies that ETtot

may be overestimated by the SWB code. Finally, the Hargreaves–Samani method is a temperature-based method for estimatingETpot and may inadequately reflect changes in ETpot related to inci-dent radiation variability (e.g., clouds or fog). In the Hawaiian Is-lands, cloud occurrence and frequency coupled with a trade windinversion have a significant impact on ETpot in mountainous areas(Lau and Mink, 2006). As a result, temperature-based methodsare generally not considered a good indicator of ETpot in Hawaiiwithin the cloud zone. Energy-based methods such as the Penmanor Penman-Montieth methods are preferred when estimating ETpot

across areas within the cloud zone in Hawaii. However, the influ-ences of cloud cover on net radiation and its impact on ETpot acrossJeju is not understood. Hence, further investigation is needed to as-sess whether temperature-based methods for estimating ETpot aresuitable for Jeju and similar environments.

The substantial hydrological effects of vegetation on ET andgroundwater recharge combined with the lack of current under-standing of these processes on Jeju stress the need for characteriz-ing water losses (i.e., interception, transpiration) in forestvegetation, grassland areas, and agricultural cropland areas (i.e.,orchards). Field studies in high rainfall, critical recharge areas onJeju would help to minimize the uncertainty on the effects of veg-etation on groundwater recharge estimates. For example, Giam-belluca et al. (2009) recently used the eddy covariance method tomeasure forest ET in Hawaii. They reported much higher rates ofevaporation than previous estimates of forest ET, and highlightedthe importance of wet-canopy evaporation in controlling varia-tions in ET and contributing to high annual ET in Hawaii. As a resultof these higher estimates of ET, they concluded that groundwaterrecharge estimates in Hawaii may need to be revised. Groundwaterrecharge on the islands of Maui and Oahu are now being updatedto reflect the increased understanding of forest ET and long-termrainfall trends (USGS, 2013a; USGS, 2013b). A similar assessmentof canopy interception and transpiration processes on Jeju wouldhelp further minimize the uncertainty in groundwater rechargeestimates.

5. Conclusions

A daily water balance analysis was conducted to estimate thespatial and temporal distribution of groundwater recharge on theisland of Jeju for baseline, drought, and climate–land use changescenarios. The SWB computer code was used to compute the waterbalance components on a grid cell basis. Until now, the SWB codehas not been used in an oceanic island setting or in mountainousterrain overlain with highly permeable lava flows. Hence, the useof SWB to estimate groundwater recharge on Jeju represents anew and unique application. This study improves upon previousrecharge estimates by using an expanded rainfall dataset, address-ing canopy interception losses, routing direct runoff to downslope

cells, and including measurements of direct runoff into model cal-ibration. Of the three recharge scenarios, the greatest recharge oc-curs under the baseline scenario (884 mm) followed by theclimate–land use change scenario (788 mm) and drought scenario(591 mm). Mean annual recharge under the baseline scenario waswithin the range of previous estimates and only slightly lower thanthe mean of four previous recharge studies (902 mm). Mean an-nual recharge as a fraction of rainfall was only 42% under the base-line scenario, which was slightly less than previous estimates of44–48%. A complete set of pumping rates from all well sources inrecent years are lacking. However, the maximum historicalreported pumping rate (c. 1993) as a fraction of mean annualrecharge was 15% and within the range of 14–16% computed fromearlier studies.

Model estimates of direct runoff were substantially less thanpreviously reported, which implies that direct runoff may havebeen overestimated by past studies. However, measurements of di-rect runoff from other parts of the island are needed to confirmwhether direct runoff has been overestimated. Total ET losses weresubstantially higher than previously reported and suggest that ETmay have been underestimated by previous studies. Although thecomputer code accounts for canopy interception losses, it doesnot correct potential ET by the amount of moisture lost to canopyinterception. Hence, model estimates of total ET may exceed actualET losses. Further field investigative study is needed to betterunderstand canopy interception and plant transpiration under dif-ferent types of dominant land cover and hydro-climatic conditions,and to determine whether temperature-based methods for esti-mating potential ET are acceptable across the range of conditionsfound on Jeju. Because SWB relies on temperature-based methodsfor estimating ETpot, it may overestimate total ET losses in environ-ments within the cloud zone. The model is currently incapable ofapplying different growing and non-growing seasons to differentparts of the model domain (e.g., growing season length adjustedto altitude) and different time periods for the purpose of comput-ing interception losses. Hence, only one growing and non-growingseason could be specified for the entire island and modeling timeperiod, which may not reflect intra-island and inter-annual grow-ing season variability. Finally, the model does not include a mech-anism for including additional sources of groundwater rechargeincluding fog drip, irrigation, and artificial recharge. Consequently,the results presented in this study quantify natural recharge onlyand represent a conservative estimate of total potential recharge.

