geomorphic and sediment volume responses of a coastal dune complex following invasive vegetation...

12
Geomorphic and sediment volume responses of a coastal dune complex following invasive vegetation removal Jordan B. R. Eamer, Ian B. Darke and Ian J. Walker * Department of Geography, University of Victoria, Victoria, British Columbia Canada Received 4 October 2011; Revised 22 January 2013; Accepted 30 January 2013 *Correspondence to: Ian J. Walker, Department of Geography, University of Victoria, Victoria, British Columbia, Canada. E-mail: [email protected] ABSTRACT: This paper documents application of an established geostatistical methodology to detect significant changes in a foredunetransgressive dune complex where Parks Canada Agency (PCA) implemented a dynamic restoration program to remove in- vasive marram grasses (Ammophila spp.) and enhance dynamic dune habitat for an endangered species. Detailed topographic surveys of a 10 320 m 2 site in the Wickaninnish Dunes in Pacific Rim National Park, British Columbia, Canada for the first year post-treatment are compared to a pre-restoration LiDAR baseline survey. The method incorporates inherent spatial structure in measured elevation datasets at the sub-landscape scale and models statistically significant change surfaces within distinct, linked geomorphic units (beach, foredune, transgressive dune complex). Seasonal and annual responses within the complex are discussed and interpreted. All geomorphic units experienced positive sediment budgets following restoration treatment. The beach experienced the highest differential volumetric change (+1656 m 3 ) and net sediment influx (+834 m 3 ,0 19 m 3 m 2 ) mostly from supply to the supratidal beach and incipient dune. This sediment influx occurred independent of the restoration effort and was available as a buffer against wave erosion and as supply to the landward dunes. The foredune received +200 m 3 (0 13 m 3 m -2 ) and its seaward profile returned to a similar pre-restoration form following erosion at the crest from vegetation removal and scarping by high water events. Sediment bypassing and minimal change was evident at the mid-stoss slope with appreciable extension of depositional lobes in the lee. The transgressive dune complex experienced high accretion following restoration activity (+201 m 3 ) and over the year (+284 m 3 ,0 07 m 3 m 2 ) mostly from depositional lobes from the foredune, precipitation ridge growth along the downwind boundary, and growth of existing lobes within the complex. Further integration of this methodology to detect significant geomorphic changes is recommended, particularly for applications where sampling densities are limited or logistically defined. Copyright © 2013 John Wiley & Sons, Ltd. KEYWORDS: foredune; coastal; aeolian; restoration; Ammophila; DEM Introduction Analysis of high resolution digital elevation models (DEMs) generated from terrestrial Light Detecting and Ranging (LiDAR) data has emerged for characterizing volumetric and geomor- phic changes in beachdune systems at the landscape scale (e.g. Woolard and Colby, 2002; Sallenger et al., 2003; Saye et al., 2005; Houser and Hamilton, 2009; Eamer and Walker, 2010). Although highly useful and of very detailed spatial reso- lution, LiDAR-derived DEMs are generally expensive and often temporally constrained. Traditional repeat topographic surveys (e.g. differential global positioning system (DGPS) or total station), however, can provide a sufficient and more affordable means for researchers and managers to assess changes in site geomorphology and sediment volume transfers following an implemented dune restoration treatment. Accurate DEMs and statistically significant change surfaces can be generated by carefully considering the sampling strat- egy, quantifiable errors, and geostatistical properties of each dataset (e.g. Chappell et al., 2003; Heritage et al., 2009; Wheaton et al., 2010). Extensive reviews of various methodological and analytical considerations for DEM modeling and interpretation in geomorphology have been provided by various authors (e.g. Heritage et al., 2009; Wheaton et al., 2010; Milan et al., 2011). Essentially, noise resulting from uncertainty in acquired topo- graphic datasets must be separated from significant elevation changes, however large or small, in each DEM. Once identified, areas of significant change can be differenced between time periods to quantify volumetric changes and related geomorphic responses. The fundamental precision and accuracy of a DEM are a product of survey point quality (resulting from instrument precision), sampling strategy (which controls point density and spacing), sampling frequency and temporal consistency, surface composition (e.g. soft sand versus stable soils), topographic complexity, and chosen spatial interpolation methods (e.g. Heritage et al., 2009; Wheaton et al., 2010). These consider- ations are essential to the ability to detect and compare signifi- cant changes in dynamic landscapes. Geostatistical interpolation methods (e.g. Weighted Least Squares, Kriging) are often used to model the spatial structure and trends in spatially discontinuous data (e.g. x, y , z-elevation data) in order to produce a continuous, representative DEM EARTH SURFACE PROCESSES AND LANDFORMS Earth Surf. Process. Landforms 38, 11481159 (2013) Copyright © 2013 John Wiley & Sons, Ltd. Published online 26 March 2013 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/esp.3403

