Remote Sensing of Environment 119 (2012) 21–34

Contents lists available at SciVerse ScienceDirect

Remote Sensing of Environment

j ourna l homepage: www.e lsev ie r .com/ locate / rse

Regional scale assessment of Submarine Groundwater Discharge in Irelandcombining medium resolution satellite imagery and geochemical tracing techniques

Jean Wilson ⁎, Carlos RochaBiogeochemistry Research Group, School of Natural Sciences, Geography Department, Trinity College Dublin, Dublin 2, Ireland

⁎ Corresponding author. Tel.: +353 1 8961121; fax:E-mail address: jewilson@tcd.ie (J. Wilson).

0034-4257/$ – see front matter © 2011 Elsevier Inc. Alldoi:10.1016/j.rse.2011.11.018

a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 August 2011Received in revised form 22 November 2011Accepted 24 November 2011Available online 14 January 2012

Keywords:Thermal remote sensingLandsat ETM+Submarine groundwater dischargeGeochemical tracingRadonSalinity

This paper sets the foundation for the use of freely available Landsat Enhanced Thematic Mapper (ETM+)thermal infrared (TIR) imagery in a regional scale assessment of submarine groundwater discharge (SGD)to coastal waters. A comprehensive, tiered, three-step approach is proposed as the most effective and afford-able means to determine the spatial extent and scale of SGD from coastal aquifers to the coastal margin. Asthe preliminary step, Sea Surface Temperature (SST) values derived from Landsat ETM+ TIR are used to suc-cessfully detect plumes of colder water eventually associated with SGD in close proximity to the shoreline.Subsequently, potential sites of SGD are linked to geological features on land acting as possible sources, bycombining within a Geographical Information System (GIS), mapped temperature anomalies with ancillaryon-shore spatial datasets describing bedrock geology including aquifer fault lines. Finally, nearshore surveysmapping the activity of 222Rn (radon) and salinity are carried out to verify the presence of SGD and provide aqualitative assessment of fresh groundwater inputs to the coastal zone.Practical application of the complete approach in the context of coastal zone management is illustratedthrough a case-study of the Republic of Ireland. As part of this study, over 30 previously unidentified linksbetween aquifers on land and the sea are shown along the Irish coast, hence illustrating the tight couplingbetween coastal waters and groundwater inputs at an unprecedented spatial scale. The study demonstratesthe potential of the combined applications of remote sensing methods and geochemical tracing techniquesfor a cost-effective regional-scale assessment of groundwater discharge to coastal waters.

© 2011 Elsevier Inc. All rights reserved.

1. Introduction and background

1.1. Overview

SGD is defined broadly as any and all flow of water across the sea-bed from land to sea (Burnett et al., 2003) and encompasses severalcomponents of subsurface flow including terrestrial freshwater andrecirculated seawater (Moore, 1999). Groundwater will seep persis-tently into the sea through permeable sediments wherever an aquiferwith positive head relative to sea-level is hydraulically connected to asurface water body (Johannes, 1980) Accordingly, SGD is a ubiquitousfeature of coastlines worldwide. SGD is extremely variable both spa-tially and temporally but occurs predominantly in the form of near-shore seepage, offshore seepage and submarine springs (Burnett etal., 2001). Submarine seepage occurring offshore is generally relatedto extensive networks of underground caves and channels includinglocal fracture systems (Shaban et al., 2005) which facilitate the trans-port of groundwater from land aquifers to points several kilometresaway from the shoreline.

+353 1 6713397.

rights reserved.

Groundwater in transit from land to sea can become contaminatedwith a variety of substances including nutrients and heavy metals(Lee et al., 2009; Swarzenski & Baskaran, 2007) hence SGD has beendefined in the literature as a potentially significant source and path-way of nutrients, dissolved substances and diffuse pollution to coastalareas particularly when originating from contaminated continentalaquifers (Leote et al., 2008). While fresh groundwater discharge isconsidered to be less than 10% of the total freshwater flux to theocean, the inputs of associated nutrients and contaminants may befar more significant because concentrations in groundwater often ex-ceed that of surface waters (Slomp & Van Capellen, 2004). Therefore,relatively small groundwater discharge rates can deliver compara-tively large quantities of solutes including nutrients to coastal areas.For example, nutrient supply via SGD has been linked to eutrophica-tion and suggested as a potential precursor of harmful algal blooms(Hu et al., 2006; Lee & Kim, 2007) or increased bacterial concentra-tions (Boehm et al., 2004).

Despite acknowledgement of its potential impact on coastal ecosys-tem functioning SGD remains a poorly-understood and often over-looked process when implementing coastal monitoring andmanagement programmes. For instance, EU directives such as theWater Framework Directive (2000/60/EC) aimed at improving thequality of the water environment do not acknowledge SGD as a

22 J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

potential nutrient source for assessment or monitoring. This is becausethe spatially and temporally heterogeneous nature of groundwaterdischarge from an essentially invisible source renders locating andquantifying rates of SGD an appreciable challenge. Consequently, thequantitative distinction between SGD and easily gauged surfacerunoff sources may be impaired when implementing coastal manage-ment policy based on current nutrient monitoring programmes forexample.

In recognition of both the significance of groundwater dischargeas a potential source of contamination and the challenges to locatingand quantifying the contribution of groundwater discharge to thecoastal zone, a comprehensive cost-effective methodology to facili-tate a regional assessment of SGD is presented here, using Ireland asa case-study. This work is based on the premise that relatively coolgroundwater discharging to warmer coastal waters manifests in thethermal band of Landsat ETM+ TIR imagery acquired during summermonths. The overarching goal of this study, the first of its kind inIreland, is to identify and characterise locations of SGD through theintegration of satellite thermal remote sensing (Landsat ETM+ TIR),ancillary geological and hydrogeological data and geochemical trac-ing (Radon-222, salinity) techniques.

1.2. Thermal remote sensing and geochemical tracing for SGD detection

Remote sensing methods for SGD detection are applicable wher-ever temperature gradients form between coastal marine watersand discharging terrestrial groundwater. Temperature has beenused successfully to study groundwater discharge by comparing therelatively constant temperature of groundwater with that of surfacewaters which fluctuate with season (Dale & Miller, 2007). In general,groundwater between approximately 5 m and 100 m depths main-tains a nearly constant temperature of 1–2 °C higher than mean an-nual air temperature (Anderson, 2004).

As water is almost opaque in the TIR, thermal remote sensing ofsurface water temperatures only provides spatially distributed valuesof radiant temperature in the “skin” layer or top 100 μm of the watercolumn (Handcock et al., 2006). Daily observations of both regionaland global SST (Kilpatrick et al., 2001; Parkinson, 2003) provide forexample direct insight into the spatial and temporal variability ofupper ocean currents, water mass boundaries and mixing (Thomaset al., 2002). This is an established practice in oceanography: oceantemperatures have been studied extensively since the late 1970sfrom a variety of satellite sensors including NOAAs Advanced VeryHigh Resolution Radiometer (AVHRR) (Fox et al., 2005); NASA'sGeostationary Operational Environmental Satellites (GOES) (Menzel& Purdom, 1994); NASAs Advanced Spaceborne Thermal Emissionand Reflection Radiometer (ASTER) (Abrams, 2000), ENVISATs Ad-vanced Along-Track Scanning Radiometer (AATSR) (Corlett et al.,2006) and NASAs Moderate Resolution Imaging Spectroradiometer(MODIS) (Esaias et al., 1998). While global measurements of sea sur-face temperature gathered by these satellites facilitate effective ob-servations of ocean-basin scale circulation patterns, the relativelycoarse spatial resolutions of these sensors (≥1 km) will not resolvefine-scale SST gradients important for the study of oceanographic fea-tures of nearshore coastal zones (Fisher & Mustard, 2004). Indeed, aspatial resolution of 1 km is not sufficient to discriminate thermalgradients in water bodies less than 3 km in width or water masses lo-cated less than 2 km from the shoreline (Wloczyk et al., 2006). How-ever, satellite sensors with spatial resolutions appropriate for trackingsmall scale thermal patterns such as Landsat (Clark, 1993; Gibbons &Wukelic, 1989; Tcherepanov et al., 2005), are limited by both theirspectral and temporal resolution. For example, atmospheric correc-tion of surface temperature values generated from image data, re-quires at least two thermal bands (Wloczyk et al., 2006) and asboth Landsat Thematic Mapper (TM) and ETM+ record thermalemissivity in one waveband with a repeat cycle of 16-days, inherent

atmospheric correction and resolution of temporal variability onscales shorter than seasonal are precluded (Thomas et al., 2002).Moreover, an error with the scan-line corrector (SLC) aboard theETM+ sensor means that post-May 2003 images include large datagaps and consequently some of the available archived images maybe unusable. While an obvious solution to the problem of low spatial,spectral and temporal resolution associated with medium resolutionsatellite systems is to use airborne systems, these tend to be extreme-ly costly (Dave, 1998) and economically unsuitable for a regionalscale assessment of potential SGD locations.