The SWB code represents a relatively simple tool for producingspatio-temporal estimates of groundwater recharge. The applica-tion of SWB to the island of Jeju represents a rigorous test of themodel’s capability to estimate recharge in a temperate-humid areaof diverse land use, high rainfall temporal and spatial variability,high topographic relief, and generally high infiltration capacity.The recharge estimates presented in this study are within therange of previous values, which suggests that the model can pro-duce reliable estimates of spatially-varying recharge in temper-ate-humid climates. However, further field investigative study onJeju is needed to validate the model estimates (potential and actualET, surface runoff, and growing season variability) and completelyassess the model’s applicability to Jeju and other island settings.

Acknowledgements

This research was supported by the Ministry of KnowledgeEconomy, Korea Institute for Advancement of Technology, and JejuLeading Industry Office through the Leading Industry Developmentfor Economic Region and Basic Research Project (12-3211) of KI-GAM. The authors would like to thank Michael Fienen, RichardHealy, Keith Lucey, and one anonymous reviewer for their con-structive comments.


Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jhydrol.2013.08.015.

References

Bailey, R.T., Jenson, J.W., Olsen, A.E., 2009. Numerical modeling of atoll islandhydrogeology. Ground Water 47 (2), 184–196.

Batelaan, O., De Smedt, F., 2007. GIS-based recharge estimation by couplingsurface–subsurface water balances. J. Hydrol. 337 (3–4), 337–355.

Boo, K.O., Kwon, W.T., Baek, H.J., 2006. Change of extreme events of temperatureand precipitation over Korea using regional projection of future climate change.Geophys. Res. Lett. 33, L01701. http://dx.doi.org/10.1029/2005GL023378.

Calder, I.R., 1998. Water use by forests, limits and controls. Tree Physiol. 18, 625–631.

Calheiros de Miranda, R., Butler, D., 1986. Interception of rainfall in a hedgerowapple orchard. J. Hydrol. 87 (3–4), 245–253.

Cassardo, C., Park, S.K., Malla Thakuri, B., Priolo, D., Zhang, Y., 2009. Soil surfaceenergy and water budgets during a monsoon season in Korea. J. Hydrometeorol.10 (6), 1379–1396.

Choi, H.-M., Lee, J.-Y., 2012. Changes of groundwater conditions on Jeju volcanicisland, Korea: implications for sustainable agriculture. Afr. J. Agr. Res. 10 (4),647–661.

Choi, G., Kwon, W.-T., Boo, K.-O., Cha, Y.-M., 2008. Recent spatial and temporalchanges in means and extreme events of temperature and precipitation acrossthe Republic of Korea. J. Korean Geogr. Soc. 43 (5), 681–700.

Chu, P.-S., Chen, H., 2005. Interannual and interdecadal rainfall variations in theHawaiian Islands. J. Clim. 18 (22), 4796–4813.

Chung, C.-H., 2007. Vegetation response to climate change on Jeju Island, SouthKorea, during the last deglaciation based on pollen record. Geosci. J. 11 (2), 147–155.

Corbett, E.S., Crouse, R.P., 1968. Rainfall Interception by Annual Grass andChaparral. . .Losses Compared. U.S. Forest Service Research Paper PSW-48, U.S.Forest Service.

Cruz, J.V., 2003. Groundwater and volcanoes: examples from the Azoresarchipelago. Environ. Geol. 44 (3), 343–355.

Deguchi, A., Hattori, S., Park, H.-T., 2006. The influence of seasonal changes incanopy structure on interception loss: application of the revised Gash model. J.Hydrol. 318 (1–4), 80–102.