Upload: ucsb

Post on 07-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

EARTH SURFACE PROCESSES AND LANDFORMSEarth Surf. Process. Landforms 38, 1148–1159 (2013)Copyright © 2013 John Wiley & Sons, Ltd.Published online 26 March 2013 in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/esp.3403

Geomorphic and sediment volume responses of acoastal dune complex following invasivevegetation removalJordan B. R. Eamer, Ian B. Darke and Ian J. Walker*Department of Geography, University of Victoria, Victoria, British Columbia Canada

Received 4 October 2011; Revised 22 January 2013; Accepted 30 January 2013

*Correspondence to: Ian J. Walker, Department of Geography, University of Victoria, Victoria, British Columbia, Canada. E-mail: [email protected]

ABSTRACT: This paper documents application of an established geostatistical methodology to detect significant changes in aforedune–transgressive dune complex where Parks Canada Agency (PCA) implemented a dynamic restoration program to remove in-vasivemarram grasses (Ammophila spp.) and enhance dynamic dune habitat for an endangered species. Detailed topographic surveysof a 10 320 m2 site in theWickaninnish Dunes in Pacific Rim National Park, British Columbia, Canada for the first year post-treatmentare compared to a pre-restoration LiDAR baseline survey. The method incorporates inherent spatial structure in measured elevationdatasets at the sub-landscape scale and models statistically significant change surfaces within distinct, linked geomorphic units(beach, foredune, transgressive dune complex). Seasonal and annual responses within the complex are discussed and interpreted.All geomorphic units experienced positive sediment budgets following restoration treatment. The beach experienced the highest

differential volumetric change (+1656 m3) and net sediment influx (+834 m3, 0 � 19 m3 m–2) mostly from supply to the supratidal beachand incipient dune. This sediment influx occurred independent of the restoration effort and was available as a buffer against waveerosion and as supply to the landward dunes. The foredune received +200 m3 (0 �13 m3 m-2) and its seaward profile returned to asimilar pre-restoration form following erosion at the crest from vegetation removal and scarping by high water events. Sediment bypassingandminimal changewas evident at themid-stoss slopewith appreciable extension of depositional lobes in the lee. The transgressive dunecomplex experienced high accretion following restoration activity (+201 m3) and over the year (+284 m3, 0 �07 m3 m–2) mostly fromdepositional lobes from the foredune, precipitation ridge growth along the downwind boundary, and growth of existing lobeswithin the complex. Further integration of this methodology to detect significant geomorphic changes is recommended, particularlyfor applications where sampling densities are limited or logistically defined. Copyright © 2013 John Wiley & Sons, Ltd.

KEYWORDS: foredune; coastal; aeolian; restoration; Ammophila; DEM

Introduction

Analysis of high resolution digital elevation models (DEMs)generated from terrestrial Light Detecting and Ranging (LiDAR)data has emerged for characterizing volumetric and geomor-phic changes in beach–dune systems at the landscape scale(e.g. Woolard and Colby, 2002; Sallenger et al., 2003; Sayeet al., 2005; Houser and Hamilton, 2009; Eamer and Walker,2010). Although highly useful and of very detailed spatial reso-lution, LiDAR-derived DEMs are generally expensive and oftentemporally constrained. Traditional repeat topographic surveys(e.g. differential global positioning system (DGPS) or totalstation), however, can provide a sufficient and more affordablemeans for researchers and managers to assess changes in sitegeomorphology and sediment volume transfers following animplemented dune restoration treatment.Accurate DEMs and statistically significant change surfaces

can be generated by carefully considering the sampling strat-egy, quantifiable errors, and geostatistical properties of eachdataset (e.g. Chappell et al., 2003; Heritage et al., 2009;Wheatonet al., 2010). Extensive reviews of various methodological and

analytical considerations for DEM modeling and interpretationin geomorphology have been provided by various authors (e.g.Heritage et al., 2009; Wheaton et al., 2010; Milan et al., 2011).Essentially, noise resulting from uncertainty in acquired topo-graphic datasets must be separated from significant elevationchanges, however large or small, in each DEM. Once identified,areas of significant change can be differenced between timeperiods to quantify volumetric changes and related geomorphicresponses. The fundamental precision and accuracy of a DEMare a product of survey point quality (resulting from instrumentprecision), sampling strategy (which controls point density andspacing), sampling frequency and temporal consistency, surfacecomposition (e.g. soft sand versus stable soils), topographiccomplexity, and chosen spatial interpolation methods (e.g.Heritage et al., 2009; Wheaton et al., 2010). These consider-ations are essential to the ability to detect and compare signifi-cant changes in dynamic landscapes.