Despite the shortcomings to the use of Landsat TIR for SST retriev-al, there have been some encouraging results. Thomas et al. (2002)quantified the variability of SST along the coastline of Maine using atime series of Landsat Thematic Mapper (TM) thermal band data. Dif-ferences in seasonal patterns of SST (spatial resolution of 120 m) ob-served during summer months in four adjacent bays were attributedto residual circulation, tidal mixing and freshwater influence. Fisherand Mustard (2004) developed a sea surface climatology trackingSST as a function of year-day, from a combination of Landsat TMand ETM+ TIR data for a study area in Southern New England. Asexpected, their results revealed that isolated and shallow water bod-ies undergomore extreme temperature variation than deeper embay-ments. Moreover, the spatial pattern of the climatology revealedanomalous patterns associated with anthropogenic thermal inputfrom a large power plant, supporting the application of Landsat ther-mal data to smaller scale studies. Wloczyk et al. (2006) calculated seaand lake surface temperatures in Northern Germany from Landsat TIRdata and compared them to in-situ measurements and an empiricalmodel developed by Germany's National Meteorological Service.Root-mean-square (RMS) deviations of 1.4 K and 2.2 K respectively,were reported.

The successful deployment of thermal remote sensing to delineategroundwater discharge to the coastal ocean documented in the liter-ature to date, centres predominantly on the acquisition and analysisof high resolution TIR imagery acquired from aerial (airborne) sur-veys. For example, Banks et al. (1996) used airborne Thermal InfraredMultispectral Scanner (TIMS) images to identify the location and thespatial extent of groundwater discharge to creeks on a peninsula inChesapeake Bay, USA. The presence of tonal differences in the TIMSimagery verified by water temperature measurements in the fieldwere interpreted to locate sites of SGD. Portnoy et al. (1998) usedaerial TIR imagery and shoreline salinity surveys to characterisegroundwater discharge to an estuary in Cape Cod, USA. Other exam-ples from the US include McKenna et al. (2001), Roseen et al.(2001), Ullman and Miller (2004) and Johnson et al. (2008) whoused airborne TIR imagery to identify groundwater discharge in Dela-ware's Inland Bays, the Great Bay Estuary (New Hampshire), Dela-ware Bay (Delaware), Waquiot Bay (Massachusetts) and Hawaiirespectively. The presence of SGD was verified in the examplesgiven through salinity and temperature measurements in the field.Elsewhere, Shaban et al. (2005) and Akawwi et al. (2008) conductedaerial TIR surveys along the Lebanese and Jordanian coastlines respec-tively to identify potential sites of SGD.

The application of Landsat TIR for SST retrieval is neither extensivenor widespread and although far less frequently reported, satellitederived SSTs have been applied with limited success to identify andcharacterise coastal SGD. Wang et al. (2008) used two Landsat ther-mal images acquired in 2000 and 2002 to successfully locate ground-water discharge areas in Rehoboth Bay, Indian River and Indian Bay,Delaware. The accuracy of the satellite derived temperatures was ver-ified through comparison with in-situ data and reported to be within1 °C of in-situ water temperatures. Varma et al. (2010) used a combi-nation of satellite (Landsat, ASTER and NOAA) thermal images to ex-amine the utility of remote sensing methods for detection of knownbut previously unmapped locations of SGD in Geographe Bay,Bunbury, Western Australia. Within their study, evidence of SGD

23J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

seepage was not detectable from either the Landsat or NOAA TIR im-ages. The failure of their approach was attributed to the poor spatialresolution of the NOAA AVHRR images (1 km) and the use of anunprocessed Landsat TIR image. The authors also discount the use ofLandsat data acquired post May 2003 (SLC off-mode) due to gaps inthe data.

Put simply, remote sensing can be used to detect SGD by examin-ing the thermal signature of discharging groundwater. Specifically,freely available Landsat ETM+TIR data, subject to the image proces-sing methodology outlined in this paper, may be successfully appliedto indicate potential locations of SGD. To actually confirm the pres-ence and occurrence of SGD at the sites identified, the thermal anom-alies observed within the satellite imagery as a signal for SGD, mustbe verified. Natural tracers of SGD, other than temperature include sa-linity and radioisotopes (e.g. Radon-222 and the radium quartetRadium-223, -224, -226, -228) which must be greatly enriched inthe discharging groundwater relative to the coastal ocean to providea detectable signal. The use of salinity as a groundwater tracer waspioneered by Johannes (1980) and has been used with success totrace SGD (Beck et al., 2007; Slater et al., 2010). However, groundwa-ter discharge to the coastal ocean may not be accompanied by a dis-cernible freshening of receiving coastal waters. In fact SGD mayinclude a major portion of recirculated seawater and the effects of di-lution for example may overshadow its nature as a tracer. Additional-ly, observed salinity differences could be the result other freshwaterprocesses such as surface water discharge.

More recent research confirms that naturally occurring radioiso-topes can be successfully applied to not only indicate the sourcesbut to quantify the magnitude or fluxes of SGD to the coastal oceanfor example see Moore (2010). Radon is an ideal tracer of SGD(Burnett & Dulaiova, 2003) as a non-reactive noble gas, radon be-haves conservatively and the only losses from the water column aredue to radioactive decay and evasion to the atmosphere (Schmidt etal., 2009). The concentration of radon in groundwater is several or-ders of magnitude higher than in seawater because groundwater isin constant contact with radon emanating aquifer material and, itshalf-life of 3.82 days is comparable to the scale of coastal circulationwith a residence time of up to 15 days (four half-lives) (Cable et al.,1997; Dulaiova et al., 2008). Most importantly, the use of an exclusivetracer of groundwater such as radon precludes any confusion as to the

Fig. 1. Location map of the study area, the Republic of Ireland, illustrating Landsat ETM+ coof six Landsat ETM+ scenes (tabulated) acquired from the USGS and GLCF between paths 20in the analysis.

sources of freshwater inputs to the coastal ocean, a characteristic thatis not available to salinity.

In summary, this study presents an integrated methodology forregional-scale assessment of SGD demonstrated through a casestudy of the Irish coastal zone. Its objectives are threefold:

1. To provide a methodology for deriving SST at 60 m spatial resolu-tion based on publically available Landsat ETM+ TIR data, includ-ing pre-processing to eliminate the effects of the atmosphere, anessential step in the use of thermal imagery for absolute tempera-ture studies. The spatial variability of groundwater dischargearound the coastline of Ireland is determined from thermal anom-alies observed in nearshore SST patterns.

2. To further evaluate the potential of the identified sites for the pres-ence of SGD, ancillary on-shore spatial datasets are combinedwithin a Geographical Information System (GIS) to link geologicaland geomorphological features on land that likely support ground-water discharge. This effectively constrains the initial results anddirects subsequent follow-up studies in the field designed tomeet the third objective.

3. To provide a qualitative assessment of surface water radon activi-ties using radioisotope tracing to confirm the presence of SGD.