Diaz-Nieto, J., Wilby, R., 2005. A comparison of statistical downscaling and climatechange factor methods: impacts on low flows in the River Thames, UnitedKingdom. Clim. Change 69 (2–3), 245–268.

Dripps, W.R., Bradbury, K.R., 2007. A simple daily soil–water balance model forestimating the spatial and temporal distribution of groundwater recharge intemperate humid areas. Hydrogeol. J. 15, 433–444.

Dripps, W.R., Bradbury, K.R., 2010. The spatial and temporal variability ofgroundwater recharge in a forested basin in northern Wisconsin. Hydrol.Process. 24 (4), 383–392.

Engott, J.A., 2011. A Water-budget Model and Assessment of Groundwater Rechargefor the Island of Hawai‘i: U.S. Geological Survey Scientific Investigations, Report2011-5078.

Engott, J.A., Vana, T.T., 2007. Effects of Agricultural Land-use Changes and Rainfallon Ground-water Recharge in Central and West Maui, Hawai‘i, 1926–2004: U.S.Geological Survey Scientific Investigations, Report 2007-5103.

ESRI, 2010. ArcInfo 10, Redlands, CA.Feinstein, D.T., Hunt, R.J., Reeves, H.W., 2010. Regional Groundwater-flow Model of

the Lake Michigan Basin in Support of Great Lakes Basin Water Availability andUse Studies: U.S. Geological Survey Scientific Investigations, Report 2010-5109.

Finch, J.W., 2001. Estimating change in direct groundwater recharge using aspatially distributed soil water balance model. Quart. J. Eng. Geol. Hydrogeol. 34(1), 71–83.

Flint, A.L., Flint, L.E., 2007. Application of the Basin Characterization Model toEstimate In-place Recharge and Runoff Potential in the Basin and RangeCarbonate-rock Aquifer System, White Pine County, Nevada, and Adjacent Areasin Nevada and Utah: U.S. Geological Survey Scientific Investigations, Report2007-5099.

Giambelluca, T.W., 1983. Water Balance of the Pearl Harbor-Honolulu Basin,Hawaii, 1946–1975. Water Resources Research Center Technical Report 151,University of Hawaii at Manoa, Honolulu.

Giambelluca, T.W., Diaz, H.F., Luke, M.S.A., 2008. Secular temperature changes inHawaii. Geophys. Res. Lett. 35 (12), L12702.

Giambelluca, T.W., Martin, R.E., Asner, G.P., Huang, M., Mudd, R.G., Nullet, M.A.,DeLay, J.K., Foote, D.E., 2009. Evapotranspiration and energy balance of nativewet montane cloud forest in Hawaii. Agric. For. Meteorol. 149, 230–243.

Gingerich, S.B., 2008. Ground-water Availability in the Wailuku Area, Maui, Hawaii.U.S. Geological Survey Scientific Investigations Report 2008-5236, 95p.

Gingerich, S.B., Jenson, J.W., 2010. Groundwater Availability Study for Guam; Goals,Approach, Products, and Schedule of Activities: U.S. Geological Survey FactSheet 2010-3084.

Gingerich, S.B., Oki, D.S., 2000. Ground-water in Hawai‘i: U.778 S. Geological SurveyFact Sheet 126-00.

Gingerich, S.B., Voss, C.I., 2005. Three-dimensional variable-density flow simulationof a coastal aquifer in southern Oahu, USA. Hydrogeol. J. 13, 436–450.

Goovaerts, P., 1997. Geostatistics for Natural Resources Evaluation. OxfordUniversity Press, New York.

Ha, K., Ko, K., 2009. Evapotranspiration Monitoring with Eddy Covariance System inSeocheon Area, Korea. Abstract #EP31A-0578, American Geophysical Union, FallMeeting 2009, San Francisco.

Hagedorn, B., El-Kadi, A.I., Mair, A., Whittier, R.B., Ha, K., 2011. Estimating rechargein fractured aquifers of a temperate humid to semiarid volcanic island (Jeju,Korea) from water table fluctuations, and Cl, CFC-12 and 3H chemistry. J.Hydrol. 409 (3–4), 650–662.