Geostatistical interpolation methods (e.g. Weighted LeastSquares, Kriging) are often used to model the spatial structureand trends in spatially discontinuous data (e.g. x, y, z-elevationdata) in order to produce a continuous, representative DEM

1149GEOMORPHIC RESPONSES TO INVASIVE VEGETATION REMOVAL

surface at a given point in time. Geostatistics capture spatial as-sociations inherent in these datasets, offer deterministic modelsfor interpolation, and minimize the use of arbitrary parameters(e.g. subjective spatial scales or ranges). Spatial structuremodels (e.g. variograms) associated with these methods areuseful for quantifying and reducing errors associated withDEM generation from field data (e.g. Desmet, 1997; Chappellet al., 2003). Variograms have an observed (experimental) andmodeled component. The observed variogram describes thespatial variation in a variable of interest and various parameters(e.g. sill, range, nugget, per Swales, 2002) are used to developthe best approximation model variogram. The model variogramhas mathematically uniform properties that enable it to be usedfor estimating properties of a surface (e.g. elevation, sedimenttransport, etc.) at unsampled locations to create a model ofthe terrain (digital terrain model, DTM) or some other geomor-phic attribute that, in turn, is useful for interpreting linkagesbetween process and form operating at different spatial andtemporal scales (Swales, 2002; Chappell et al., 2003).Geographic information system (GIS) and geostatistics have

been used increasingly to visualize and analyze spatial-temporalchanges from DEMs of coastal landscapes (e.g. Andrews et al.,2002; Swales, 2002; Woolard and Colby, 2002; Mitasova et al.,2005; Anthony et al., 2006). Different methods of data acquisi-tion (e.g. LiDAR, real time kinematic- global positioning system(RTK-GPS), digital photogrammetry, laser total station surveys)and spatial interpolation models (e.g. Inverse DistanceWeighted,Regularized Spline with Tension, Kriging) have been used, eachwith their respective uncertainties and errors, which makes inter-pretation and comparison of results sometimes difficult. Oftenthese uncertainties are not specified (e.g. Mitasova et al., 2005)or are assigned somewhat arbitrarily as a minimum detectionthreshold (often 5 cm vertical accuracy, e.g. Anthony et al.,2006). If uncertainties are unspecified or unexplored, reportedvolumetric changes and patterns may include a component thatis statistically insignificant and misleading, whereas assignedthresholds may filter out small, but real, changes and patterns.Recent dune restoration efforts provide distinct research oppor-

tunities to quantify sedimentmass transfer processes and interpretmorphodynamic responses that, in turn, can provide usefulinformation for assessing the effectiveness and refinement ofimplemented treatment regimes and broader ecosystemmanage-ment initiatives. Geomorphic research to date has focusedprimarily on measuring changes in active sand surface area orinterpreting cross-shore topographic profiles, with only indirector relatively qualitative measures of aeolian activity and land-scape change (e.g. Arens et al., 2004; Arens and Geelen, 2006;Hilton et al., 2009). The purpose of this research is to apply anestablished geostatistical methodology to quantify and describevolumetric and resulting geomorphic responses within a recentlydestabilized foredune–transgressive dune system on the westcoast of Vancouver Island, British Columbia, Canada. In particu-lar, the study applies statistical change detection methods tohigh-resolutionDEMs obtained fromLiDAR and subsequent lasertotal station surveys to estimate seasonal sand volume changesand describe resulting geomorphic responses.

Study Setting

Study area and environmental setting

The study site is located in the Wickaninnish Dunes complexwithin Pacific Rim National Park Reserve (PRNPR) on the westcoast of Vancouver Island near Ucluelet, British Columbia,Canada (Figure 1). Regional climate is marine west coast cool(Cfb) with an annual average air temperature of 12 �8 �C

Copyright © 2013 John Wiley & Sons, Ltd.

(1971–2000) and high year-round precipitation (3305 �9 mm).The regional wind regime is seasonally bimodal and dominatedby frequent west-northwest (WNW) summer winds, which alignwell with erosional blowouts and depositional lobes in the trans-gressive dune complex, and stronger (and wetter) southeast (SE)winter storm winds, which skew the resultant aeolian sand trans-port potential vector (Beaugrand, 2010) (Figure 2).