The developed methodology is not limited to the coastal zone orstudy area but can be directly applied to a regional assessment ofSGD elsewhere with potential application to lake systems also.

2. Methodology

2.1. Study location

The study area comprises the island of Ireland (Fig. 1), demarcatedby the intercept of a total of 6 Landsat ETM+ scenes from Path 206Row 024 through Path 208 Row 024 of the WRS-2 (Landsat ETM+Worldwide Reference System) coordinate system. The area spansthe coastline from Mayo in the north west (54.30°N, 10.00°W)extending southeast to Wexford (52.34°N, 6.44°W). Ireland's maingeographical features include low lying central plains comprisinglimestone, covered with glacial deposits of clay and sand with numer-ous bogs and lakes bordered by a ring of coastal mountains. The

verage areas by county. The bold outline demarcates the combined geographical extent6 and 208 and rows 022 to 024 of the Landsat World Reference System (WRS) and used

24 J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

mountain ranges in Galway, Mayo and Wicklow are mainly granite(Woodcok & Strachan, 2000). The western and southern coastlinesare rugged with many islands, peninsulas headlands and bays where-as the eastern coastline comprises low, sandy dune backed beacheswith intermittent rocky headlands. The climate of Ireland is definedas temperate maritime moderated by the warm influence of theGulf stream producing generally warm summers and mild winters.Mean annual precipitation in the west of Ireland ranges between1000 and 1400 mm and in the east is around 700 mm (Kiely, 1999).

2.2. Landsat ETM+ thermal data acquisition

To develop a methodology which would facilitate a regional scaleassessment of groundwater discharge from coastal aquifers, the opti-mal time period for SGD detection via remote sensing techniquesmust firstly be established. A comparison of SST acquired from theMarine Institute (MI) Ireland and groundwater temperature datafrom a selection of boreholes sampled by the Environmental Protec-tion Agency (EPA) Ireland in 2008, revealed that maximum tempera-ture differences occur through the summer months (Fig. 2).Consequently, this study is based on the fact that inflow of coolergroundwater into warmer nearshore waters results in buoyantplumes of low salinity water and lower temperatures at the sea sur-face in the zones of groundwater discharge. The Landsat ETM+ TIRwaveband (band 6) is sensitive in the electromagnetic region be-tween 10.4 μm and 12.5 μm, capturing radiant thermal energy at theatmospheric window between ozone and carbon dioxide atmospher-ic absorption, with a spatial resolution of 60 m (Jensen, 2007) and anaccuracy of 0.4 °C–0.6 °C (Tcherepanov et al., 2005).

A total of thirty publically available Landsat ETM+ TIR images ofIreland spanning the time period between May 2001 and June 2010,were acquired from a variety of sources including the EuropeanSpace Agency (ESA), the US Geological Survey (USGS) and the GlobalLand Cover Facility (GLCF) at the University of Maryland USA, six ofwhich are presented as examples here (Fig. 1). The images obtainedwere mostly cloud free with a scene centre flyover time of between11:15 h and 11:30 h GMT (local time). Land pixels in each scenewere masked based on a threshold of Landsat ETM+ band 5 imagevalues from a clear, cloud-free day. The available archived images en-velope almost the entire coastline of Ireland and were used to createtemperature anomaly maps of coastal waters for use in a regionalscale assessment of SGD.

Fig. 2. Comparison of groundwater temperature values recorded from the Bog of Ringboreholes located in Dublin (identified through unique codes and sourced from theEPA) and SST measurements recorded across Dublin Bay by the Celtic Voyager (sourcedfrom the Marine Institute) in 2008.

2.3. Deriving SST values from Landsat ETM+ TIR data

SST values were generated from Landsat ETM+ TIR data usingcommercially available image processing software (ERDAS ImagineAdvantage). To retrieve SST, in this analysis, raw thermal data valueswere converted from calibrated digital numbers (DN) back to at-satellite or top of atmosphere (TOA) radiance (Eq. 1) to correct forgain and bias i.e. the original scaling factors at the detector(Chander & Markham, 2003; Landsat Project Science Office, 2003),as follows:

Lλ ¼ “Gain”� Qcalþ “Bias or of f setð Þ”which is also expressed as:

Lλ ¼ LMAXλ−LMINλ

QCalMAX

� �Qcal−QcalMINð Þ þ LMINλ

ð1Þ

where, Lλ is spectral radiance received at the sensor in watts permetre squared * ster * μm (Wm−2 sr−1 μm−1), Gain is rescaledgain contained in the image product header file (W m−2 sr−1 μm−1),Qcal is quantised calibrated pixel values in DN, Bias (or offset) is therescaled bias contained in the image product header file(W m−2 sr−1 μm−1), QcalMAX is the maximum quantised calibratedpixel value (corresponding to LMAXλ) in DN, QcalMIN is the minimumquantised calibrated pixel value (corresponding to LMINλ) in DN,LMAXλ is spectral radiance that is scaled to QcalMAX(Wm−2 sr−1 μm−1), and LMINλ is spectral radiance that is scaledto QcalMIN (Wm−2 sr−1 μm−1).

LMINλ and LMAXλ radiance values for Landsat ETM+ band 6 weredetermined from the literature and are 3.2 Wm−2 sr−1 μm−1 and12.65 W m−2 sr−1 μm−1 respectively (Landsat Project ScienceOffice, 2003). The DN value for QcalMIN may be 0 or 1 (dependingon when and how the image was processed) and is determinedfrom the accompanying image header file, the DN value for QcalMAXis 255. For each of the datasets processed here, the QcalMin value forband 6 is 1.

Radiance (Wm−2 sr−1 μm−1) converted from DN does not rep-resent true surface radiance but a mixed signal or sum of three differ-ent fractions of energy including emitted radiance from the Earth'ssurface, upwelling radiance from the atmosphere and downwellingradiance from the sky. Removing these atmospheric effects isdescribed as an essential step in the application of thermal band im-agery to absolute temperature studies (Barsi et al., 2003). As no ancil-lary atmospheric data were available at the time of satellite overpassfor the archive of imagery used in this study, an atmospheric correc-tion tool developed for public web site access for the Landsat ETM+thermal band by NASA (which uses the National Centre for Environ-mental Prediction (NCEP) modelled atmospheric global profiles)was used in lieu (Barsi et al., 2003). The correction tool has been ap-plied successfully to the retrieval of sea, lake and land surface temper-atures elsewhere (Srivastava et al., 2009; Wloczyk et al., 2006). Theliterature reports that while temperature values or at-satellitebrightness can be derived from TOA radiance values, in the absenceof atmospheric correction however the difference between TOAbrightness and absolute temperature can range from between 1 Kand 5 K.

Using image acquisition date, time and geographical location asinput, site specific atmospheric parameters transmission, upwellingand downwelling radiances were calculated above (Table 1) and ap-plied to derive water surface radiances (Srivastava et al., 2009)

LλT ¼ LλTOA−Lλupτε

− 1−εð Þε

Lλdown ð2Þ

where LλT is the (surface) radiance of a blackbody target of kinetictemperature T (Wm−2 sr−1 μm−1), LλTOA is TOA or at-satellite radi-ance measured by the instrument (Wm−2 sr−1 μm−1), τ is the

Table 1Landsat ETM+ TIR imagery acquired for Ireland detailing time of satellite overpass and atmospheric correction parameters (upwelling and downwelling radiances, atmospherictransmission) derived from an online atmospheric correction parameter tool (http://atmcorr.gsfc.nasa.gov/) used to derive scene at-surface kinetic temperature values (SST).