Hahn, J., Lee, Y., Kim, N., Hahn, C., Lee, S., 1997. The groundwater resources andsustainable yield of Cheju volcanic island, Korea. Environ. Geol. 33 (1), 43–53.

Hargreaves, G.H., Samani, Z.A., 1985. Reference crop evapotranspiration fromtemperature. Appl. Eng. Agric. 1, 96–99.

Hayhoe, K., Wake, C., Huntington, T., Luo, L., Schwartz, M., Sheffield, J., Wood, E.,Anderson, B., Bradbury, J., DeGaetano, A., Troy, T., Wolfe, D., 2007. Past andfuture changes in climate and hydrological indicators in the US Northeast. ClimDyn 28 (4), 381–407.

Hay, L.E., Wilby, R.L., Leavesley, G.H., 2000. A comparison of delta change anddownscaled GCM scenarios for three mountainous basins in the United States.JAWRA J. Am. Water Resour. Assoc. 36 (2), 387–397.

Im, E.S., Jung, I.W., Bae, D.H., 2011. The temporal and spatial structures of recent andfuture trends in extreme indices over Korea from a regional climate projection.Int. J. Climatol. 31 (1), 72–86.

Izuka, S.K., Oki, D.S., Chen, C., 2005. Effects of Irrigation and Rainfall Reduction onGround-water Recharge in the Lihue Basin, Kauai, Hawaii: U.S. GeologicalSurvey Science Investigations Report 2005-5146.

Izuka, S.K., Perreault, J.A., Presley, T.K., 2007. Areas Contributing Recharge to Wellsin the Tafuna-Leone Plain, Tutuila, American Samoa: U.S. Geological SurveyScientific Investigations Report 2007-5167.

Jeju Special Self-Governing Province, 2013. Facts and statistics. <http://english.jeju.go.kr/> (accessed 12.03.13).

Jeong, S.-J., Ho, C.-H., Choi, S.-D., Kim, J., Lee, E.-J., Gim, H.-J., 2013. Satellite data-based phenological evaluation of the nationwide reforestation of South Korea.PLoS ONE 8 (3), e58900.

Johnson, A.G., 2012. A Water-budget Model and Estimates of GroundwaterRecharge for Guam: U.S. Geological Survey Scientific Investigations, Report2012-5028.

Jung, I.-W., Bae, D.-H., Kim, G., 2011. Recent trends of mean and extremeprecipitation in Korea. Int. J. Climatol. 31 (3).

Kang, H.-J., Lee, S.-C., Yang, S.-H., Lee, S.-S., Koh, S.-J., Ha, S.-K., Hur, S.-O., 2010. SoilErosion Characteristics Under Rainfall Simulator Conditions of the Jeju Soil inKorea, 19th World Congress of Soil Science. Soil Solutions for a Changing World,Brisbane, Australia.

KIGAM, 2010a. Daily Minimum, Maximum, and Average Temperatures at Jeju Cityand Seogwipo City from January 1, 1961 to December 31, 2009. Korea Instituteof Geoscience and Mineral Resources.

KIGAM, 2010b. Daily Minimum, Maximum, and Average Temperatures Collectedfrom 22 Monitoring Stations on Jeju Island from January 1, 2006 to December31, 2009. Korea Institute of Geoscience and Mineral Resources.

KIGAM, 2010c. Daily Rainfall Dataset for Island of Jeju, Korea from January 1, 1924to February 28, 2010. Korea Institute of Geoscience and Mineral Resources.

KIGAM, 2010d. Digital Elevation Model (DEM) of the island of Jeju, Korea. KoreaInstitute of Geoscience and Mineral Resources.

KIGAM, 2010e. Direct Runoff Measurements at Gangjeongcheon, Jungmuncheon,Oedocheon, and Ongpocheon Stream Gauge Sites from 2006 to 2009. KoreaInstitute of Geoscience and Mineral Resources.

KIGAM, 2010f. Land Use/Land Cover Map of Jeju Island. Korea Institute ofGeoscience and Mineral Resources.

KIGAM, 2010g. Soil Series Map of Jeju Island. Korea Institute of Geoscience andMineral Resources.