Established foredunes in the study area range in height from1 to 5 m and are prograding at a rate of approximately 0 � 2 to0 �5 m a–1 (Heathfield and Walker, 2011) in response to highonshore sand supply and regional tectonic uplift along theCascadia Subduction Zone that is causing relative sea level todrop at a rate of �0 �9 mm a–1 (Mazzotti et al., 2008). The tiderange is mesotidal (spring tide range of 4 � 2 m) and wave regimein the study region is very energetic (average winter significantwave height of 2 �47 m and period of 12 �07 seconds)(Beaugrand, 2010). Erosion by high water events that exceedthe locally derived threshold elevation of the beach–dune junc-tion (5 �5 m above chart datum, or maCD) occurs frequently atthe site with a recurrence interval of approximately 1 � 53 years(annual probability of 65%) (Beaugrand, 2010). Erosive eventsare driven principally by enhanced wave conditions (61 � 5%)and elevated storm surge events (21 �8%) (Heathfield et al.,2012). Dune rebuilding occurs rapidly, however, by way of sandramp development and incipient dune growth, often in thepresence of large woody debris (Heathfield and Walker, 2011).

Dune ecosystems in the region host a number of native plantspecies, predominantly American dune (wildrye) grass (Leymusmollis), beach morning glory (Convolvulus soldanella), beachcarrot (Glehnia littoralis), and yellow sand verbena (Abronialatifolia). Dominant species on the foredune at Wickaninnish,however, are the introduced American and European beach(marram) grasses (Ammophila breviligulata and Ammophilaarenaria, respectively). In the transgressive dunes, vegetationsuccession has lead to encroachment by native Kinnikinnick(Arctostaphylos uva-ursi) and Sitka spruce (Picea sitchensis).

The introduced Ammophila is of special concern due to itsaggressive expansion on foredunes, reduction of dune ecosys-tem biodiversity, and ability to significantly alter foredunesediment budgets and morphodynamics (e.g. Wiedemann andPickart, 1996; Hesp, 2002; Zernetske et al., 2012). At this site,Ammophila spp. have reduced habitat for several endangeredspecies, most notably the provincially endangered Gray beachpea vine (Lathyrus littoralis) and the Pink sand verbena (Abroniaumbellate var breviflora), which is federally red-listed as endan-gered under the Canadian Species at Risk Act (SARA). ParksCanada Agency (PCA) is legally obligated under SARA todevelop a recovery strategy for Pink sand verbena in thePRNPR and in 2009 a five-year dune restoration program wasimplemented and provided the opportunity for this research.The strategy employed by PCA was essentially a ‘processmanagement’ approach (e.g. Riksen et al., 2006) that removesvegetation over large areas with no specific pattern (i.e. broadscale removal on the foredune) bounded by distinct, artificialborders in an effort to enhance aeolian activity and dunemorphodynamics in targeted areas. This was designed to testthe hypothesis that dense Ammophila colonization of theforedune had reduced sand supply to the transgressive dunesystem at Wickaninnish resulting in surface stabilization and adecline of 28% in active sand surface from 1973 to 2007(Heathfield and Walker, 2011). In turn, it was hypothesized thatthis stabilization reduced the availability and suitability of habitatfor the aforementioned endangered species. This research wassupported by PCA as a means to assess initial geomorphic re-sponses and outcomes (i.e. increased aeolian activity, enhancedsand tranfers, etc.) of the restoration treatment as a step toward alonger-term plan to improve and restore dynamic habitat.

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

Figure 1. Study area showing the nearby town of Ucluelet, British Columbia, Canada and inset annual wind rose for the study region derived fromthe Environment Canada Tofino Airport climate station [EC-ID 1038205] for the period 1971 to 1977 by Beaugrand (2010). The study site is locatedwithin the Wickaninnish Dunes complex within the Pacific Rim National Park Reserve (PRNPR).

1150 J. B. R. EAMER ET AL.

Study site location and rationale

A stretch of approximately 3 km of foredune was identified byPCA in 2009 for restoration and, during the first phase of the pro-ject on September 21, 2009, PCA removed Ammophila fromapproximately 200 m of foredune at several sites that straddledpre-existing, cross-shore coastal erosion monitoring profiles(Walker and Beaugrand, 2008; Darke et al., 2013) (Figure 3). Pro-files extend from the forest margin at the back of the transgressivedune complex, over the foredune and down the beach intothe intertidal zone. At each site, plant cover was removedmechanically by a backhoe equipped with a specialized fingerbucket designed to reduce the amount of sand loss duringremoval. Extracted plant debris was buried by PCA in deeptrenches at nearby locations outside of the restoration zones.This study examines changes over the first year following veg-

etation removal at the southeastern-most foredune–transgressivedune complex (Figure 3). The site is fronted by 75 m of fully de-nuded foredune, has an active sand surface area of 10 320 m2,and a perimeter of 924 m encompassed on all landward bound-aries by forest vegetation. These dimensions remained essentiallyconstant over the period of study. This site was selected as it pro-vides a spatially discrete entity for quantifying volumetric andmorphological responses to the restoration treatment. The more

Copyright © 2013 John Wiley & Sons, Ltd.

northerly sites in the larger transgressive dune complex have con-siderably more active sand surface and much longer and openfetch distances in the transgressive dune complex resulting in ap-preciable lateral sand transfers and dunemigration between sites.Discussion of broader rationale for the restoration project as wellas responses at other locations within the Wickaninnish Dunescomplex are provided in Darke et al. (2013).