Scene acquisition date(yy/mm/dd — hh:mm GMT)

UpwellingWm−2 sr−1 μm−1

DownwellingWm−2 sr−1 μm−1

Transmission%

1999/08/21 — 11:17 No data available for imagery pre 20002000/07/22 — 11:26 2.03 3.24 0.732007/06/08 — 11:25 1.50 2.42 0.792004/05/23 — 11:18 0.99 1.65 0.852004/06/15 — 11:24 1.52 2.43 0.802010/06/02 — 11:15 1.25 2.05 0.82

25J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

atmospheric transmission (unitless), and Lλup is upwelling or atmo-spheric path radiance (Wm−2 sr−1 μm−1) and Lλdown is downwel-ling or sky radiance. The emissivity of water in the Landsat ETM+TIR band ranges from 0.98 to 0.99 and in this study, a constant emis-sivity of 0.989 was used as reported elsewhere (Snyder et al., 1998;Srivastava et al., 2009).

Finally, water surface radiance values derived from Landsat ETM+TIR data are converted into SST (Chander & Markham, 2003; LandsatProject Science Office, 2003):

Tss ¼ K2

ln K1LλT

þ 1� � ð3Þ

where Tss is SST (Kelvin), K1 and K2 are pre-launch thermal calibra-tion constants (666.09 Wm−2 sr−1 μm−1 and 1282.71 K respective-ly) derived from the literature (Landsat Project Science Office, 2003),LλT is the (surface) radiance of a blackbody target of kinetic tempera-ture T (Wm−2 sr−1 μm−1) derived from Eq. (2).

2.4. Deriving Temperature Anomaly (TA) and Standardised TemperatureAnomaly (STA) maps from SST values for SGD detection

In the absence of in-situ temperature data at the time of image ac-quisition, it is extremely difficult to validate sea surface temperaturevalues generated from archived Landsat ETM+ TIR imagery. Howev-er, as the primary objective here is to firstly determine the geograph-ical location of potential sites of SGD along the shoreline and secondlyto determine the relative significance of the temperature patterns ob-served prior to the selection of field sites for (more costly) in-situ ver-ification, the accuracy of the absolute temperature values per se is lessimportant. Of more importance is the ability to produce a standar-dised set of fully inter-comparable thermal anomaly maps that willallow the comparison of the SST values across locations generatedfrom imagery acquired on different calendar dates. To this end, a setof Temperature Anomaly (TA) and Standardised TemperatureAnomaly (STA) maps was derived from each of the SST images.

TA as part of this work is defined as the difference between theSST of each pixel in the image and the average SST temperature

Table 2Average SST and standard deviation values generated for each Landsat ETM+ scene used to c

Scene acquisitiondate yyyy/mm/dd

Coverage area Min SSTK (°C)

1999/08/21 Mayo 284.87 (11.72)2000/07/22 Mayo/Galway 285.09 (11.94)2007/06/08 Galway/Clare 285.41 (12.26)2004/06/15 Kerry/Cork 284.15 (11.00)2004/05/23 Cork 284.40 (11.25)2010/06/02 Waterford/Wexford 284.20 (11.05)

value generated for the coastal water body defined by the spatial ex-tent of each Landsat scene:

TA ¼ Tp−T ð4Þ

where TA denotes temperature anomaly (°C), Tp is the temperaturevalue specific to each pixel in the scene (°C), and �T is the average tem-perature value for the scene (°C).

STA (dimensionless) is defined as the division of TA (Eq. 4) by thestandard deviation (σ) of SST values computed for each of the sixLandsat scenes analysed (Table 2):

STA ¼ TAσ

: ð5Þ

These transforms maintain the spatial pattern of sea surfacetemperatures but all pixel values are rescaled within the TA maps,from −4.0 °C to +3.5 °C and STA maps, from −5.0 to +4.0(dimensionless).

2.5. Integration of ancillary spatial datasets within a GIS to characterisepotential SGD sites

To gauge the potential for groundwater discharge at the sites iden-tified and help direct subsequent in-situ verification in addition to useof the standardised maps, spatial datasets describing onshore bedrockgeology and characterising aquifer groundwater body type were ac-quired from the Geological Survey of Ireland (GSI, 2010). Using thisinformation combined with the results from the thermal mappingwithin a GIS, the locations identified around the coastline can be fur-ther characterised and ranked in their importance as potentialgroundwater sources.

2.6. Geochemical tracing to verify presence of SGD

To verify that the locations of plumes of cooler water observedalong the shoreline are an indication for sites of SGD, geochemicaltracing techniques are employed to provide a qualitative assessmentof surface water radon activities to confirm the presence of ground-water. A pilot survey was conducted around Hook Head off the

reate Temperature Anomaly (TA) and Standardised Temperature Anomaly (STA) maps.

Max SSTK (°C)

Average SST K (°C) No. of pixelsper scene

288.84 (15.69) 286.24 (13.09)±0.425 1.9×107

291.76 (18.61) 289.37 (16.22)±0.929 2.9×105

291.97 (18.82) 287.97 (14.82)±1.192 3.4×106

289.84 (16.69) 287.58 (14.43)±1.151 4.5×106

289.19 (16.04) 285.82 (12.67)±0.475 3.9×106

289.49 (16.34) 287.08 (13.93)±1.041 2.4×107

26 J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

southeast coast of Ireland (Fig. 1) following the results of the thermalimage analysis which consistently revealed the presence of large, coldwater plumes emanating from the peninsula. The Hook, characterisedby an absence of surface (river/stream) drainage and faulted bedrockgeology highly conducive to the transmission of groundwater, wasdeemed a prime site for further investigation. During the surveytwo Durridge RAD-7 radon-in-air monitors in conjunction with twoRADAqua air–water exchanger attachments were used to monitorthe spatial distribution of 222Rn activity in dissolved gas in thewater column continuously from a moving vessel. The air–water ex-changers, fed from two submersible pumps positioned approxi-mately 1 m below the surface, allow radon dissolved in seawaterto de-gas into a closed air loop. The radon-rich air travels througha tube of desiccant before reaching the RAD7 detection chamberwhere counts of the decay of 222Rn to the alpha emitting poloniumdaughters 218Po+ are recorded as a measure of the radon activity inair. Equilibrium between the circulating air and continuouslypumped seawater is typically established after 20–25 min (Lane-Smith et al., 2002). The 222Rn activity in water is calculated from theradonmeasured in the air-loop, determined by the temperature depen-dent air/water equilibrium Fritz–Weigel equation (Weigel, 1978):

222Rnwater¼222Rnair � 0:105þ 0:405e−0:0502T� �

ð6Þ

where 222Rnwater and 222Rnair are the radon activity in water andair respectively and T is the temperature in °C recorded within theair–water exchanger. The RAD7s are factory calibrated and 222Rn activ-ities are presented in this paper using the SI unit Becquerel per metrecubed (Bqm−3).

Data are continuously recorded by both RAD7s in 20 minute inter-vals and the monitor integrated counting intervals are dephased by10 min to provide a unique radon activity reading every 10 minalong the path taken during the pilot survey. 222Rn was continuouslysampled at a distance of between approximately 1 km and 5 km fromthe coast for the first survey and approximately 0.5 km and 1.5 kmfrom the coast for the second. During the first survey several sweepsof the coastline were completed with increasing distance from shore,and during the second survey radon activities were monitored closerto the shoreline in order to determine specific “hot spots” or exitpoints of SGD. A boat speed of 4 knots was maintained whilst sailingparallel to the coastline. The boat position was continuously recordedusing a Garmin GPS and water depths were extracted from nauticalcharts. A CTD probe (Schlumberger Water Services) was positionedabove the pumps beneath the water to continuously record conduc-tivity and temperature from which continuous measurements ofsalinity were derived. In addition, grab sampling of radon was under-taken at two natural springs in close proximity to the shoreline atHook to determine background concentrations of radon within thecoastal aquifer.

Finally, the recorded activity levels are corrected to determineexcess radon activity above background or ambient levels i.e. unsup-ported by Radium-226 (226Ra) present in the ocean. A radium activityvalue of 1.13 Bqm−3 was derived from the literature for the Hook re-gion (Schmidt et al., 1998) and subtracted from the measured activitylevels to determine excess radon activity.