Kim, K.H., Woo, B.M., 1988. Study on rainfall interception loss from canopy in forest.J. Korean For. Soc. 77 (3), 331–337 (in Korean).

Kim, Y., Lee, K.S., Koh, D.C., Lee, D.H., Lee, S.G., Park, W.B., Koh, G.W., Woo, N.C., 2003.Hydrogeochemical and isotopic evidence of groundwater salinization in acoastal aquifer: a case study in Jeju volcanic island, Korea. J. Hydrol. 270, 282–294.

Kim, K.-Y., Seong, H., Kim, T., Park, K.-H., Woo, N.-C., Park, Y.-S., Koh, G.-W., Park, W.-B., 2006. Tidal effects on variations of fresh–saltwater interface andgroundwater flow in a multilayered coastal aquifer on a volcanic island (JejuIsland, Korea). J. Hydrol. 330, 525–542.

Koh, D.-C., Chang, H.-W., Lee, K.-S., Ko, K.-S., Kim, Y., Park, W.-B., 2005.Hydrogeochemistry and environmental isotopes of ground water in Jejuvolcanic island, Korea: implications for nitrate contamination. Hydrol.Process. 19, 2225–2245.

Koh, D.-C., Plummer, L.N., Solomon, D.K., Busenberg, E., Kim, Y.-J., Chang, H.-W.,2006a. Application of environmental tracers to mixing, evolution, and nitratecontamination of ground water in Jeju Island, Korea. J. Hydrol. 327, 258–275.

Koh, G.-W., Kang, B.-R., Moon, D.-C., 2006b. Hydrogeological features andgroundwater management systems of Jeju Island. Proc. Jeju Hawaii WaterForum 2006, 325–362.

Koh, D.-C., Chae, G.-T., Yoon, Y.-Y., Kang, B.-R., Koh, G.-W., Park, K.-H., 2009. Baselinegeochemical characteristics of groundwater in the mountainous area of JejuIsland, South Korea: implications for degree of mineralization and nitratecontamination. J. Hydrol. 376, 81–93.



http://refhub.elsevier.com/S0022-1694(13)00593-3/h0005




http://dx.doi.org/10.1029/2005GL023378































































http://english.jeju.go.kr/

http://english.jeju.go.kr/



































Koh, D.-C.., Ha, K., Lee, K.-S., Yoon, Y.-Y., Ko, K.-S., 2012a. Flow paths and mixingproperties of groundwater using hydrogeochemistry and environmental tracersin the southwestern area of Jeju volcanic island. J. Hydrol. 432-433, 61–74.

Koh, E.-H., Kaown, D., Mayer, B., Kang, B.-R., Moon, H.S., Lee, K.-K., 2012b.Hydrogeochemistry and Isotopic Tracing of Nitrate Contamination of TwoAquifer Systems on Jeju Island, Korea. J. Environ. Qual. 41 (6), 1835–1845.

Kosugi, Y., Katsuyama, M., 2007. Evapotranspiration over a Japanese cypress forest.II. Comparison of the eddy covariance and water budget methods. J. Hydrol. 334(3–4), 305–311.

KOWACO, 2003a. Comprehensive Investigation of Hydrogeology and GroundwaterResources in Jeju Island, Jeju Province, Korea Water Resources Corporation (inKorean).

KOWACO, 2003b. Meteorology, Hydrology, and Water Budget Analysis, TechnicalReport, Jeju Province, Korea Water Resources Corporation.

Lau, L.S., Mink, J.F., 2006. Hydrology of the Hawaiian Islands. University of HawaiiPress, Honolulu.

Lee, D.K., Kim, Y.S., Kim, G.T., Joo, K.Y., 1997. Throughfall, stemflow and rainfallinterception loss in Pinus koraiensis Sieb. et Zucc., Larix leptolepis (Sieb. et Zucc.)Gordon and Quercus species stand at Kwangju-gun, Kyunggi-Do. J. Korean For.Soc. 86 (2), : 200–207 (in Korean).

Mair, A., El-Kadi, A., Ha, K., Koh, G.-W., 2013. Temporal and spatial variability ofrainfall and climate trend on Jeju Island. Geosci. J. 17 (1), 75–85.