3. Results and discussion

3.1. SST, TA and STA values derived from Landsat ETM+ TIR data of Irishcoastal waters

Clearly discernible cold water plumes emanate from nearshorewaters along the coastline of Ireland (Fig. 3). These constitute thedominant SST feature observed from a selection of temperaturemaps derived from 60 m resolution Landsat ETM+ TIR images

acquired between 11.15 am and 11.30 am GMT (local time), duringsummer months (May to August) between 1999 and 2010. SST inthe processed imagery ranged from a minimum of ~11 °C to a maxi-mum of ~19 °C, with averages falling between 13.09±0.425 °C and16.22±0.929 °C, where error reports the standard deviation betweenall pixels (Table 2).

Average groundwater temperatures in Ireland typically range be-tween 9.5 °C and 11 °C (Aldwell & Burdon, 1986). This is illustratedin Fig. 2 for a collection of boreholes sampled between July andDecember, 2008. The minimum surface temperatures recorded with-in the SST maps range between 11.0 °C and 12.3 °C and manifestwithin the large buoyant cold water plumes off the coastlines ofMayo, Galway, Clare, Kerry, Cork and Wexford.

The significance of the observed temperature anomalies is furtherillustrated through the TA and STA maps. Here, cold water plumes areshown in clearer detail relative to surrounding waters and can beinterpreted to delineate the location and extent of groundwaterdischarge originating from nearshore waters extending to tens ofkilometres offshore. Temperature anomaly defined earlier as the dif-ference between SST and average SST across an entire scene(Section 2.4) ranges from −4.0 °C to +3.5 °C illustrated in Fig. 4. Tofacilitate a context-based inter-comparison of temperature anomalyvalues, the standardised anomaly maps (Fig. 5) reveal the relative sig-nificance of the anomalies observed at different locations. STA rangesfrom −5.0 to +4.0. The largest negative STA values were detectedwithin plumes mapped off the coastline north of Erris Head, Co.Mayo; northwest of Cleggan Co. Galway; off Mizen Head, Co. Corkand off Hook Head, Co. Wexford. These plumes form within metresof the shoreline and extend from over 2 km (e.g. Fig. 5 (d) and (f))to distances greater than 20 km offshore (e.g. Fig. 5 (a) and (b)).

When combined with additional ancillary spatial datasets sourcedfrom the Geological Survey of Ireland (GSI, 2010) the onshore loca-tions of potential sites of SGD, from which the nearshore cold waterplumes appear to originate, lend support to the observation thatadjacent geological features on land may be acting as possiblesources. For instance, visual inspection of the processed Landsatscenes revealed potential SGD sites at 35 locations around the Irishcoastline (Fig. 6), highlighting numerous previously unidentifiedlinks between aquifers on land and the sea. The on-shore location ofthese potential sites is characterised by a faulted, fractured and per-meable bedrock geology comprising predominantly limestone, mud-stone or sandstone (Table 3) associated with locally important(productive) aquifer types (DoELG, 1999) highly conducive to thetransmission of water.

Given these results, it is quite probable that the presence of geo-logic structures on-shore such as karst, bedrock fissures and faults ad-jacent to the thermal plumes is serving as a hydrological pathwaytransporting potentially large volumes of groundwater and associatedmaterials to the sea. The SST maps (Fig. 3) show that the pattern oflarge cold water plumes corresponds almost exclusively to the pres-ence of aquifer fault lines that intersect the shoreline. Additionally,some of the largest negative temperature anomalies were recordedat locations where bedrock fault lines extend several kilometres off-shore. Erris Head Co. Mayo, Cleggan Co. Galway and Hook Head, Co.Wexford (Figs. 4 and 5 (a), (b) and (f) respectively) are particularlygood examples of this and it is most likely that the analysis has iden-tified large offshore submarine springs at these locations.

It is clear that the results of the thermal analysis have highlighted avisual spatial correlation between the location of the thermal plumesand onshore bedrock geology, warranting further investigation.

Sources of freshwater to coastal nearshore waters include surfacerunoff, rivers and groundwater seepage and it cannot be assumedthat the thermal signatures observed are solely due to the presenceof groundwater, nor can one assume that all groundwater seepagepoints from coastal aquifers can be detected via remote sensingtechniques as buoyancy will strongly influence the capacity of the

Fig. 3. Series of SSTmaps (°C) derived from Landsat ETM+TIR imagery acquired on (a) 21/08/1999, (b) 22/07/2000, (c) 08/06/2007, (d) 15/06/2004, (e) 23/05/2004 and (f) 02/06/2010illustrating buoyant cold water plumes emanating from nearshore coastal waters and revealing potential sites for SGD off the Irish coastline. Dashed lines represent bedrock fault linessourced from the GSI.

27J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

thermal sensor to detect a surface signature (Becker, 2005). For in-stance, as fresh water is relatively buoyant compared to saline estuarywaters, thermal signatures of groundwater discharge are easier to de-tect in estuaries compared with fresh water–fresh water interfaces

where relatively cold water will not be detected immediately at thesurface (Moore, 1996).

Successful application of thermal imaging to identify sources ofcoastal groundwater discharge is also constrained by the spatial

Fig. 4. Series of Temperature Anomaly maps (°C) derived from Landsat ETM+ TIR imagery acquired on (a) 21/08/1999, (b) 22/07/2000, (c) 08/06/2007, (d) 15/06/2004, (e) 23/05/2004and (f) 02/06/2010 delineating anomalous cold water plumes off the Irish coastline. Large negative anomalies are illustrated through dark blue through light blue tones and all positiveanomalies greater than or equal to zero are presented in yellow.

28 J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

resolution of the remote sensing system employed. In this study,coastal circulation patterns in nearshore waters derived from SSTswere resolved at 60 m resolution. While the results are very promis-ing for regional scale assessments, it must be acknowledged that the

acquisition of higher resolution imagery obtained through airbornesurveys would likely serve to elucidate finer scale patterns of coastalwater discharge and by doing so, highlight potentially numerous andsignificant inputs of SGD on a local scale. However, the latter is not

Fig. 5. Series of Standardised Temperature Anomalymaps (unitless) derived from Landsat ETM+ TIR imagery acquired (a) 21/08/1999, (b) 22/07/2000, (c) 08/06/2007, (d) 15/06/2004,(e) 23/05/2004 and (f) 02/06/2010 facilitating inter-scene comparison of observed temperature anomaly. Large negative anomalies displayed in dark blue tones reveal the location of coldwater plumes and the relative significance of the anomalies observed at different locations. Areas exhibiting higher than scene average SSTs are displayed through green tones.

29J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

the major objective of this work as smaller scale patterns would bebetter studied perhaps with recourse to actual tracers of groundwaterin-situ such as radon. As the objective here is to constrain monitoringstudies including geochemical surveys to areas most likely to host

SGD to save significant time and resources, the current technique isextremely effective, particularly if potential local constraints towater density differences and their seasonality are incorporated inthe selection of appropriate imagery.

Fig. 6. Location of 35 potential sites of SGD based on visual inspection of SST, temper-ature anomaly and standardised temperature anomaly maps derived from availableLandsat ETM+ TIR images of Ireland acquired during summer months, May–August,from 1999 to 2010.

30 J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

3.2. Geochemical tracing using Radon-222 (222Rn) and salinity

A pilot study was conducted in the southeast of Ireland along theWexford coastline from Ballyhack around Hook Head during August

Table 3On-shore locations of nearshore cold water anomalies and potential sites of SGD detected fring bedrock geology and aquifer type sourced from the GSI (DoELG, 1999).

Code Site County Geology

1 Downpatrick Head Mayo Limestone an2 Glinsk Psammites a3 Benwee Head4 Erris Head5 Sleibhmore Achill6 Clew Bay Limestone an7 Cooltraw Strand8 Roonah Point Sandstone an9 Inishboffin Galway Schist10 Culfin Mudrock and11 Cleggan Quartizite12 Ballyvaughen Bay Clare Limestone an13 Blackhead14 Doolin Cherty limes15 Freaghcastle Sandstone, s16 Caherrush Point17 Carricknola18 Kilkee Siltstone and19 Loop Head Sandstone20 Bolus Kerry Sandstone an21 Dursey Island Cork Sandstone an22 Crows Head23 Sheep's Head Sandstone an24 Dunmanus25 Three Castle Head26 Mizen Head27 Barley Cove28 Cape Clear29 Sherkin Island Mudstone, si30 Toe Head31 Galley Head32 Old Head of Kinsale33 Brownstown Head Waterford Red conglom34 Swines Head35 Hook Head Wexford Limestone an

2010 (Fig. 7) to groundtruth the seasonal thermal anomalies ob-served using Landsat ETM+ TIR imagery acquired in June 2010 (seeFigs. 3–5 (f)) and to verify the presence of SGD while providing aqualitative assessment of fresh groundwater inputs to the coastalzone.

The Hook peninsula and surrounding area form the easternboundary of Waterford Harbour, a natural harbour at the mouth ofthe Three Sisters, the River Nore, Suir and Barrow. The region waschosen as a prime study site for SGD following thermal image analysiswhich revealed seasonally persistent buoyant plumes of cold waterforming within metres of the shoreline of the peninsula. The bedrockgeology of the Hook comprises predominantly permeable rocks in-cluding Carboniferous limestones, sandstones, mudstones, siltstonesand shales lying mostly at less than 60 m elevation (Colfer, 2004)and hence favourable to the presence of SGD (Johannes, 1980). TheHook peninsula is characterised as a locally important productiveaquifer type (DoELG, 1999) as illustrated in Fig. 7. Large deep fissures(known locally as chans) have formed through erosion of bedrockalong the many faults that characterise the coastline. The local soilsare freely drained grey-brown podzolics (till) predominantly of lime-stone composition. There is no surface drainage on the peninsula butthe region is drained further north by a small stream feeding intoDuncannon Beach at Ballystraw.

Results from the surveys determining the spatial distribution ofradon in surface waters around the Hook peninsula confirm the pres-ence of SGD. Excess radon activities in the first survey (Fig. 7 (a))measured between 1 Bqm−3 and 25.6 Bqm−3 for a 25 km stretch ofcoastline and approximately 100 km2 survey area. The highest activ-ity values in the first survey (Fig. 8 (a)) were measured within a kilo-metre from the shoreline at Duncannon extending southwardsaround the peninsula. The second survey found excess radon activity

om SST and TA maps retrieved from available Landsat ETM+ TIR data of Ireland includ-

Aquifer type

d shale Locally important, karstifiednd schists Poorly Productive

d shale Regionally important, karstifiedLocally important, karsified

d siltstone Poor except for local zonesPoor except for local zones

siltstone

d dolomite Regionally important, karstified

toneiltstone and mudstone Locally important, moderately productive

sandstone

d siltstone Locally important, moderately productived siltstone Locally important moderately productive

Poor except for local zonesd mudstone Locally important moderately productive

ltstone, sandstone

Poor except for local zonesLocally important moderately productive

erates, sandstone, mudstone Locally important moderately productive

d dolomites Locally important, karstified

Fig. 7. Hook Head Co. Wexford, location of a pilot study for continuous monitoring of 222Rn to verify the presence of SGD detailing survey tracks completed on a) August 5th andb) August 11th 2010. Sample points above and below the estuarine-seawater mixing line are illustrated with black and white dots respectively. Stars depict sample points above thefreshwater–seawater mixing line (radon hotspots).

31J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

levels between 3.7 Bqm−3 and 78.1 Bqm−3 (Fig. 8 (b)) along thecoastline. Elevated radon activities were measured at a number of lo-cations (radon hotspots) along the peninsula that appear to be linkedto the local geology and geomorphology (Fig. 7 (b)): radon hotspotscoincide spatially with the presence of fault lines and aquifer bodytypes conducive to the transmission of water (Fig. 8 (b)).

The amount of radon present in seawater depends on a number offactors, but essentially, upon the rates of seawater pumping, radonactivity levels within discharging river and groundwater, the produc-tion rate of seafloor sediments, water depth and offshore mixing(Stieglitz et al., 2010). As groundwater or river water discharges tothe sea, radon activities decrease as a result of degassing to the atmo-sphere, radioactive decay and mixing of fresh groundwater with theopen ocean. As such, the position of a sampling location relative to ariver mouth, coastal spring or the open ocean, will affect the radon ac-tivities recorded there. To distinguish between the possible sources ofradon contributing to the activities observed within the nearshorewaters off Hook and more specifically to evaluate whether these arethe result of groundwater seepage at the coast, concurrent salinitysamples were gathered to support the results from the radon surveyand provide a qualitative assessment of apparent fresh-groundwaterinputs.

An inverse relationship between radon activity concentrations ofsurface seawater and salinity was observed for sample pointsmeasured along the first and second surveys (Pearson correlation of−0.81 and −0.86 respectively). The observed negative correlationbetween radon and salinity indicates that waters with lower salinityhave higher radon concentrations due to an admixture of groundwa-ter and this addition of groundwater to the estuary can be furtherexplained by examining the distribution of radon against salinityaround freshwater–seawater mixing lines. For example, if radon asit enters the estuary north of Hook is subject only to mixing anddecay, then the plot of radon against salinity for a set of sample pointsalong a survey through the estuary and beyond the peninsula, in theabsence of the further addition of radon will be a straight line with

a negative slope, due to the higher activity of radon within riverwater compared to seawater.

Fig. 8 (c) displays the plot of radon against salinity for the first sur-vey and includes three theoretical mixing lines generated using fresh-water end-member values from two coastal springs (Rock Well andDuffins Well, Hook Head), an estuarine end-member recorded fromBallystraw Stream at Duncannon and a seawater end-memberrecorded 5 km from the Hook peninsula. Freshwater end-members,particularly the coastal springs, display very high radon activityvalues (16,157 Bqm−3 and 23,516 Bqm−3 respectively) and low sa-linities (0.1 ppt) relative to the seawater end-member which mea-sured very low radon activity (1.05 Bqm−3) and high salinity(31.7 ppt). The mixing lines display the rate of dilution that wouldbe expected between the provided freshwater and seawater end-members. If the observed decrease in radon is primarily due to mixingbetween end-members then radon and salinity should be affectedequally and the spread of radon measurements from the surveywould fall on a straight line between the freshwater and seawaterend-members. Conversely, if the decrease in radon is largely due toatmospheric degassing and radioactive decay, then radon would de-crease more rapidly than salinity and the sample points will fallbelow the mixing line. From the analysis it is clear that all of the sur-vey points are above the theoretical estuarine mixing line implyingthe addition of radon from a local source, thus precluding the estuaryas the primary origin of radon measured in coastal waters off Hook(Fig. 7 (a)).

Fig. 8 (d) displays the plot of radon against salinity for the secondsurvey, where radon was continuously sampled at locations within1 km of the shoreline at Hook. The spread of sample points around thetheoretical mixing lines once again demonstrates the addition ofradon from a source outside the estuarywith the exception of four sam-ple points measured in the upper reaches of the estuary betweenBallyhack and Duncannon (Fig. 7 (b)). Additionally, a number of pointsin the survey plotted just above the fresh groundwater–seawater mix-ing lines revealing radon hotspots or potential sources of SGD at specific

Fig. 8. Excess radon activity values for the coastal waters surrounding Hook recorded during surveys undertaken on August 5th (a) and August 11th (b) 2010. Higher radon activityvalues are displayed in orange through pink to red contour lines, and lower radon activity values are displayed in blue contour lines in units Bqm−3. The plot of salinity versus radonactivity including fresh-seawater and estuarine-seawater mixing lines is also shown for the August 5th (c) and August 11th (d) surveys. Dashed lines represent bedrock fault linessourced from the GSI.