Min, H.J., Woo, B.M., 1995. Throughfall, stemflow and interception loss at Pinustaeda and Pinus densiflora stands. J. Korean For. Soc. 84 (4), 502–516 (in Korean).

Min, S.-K., Park, E.-H., Kwon, W.-T., 2004. Future projections of east Asian climatechange from multi-AOGCM ensembles of IPCC SRES scenario simulations. J.Meteorol. Soc. Jpn 82 (4), 1187–1211.

Moon, S.-H., Lee, J.-Y., Lee, B.-J., Park, K.-H., Jo, Y.-J., 2012. Quality of harvestedrainwater in artificial recharge site on Jeju volcanic island, Korea. J. Hydrol. 414-415, 268–277.

Oki, D.S., 2004. Trends in Streamflow Characteristics at Long-term Gagingstations, Hawaii: U.S. Geological Survey Scientific Investigations Report 2004-5080.

Oki, D.S., 2005. Numerical Simulation of the Effects of Low-Permeability Valley-fillBarriers and the Redistribution of Ground-water Withdrawals in the PearlHarbor Area, Oahu, Hawaii: U.S. Geological Survey Scientific InvestigationsReport 2005-5253.

Oki, D.S., 2008. The Significance of Accounting Order for Evapotranspiration andRecharge in Monthly and Daily Threshold-type Water Budgets: U.S. GeologicalSurvey Scientific Investigations, Report 2008-5163.

Park, J.-S., Kang, H.-S., Lee, Y.S., Kim, M.-K., 2011. Changes in the extreme dailyrainfall in South Korea. Int. J. Climatol. 31 (15), 2290–2299.

Presley, T.K., 2005. Effects of the 1998 Drought on the Freshwater Lens in the LauraArea, Majuro Atoll, Republic of the Marshall Islands: U.S. Geological SurveyScientific Investigations Report 2005-5098.

Rim, C.-S., 2004. A sensitivity and error analysis for the penman evapotranspirationmodel. KSCE J. Civil Eng. 8 (2), 249–254.

Rim, C.-S., 2009. The effects of urbanization, geographical and topographicalconditions on reference evapotranspiration. Clim. Change 97 (3), 483–514.

Rushton, K.R., Ward, C., 1979. The estimation of groundwater recharge. J. Hydrol. 41(3–4), 345–361.

Scanlon, B., Healy, R., Cook, P., 2002. Choosing appropriate techniques forquantifying groundwater recharge. Hydrogeol. J. 10 (1), 18–39.

Schroeder, P.R., Aziz, N.M., Lloyd, C.M., Zappi, P.A., 1994. The Hydrologic Evaluationof Landfill Performance (HELP) Model, User’s Guide for Version 3, EPA/600/R-94/168a, U.S. Environmental Protection Agency Office of Research andDevelopment, Washington, DC.

Shade, P.J., 1995. Water Budget for the Island of Kauai, Hawaii: U.S. GeologicalSurvey Water-Resources Investigations Report 95-4128.

Shade, P.J., Nichols, W.D., 1996. Water Budget and the Effects of Land-use Changeson Ground-water Recharge, Oahu, Hawaii: U.S. Geological Survey ProfessionalPaper 1412-C.

Shimizu, A., Shimizu, T., Miyabuchi, Y., Ogawa, Y., 2003. Evapotranspiration andrunoff in a forest watershed, western Japan. Hydrol. Process. 17 (15), 3125–3139.

Silva, I.C., Okumura, T., 1996. Throughfall, stemflow and interception loss in a mixedwhite oak forest (Quercus serrata Thunb.). J. For. Res. 1 (3), 123–129.

Spalding, R.F.U.Z.K., Hyun, S.W., Martin, G.E., Burbach, M.E., Yang, S.I.I., Kim, M.,Exner, M.E., Song, S.J., 2001. Source identification of nitrate on Cheju Island,South Korea. Nutr. Cycl. Agroecosyst. 61 (3), 237–246.

Thornthwaite, C.W., Mather, J.R., 1955. The water balance. Climatology 8 (1), 1–104.Thornthwaite, C.W., Mather, J.R., 1957. Instructions and tables for computing

potential evapotranspiration and the water balance. Publ. Climatol. 10, 185–311.