32 J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

locations along the peninsula. These are spatially coincident with thethermal anomalies (Figs. 4 and 5 (f)) and geological features such asthe presence of fault lines and productive aquifer types (Fig. 7 (b)).Data points below the estuarine and groundwater mixing lines illus-trate the impact of decay and atmospheric degassing on measuredradon activities. Minimum salinity values for the second survey(21.1 ppt) are lower than for thefirst (28.4 ppt) indicating the detectionof a larger overall input of fresh groundwater discharge to the coast dur-ing the second survey. The first survey spans a much larger spatial areaallowingmore time for degassing, decay andmixing to occur and this isreflected in the results where radon activities are lower across the firstcompared to the second survey.

4. Conclusions

This research presents a comprehensive cost-effective techniquefor deriving SSTs from the Landsat ETM+ TIR waveband, highlightingthe suitability of the approach for a regional scale survey of potentialSGD locations. The approach was applied to the Irish coastline, iden-tifying 35 potential sites of groundwater seepage from coastal aqui-fers to the sea around the western and southern coast of Irelandand signalling numerous offshore sources.

Once identified by remote sensing, potential sites of SGD can befurther investigated using ancillary spatial datasets describing localgeology (e.g. aquifer productivity and structural geology) and geo-chemical tracers. We present a case-study where concurrent radonand salinity data permit a qualitative assessment of the stretches of

coastline along the Hook peninsula where land–sea fluxes areshown to occur. This analysis supports the remote sensing findingsby revealing a link between the location of fresh cold water plumes,sites of elevated radon activities and on-shore geological features.Mixing curves help distinguish between the possible sources ofradon to coastal waters by eliminating estuarine outflow as the po-tential origin of the observed thermal anomalies and their concurrentradon activities. Additional data such as atmospheric degassing rates,sediment diffusion rates, mixing rates and the spatial distribution ofradon in groundwater and river water for instance would be requiredto provide quantitative estimates of SGD and by doing so, a quantita-tive understanding of groundwater discharge at the site.

The methodology outlined is not limited to the study-area orcoastal zone but can be applied wherever temperature gradientsexist between discharging groundwater and surface water bodiesthus presenting a comprehensive cost-effective tool for coastal man-agers to detect potential land-based sources of pollution to nearshorewaters.

Acknowledgements

This research is funded by the Environmental Protection Agency(EPA), Ireland under the STRIVE initiative (project code 2008-FS-W-S5)hosted by the Biogeochemistry Research Group within the Departmentof Geography, School of Natural Sciences at Trinity College Dublin. Theauthors wish to thank Ms Sinead Kehoe for her help with the radonsurveys.

33J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

References

Abrams, M. (2000). The Advanced Spaceborne Thermal Emission and ReflectionRadiometer (ASTER): Data products for the high spatial resolution imager on NASA'sTerra platform. International Journal of Remote Sensing, 21(5), 847–859.

Akawwi, E., Al-Zouabi, A., Kakish, M., Koehn, F., & Sauter, M. (2008). Using thermalinfrared imagery (TIR) for illustrating the submarine groundwater discharge intothe eastern shoreline of the Dead Sea-Jordan. American Journal of EnvironmentalSciences, 4, 693–700.

Aldwell, C. R., & Burdon, D. J. (1986). Temperature of infiltration and groundwater.Conjunctive Water Use. Proceedings of the Budapest Symposium, July 1986. IAHSPublication Number 156.

Anderson, M. P. (2004). Heat as a ground water tracer. Ground Water, 43, 951–968.Banks, W. S. L., Paylor, R. L., & Hughes, W. B. (1996). Using thermal infrared imagery to

delineate groundwater discharge. Ground Water, 43, 434–443.Barsi, J. A., Schott, J. R., Palluconi, F. D., Helder, D. L., Hook, S. J., & Markham, B. L. (2003).

Landsat TM and ETM+ thermal band calibration. Canadian Journal of RemoteSensing, 29, 141–153.

Beck, A. J., Tsukamoto, Y., Tovar-Sanchez, A., Huerta-Diaz, M., Bokuniewicz, H. J., &Sanudo-Wilhelmy, S. A. (2007). Importance of geochemical transformations in de-termining submarine groundwater discharge-derived tracer metal and nutrientfluxes. Applied Geochemistry, 22, 477–490.

Becker, M. W. (2005). Potential for satellite remote sensing of groundwater. GroundWater, 44, 306–318.

Boehm, A. B., Shellanbarger, G. G., & Paytan, A. (2004). Groundwater discharge: Poten-tial association with fecal indicator bacteria in the surf zone. Environment Science &Technology, 38, 3558–3566.

Burnett, W. C., & Dulaiova, H. (2003). Estimating the dynamics of groundwaterinput into the coastal zone via continuous Radon-222 measurements. Journalof Environmental Radioactivity, 69, 21–35.

Burnett, W. C., Taniguchi, M., & Oberdorfer, J. A. (2001). Assessment of submarinegroundwater discharge into the coastal zone. Journal of Sea Research, 46, 109–116.

Burnett, W. C., Bokuniewicz, H., Huettel, M., Moore, W. S., & Taniguchi, M. (2003).Groundwater and pore water inputs to the coastal zone. Biogeochemistry, 66,3–33.

Cable, J. E., Burnett, W. C., & Chanton, J. P. (1997). Magnitude and variations of ground-water seepage along a Florida marine shoreline. Biogeochemistry, 38, 189–205.

Chander, G., & Markham, B. (2003). Revised Landsat-5 TM radiometric post-calibrationdynamic ranges. IEEE Transactions on Geoscience and Remote Sensing, 41(11),2674–2677.

Clark, C. D. (1993). Satellite remote sensing of marine pollution. International Journal ofRemote Sensing, 14(16), 2985–3004.

Colfer, B. (2004). Irish rural landscapes: Volume two: The Hook Peninsula. CountyWexford:Cork University Press.

Corlett, G. K., Barton, I. J., Donlon, C. J., Edwards, M. C., Good, S. A., Horrocks, L. A., et al.(2006). The accuracy of SST retrievals from AATSR: An initial assessment throughgeophysical validation against in situ radiometers, buoys and other SST data sets.Advances in Space Research, 37(4), 764–769.

Dale, R. K., & Miller, D. C. (2007). Spatial and temporal patterns of salinity and temper-ature at an intertidal groundwater seep. Estuarine, Coastal and Shelf Science, 72,283–298.

Dave, A. (1998). The application of thermal remote sensing to effluent monitoring: TheMount Hope Bay case study: Thesis: Environmental Studies. Providence, RhodeIsland: Brown University, 42 pp.

DoELG (Department of the Environment and Local Government) (1999). GroundwaterProtection Schemes. Dublin: Department of the Environment and LocalGovernment.

Dulaiova, H., Gonneea, M. E., Henderson, P. B., & Charette, M. A. (2008). Geochemicaland physical sources of radon variation in a subterranean estuary — Implicationsfor groundwater radon activities in submarine groundwater discharge studies.Marine Chemistry, 110, 120–127.

Esaias, W. E., Abbott, M., Barone, L., Brown, O., Campbell, J., Carder, K., et al. (1998). Anoverview of MODIS capabilities for ocean science observations. IEEE Transactions onGeoscience and Remote Sensing, 36, 1250–1265.

Fisher, J. I., & Mustard, J. F. (2004). High spatial resolution sea surface climatology fromLandsat thermal infrared data. Remote Sensing of Environment, 90, 293–307.

Fox, M. F., Kester, D. R., & Yoder, J. A. (2005). Spatial and temporal distributions of sur-face temperature and chlorophyll in the Gulf of Maine during 1998 using SeaWiFSand AVHRR imagery. Marine Chemistry, 97, 104–123.

Gibbons, D. E., &Wukelic, G. E. (1989). Application of Landsat thematic mapper data forcoastal thermal plume analysis at Diablo Canyon. Photogrammetric Engineering andRemote Sensing, 55, 903–909.

GSI (2010).DCENR Spatial Data. URL:. http://www.dcenr.gov.ie/Spatial+Data/Geological+Survey+of+Ireland/GSI+Spatial+Data+Downloads.htm last accessed May 25th2011.