Thurow, T.L., Blackburn, W.H., Warren, S.D.C.A., Taylor, J., 1987. Rainfall interceptionby midgrass, shortgrass, and live oak mottes. J. Range Manage. 40 (5), 455–460.

Toba, T., Ohta, T., 2005. An observational study of the factors that influenceinterception loss in boreal and temperate forests. J. Hydrol. 313 (3–4), 208–220.

Todd Engineers, ETIC Engineering, 2005. Development of a GroundwaterManagement Model, Honolulu Area of the Southern O‘ahu GroundwaterSystem, Final Report, City and County of Honolulu.

Tribble, G., 2008. Ground Water on Tropical Pacific Islands – Understanding a VitalResource: U.S. Geological Survey Circular 1312.

USDA-NRCS, 2003. National Engineering Handbook, Part 630, Hydrology. U.S.Department of Agriculture, Natural Resources Conservation Service. <http://directives.sc.egov.usda.gov/.

USGS, 2008. Documentation of Computer Program INFIL3.0—A Distributed-parameter Watershed Model to Estimate Net Infiltration Below the RootZone: U.S. Geological Survey Scientific Investigations Report 2008-5006.

USGS, 2013a. Pacific Islands Water-resource Study Titled ‘‘Groundwater Rechargeand Availability in the Pearl Harbor Aquifer, Oahu, Hawaii.’’ <http://hi.water.usgs.gov/studies/pearlharboraquifer/> (accessed 01.17.13). U.S.Geological Survey Pacific Islands Water Science Center.

USGS, 2013b. Pacific Islands Water-resource Study Titled ‘‘Reassessment ofGroundwater Recharge for Central and West Maui.’’ <http://hi.water.usgs.gov/studies/CWmaui_recharge/> (accessed 01.17.13). U.S. Geological Survey PacificIslands Water Science Center.

Voss, C.I., Wood, W.W., 1993. Synthesis of Geochemical, Isotopic and GroundwaterModelling Analysis to Explain Regional Flow in a Coastal Aquifer of SouthernOahu, Hawaii, Mathematical Models and their Applications to Isotope Studies inGroundwater Hydrology. International Atomic Energy Agency, Vienna, Austria,pp. 147–178.

WCRP, 2010. Climate Change Indices. <http://cccma.seos.uvic.ca/ETCCDMI/list_27_indices.shtml>. World Climate Research Programme (accessedNovember 2010).

Westenbroek, S.M., Kelson, V.A., Dripps, W.R., Hunt, R.J., Bradbury, K.R., 2010. SWB –A modified Thornthwaithe-Mather Soil–Water-Balance Code for EstimatingGroundwater Recharge: U.S. Geological Survey Techniques and, Methods G-A31.

Won, J.-H., 2004. Hydrogeological Investigation and Hydrologic Budget ofGroundwater Resources on Jeju Island, Korea. PhD Thesis, Seoul NationalUniversity, Korea (in Korean).

Won, J.-H., Kim, J.-W., Koh, G.-W., Lee, J.-Y., 2005. Evaluation of hydrogeologicalcharacteristics in Jeju Island, Korea. Geosci. J. 9 (1), 33–46.

Won, J.-H., Lee, J.-Y., Kim, J.-W., Koh, G.-W., 2006. Groundwater occurrence on JejuIsland, Korea. Hydrogeol. J. 14, 532–547.

Xiao, Q., McPherson, E.G., Simpson, J.R., Ustin, S.L., 1998. Rainfall interception bySacramento’s urban forest. J. Arboric. 24 (4), 235–244.




















































http://directives.sc.egov.usda.gov

http://directives.sc.egov.usda.gov

http://hi.water.usgs.gov/studies/pearlharboraquifer/

http://hi.water.usgs.gov/studies/pearlharboraquifer/

http://hi.water.usgs.gov/studies/CWmaui_recharge/

http://hi.water.usgs.gov/studies/CWmaui_recharge/






http://cccma.seos.uvic.ca/ETCCDMI/list_27_indices.shtml

http://cccma.seos.uvic.ca/ETCCDMI/list_27_indices.shtml







journal of hydrology et al... · however, these water balance studies were computed using either...

Documents