Handcock, R. N., Gillespie, A. R., Cherakauer, K. A., Kay, J. E., Burges, S. J., & Kampf, S. K.(2006). Accuracy and uncertainty of thermal-infrared remote sensing of streamtemperatures at multiple spatial scales. Remote Sensing of Environment, 100,427–440.

Hu, C., Muller-Karger, F. E., & Swarzenski, P. W. (2006). Hurricanes, submarine ground-water discharge and Florida's red tides. Geophysical Research Letters, 33, L11601.

Jensen, J. R. (2007). Remote sensing of the environment: an earth resource perspective.Upper Saddle River, New Jersey: Pearson Prentice Hall.

Johannes, R. (1980). The ecological significance of the submarine discharge of ground-water. Marine Ecological Progress Series, 3, 365–373.

Johnson, A. G., Glenn, C. R., Burnett, W. C., Peterson, R., & Lucey, P. G. (2008). Aerialinfrared imaging reveals large nutrient-rich groundwater inputs to the ocean.Geophysical Research Letters, 35, 1–6.

Kiely, G. (1999). Climate change in Ireland from precipitation and streamflow observa-tions. Advances in Water Resources, 23, 141–151.

Kilpatrick, K. A., Podesta, G. P., & Evans, R. (2001). Overview of the NOAA/NASA ad-vanced very high resolution radiometer Pathfinder algorithm for sea surface tem-perature and associated matchup database. Journal of Geophysical Research-Oceans,106, 9179–9197.

Landsat Project Science Office (2003). Landsat 7 Science Data User's handbook. NASAWashington, D.C: Goddard Space Flight Centre.

Lane-Smith, D., Burnett, W. C., & Dulaiova, H. (2002). Continuous Radon-222 measure-ments in the coastal zone. Sea Technology (October 2002), 37–45.

Lee, Y. -W., & Kim, G. (2007). Linking groundwater-borne nutrients and dinoflagellatered-tide outbreaks in the southern sea of Korea using a Ra tracer. Estuarine, Coastaland Shelf Science, 71, 309–317.

Lee, Y. -W., Hwang, D. -W., Kim, G., Lee, W. -C., & Oh, H. -T. (2009). Nutrient inputs fromsubmarine groundwater discharge (SGD) in Masan Bay, an embayment surroundedby heavily industrialized cities, Korea. The Science of the Total Environment, 407,3181–3188.

Leote, C., Ibanhez Severino, J., & Rocha, C. (2008). Submarine groundwater discharge as anitrogen source to the Ria Formosa studiedwith seepagemetres. Biogeochemistry, 88,185–194.

McKenna, T., Andres, A., & Deliberty, T. (2001). Mapping locations of ground-waterdischarge in Rehoboth and Indian River bays, Delaware using thermal imagery.Geological Society of America Abstracts with Programs, 33, p44.

Menzel, W. P., & Purdom, J. (1994). Introducing GOES-I, 1994: The first of a new gener-ation of geostationary operational environmental satellites. Bulletin of the AmericanMeteorological Society, 75, 757–781.

Moore, G. K. (1996). Using thermal-infrared imagery to dilineate groundwater discharge— Discussion. Ground Water, 34 962-962.

Moore, W. S. (1999). The subterranean-estuary: a reaction zone of groundwater andsea water. Marine Chemistry, 65, 111–125.

Moore, W. S. (2010). A reevaluation of submarine groundwater discharge alongthe southeastern coast of North America. Global Biogeochemical Cycles,24(GB4005).

Parkinson, C. L. (2003). Aqua: An earth-observing satellite mission to examine waterand other climate variables. IEEE Transactions on Geoscience and Remote Sensing,41, 173–183.

Portnoy, J. W., Nowicki, B. L., Roman, C. T., & Urish, D. W. (1998). The discharge ofnitrate-contaminated groundwater from developed shoreline to marsh-fringed es-tuary. Water Resources Research, 34, 3095–3104.

Roseen, R. M., Brannaka, L. K., & Ballestero, T. P. (2001). Determination of nutrientloading from groundwater discharge into an inland estuary using airborne thermalimagery. Proceedings of the 12th Biennial Coastal Zone Conference, Cleveland, OH,USA.

Schmidt, A., Reyss, J. -L., Landré, F., & Boust, D. (1998). Distribution and flux of 226Raand 228Ra in the Irish Sea and in the English Channel, in relation with hydrologicalconditions and sediment interactions. Radiation Protection Dosimetry, 75, 65–67.

Schmidt, A., Stringer, C. E., Haferkorn, U., & Schubert, M. (2009). Quantification ofgroundwater discharge into lakes using radon-222. Environmental Geology, 56,855–863.

Shaban, A., Khawlie, M., Abdallah, C., & Faour, G. (2005). Geologic controls of subma-rine groundwater discharge: Application of remote sensing to north Lebanon.Environmental Geology, 47, 512–522.

Slater, L. D., Ntarlagiannis, D., Day-Lweis, F., Mwakanyamale, K., Versteeg, R. J.,Ward, A., et al. (2010). Use of electrical imaging and distributed temperaturesensing methods to characterise surface/groundwater exchange regulatinguranium transport at the Hanford 300 Area, Washington. Water ResourcesResearch, 46, 13.

Slomp, C. P., & Van Capellen, P. (2004). Nutrient inputs to the coastal ocean throughsubmarine groundwater discharge: controls and potential impact. Journal ofHydrology, 295, 64–86.

Snyder, W. C., Wan, Z., Zhang, Y., & Feng, Y. -Z. (1998). Classification-based emissivityfor land surface temperature measurement from space. International Journal ofRemote Sensing, 19(14), 2753–2774.

Srivastava, P. K., Majumdar, T. J., & Bhattacharya, A. K. (2009). Surface temperature es-timation in Singhbum Shear Zone of India using Landsat-7 ETM+ thermal infrareddata. Advances in Space Research, 43, 1563–1574.

Stieglitz, T. C., Cook, P. G., & Burnett, W. C. (2010). Inferring coastal processes fromregional-scale mapping of 222Radon and salinity: Examples from the Great BarrierReef, Australia. Journal of Environmental Radioactivity, 101, 544–552.

Swarzenski, P. W., & Baskaran, M. (2007). Uranium distribution in the coastal watersand pore waters of Tampa Bay, Florida. Marine Chemistry, 104, 43–57.

Tcherepanov, E. N., Zlotnik, V. A., & Henebry, G. M. (2005). Using Landsat thermalimagery and GIS for identification of groundwater discharge into shallowgroundwater-dominated lakes. International Journal of Remote Sensing, 26,3649–3661.

Thomas, A., Byrne, D., & Weatherbee, R. (2002). Coastal sea surface temperaturevariability from Landsat infrared data. Remote Sensing of Environment, 81,262–272.

Ullman, W. J., & Miller, D. C. (2004). Ecological consequences of ground water dischargeto Delaware Bay, United States. Ground Water, 42, 959–970.

Varma, S., Turner, J., & Underschultz, J. (2010). Estimation of submarine groundwaterdischarge into Geographe Bay, Bunbury, Western Australia. Journal of GeochemicalExploration, 106, 197–210.

34 J. Wilson, C. Rocha / Remote Sensing of Environment 119 (2012) 21–34

Wang, L. T., McKenna, T. E., & Deliberty, T. L. (2008). Report of Investigations No. 74:Locating ground-water discharge areas in Rehoboth and Indian River Bays and IndianRiver Delaware using Landsat 7 imagery. Newark, State of Delaware: DelawareGeological Survey.

Weigel, F. (1978). Radon. Chemiker Zeitung, 102, 282.

Wloczyk, C., Richter, R., Borg, E., & Nueberts, W. (2006). Sea and lake surface tempera-ture retrieval from Landsat thermal data in Northern Germany. InternationalJournal of Remote Sensing, 27(12), 2489–2502.

Woodcok, N. H., & Strachan, R. A. (2000). Geological history of Britain and Ireland.London: Blackwell Science Ltd.