Methods and Data

To assess changes in site geomorphology and sediment volumetransfers, repeat (approximately bi-monthly) DEMs were pro-duced using detailed topographic survey data imported into aGIS database from which surface geostatistics and change de-tection maps were generated. The methodology for generatingaccurate DEMs with quantifiable error using careful consider-ation of the geostatistical properties of the site is described later.

Topographic survey data

A bare earth basemap of the entireWickaninnishDunes complexwas derived from airborne LiDAR flown on August 27, 2009 just

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

Figure 2. Aeolian sand transport potential roses for the study area (modified from Beaugrand, 2010), including resultant vectors (arrows) showing nettransport magnitudes and directions for the entire year (left) and for selected months (left) that correspond to the modeled change surfaces shown inFigure 5. A strongly bimodal annual transport regime is evident with transport from theWNW prevailing in summer months and from the SE prevailingin winter. Axes represent azimuth (angle) and transport potential (magnitude) in m3 m–1 month–1.

Figure 3. Aerial view of the study site (lower right) and other areaswithin the larger Wickaninnish Dunes complex where invasiveAmmophila spp. was removed (polygons) by PCA as part of a dynamicdune restoration program. This figure is available in colour online atwileyonlinelibrary.com/journal/espl

1151GEOMORPHIC RESPONSES TO INVASIVE VEGETATION REMOVAL

prior to vegetation removal in September. This DEM served as abaseline control for the study and was based on 11 601 pointmeasurements and had an average point density of 1 �13 pointsm–2 (Table I). Subsequent surveyswere conducted using a Topcon

Copyright © 2013 John Wiley & Sons, Ltd.

GTS226 laser total station approximately bi-monthly betweenAugust 2009 and August 2010 (Table I) and measurements weretied into a local grid of georeferenced control points that wereestablished using a RTK-GPS unit (Beaugrand, 2010). The surveydata collection strategy was a systematic nested pattern (e.g.Chappell et al., 2003) (versus a fixed positional grid of recurrentlysurveyed points) that captured detail of significant features and slopeinflection points with grid densities of 0 �04 to 0 �09 points m–2

and closing errors of 0 � 04 to 1 � 65 cm (see Table I). Intervalsurveys were compared to the baseline LiDAR dataset viainterpolated DEMs with similar resolution (described later).

Geomorphic unit delineation

A digital orthophotograph mosaic was constructed for the studysite using imagery obtained during the LiDAR mission. Interpre-tation of the photo mosaic, combined with field reconnais-sance, was then used to delineate the study site into threediscrete geomorphic units: beach, foredune, and transgressivedune complex (Figure 4). These units were held fixed for theperiod of study to allow volumetric response comparisonswithin the same unit over time. As different formative processescontrol sedimentary dynamics within and across these units(i.e. swash zone dynamics, aeolian transport by saltation,grainfall avalanching, etc.), this delineation allows for refine-ment of the interpolation method and provides some geomor-phic rationale for representative geostatistical modeling. Forexample, the observed variogram generated for the beachgeomorphic unit was defined largely by a gradual, upwardlysloping shoreward trend surface and beach cusps. Likewise,the observed variogram for the transgressive dune complexwas defined by erosional blowouts and depositional lobesformed by aeolian processes.

The beach unit was approximately 80 m wide and wasdefined by the seaward limit of the survey on the beach and,on the landward margin, by the toe of the established foredune.

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

Table I. Study site survey metadata including those for the LiDAR base map and subsequent total station surveys.

Date of survey (Julian day) Number of points Point density (pt m–2) Horizontal closing error (m) Vertical closing error (m)

August 27, 2009 (LiDAR) (239) 11601 1 �13 N/A 0 �1500 (assumed)September 24, 2009 (267) 926 0 � 09 0 � 152 0 �0020October 23, 2009 (296) 455 0 � 04 0 � 080 0 �0051December 8, 2009 (342) 480 0 � 05 0 � 078 0 �0111January 15, 2010 (15) 501 0 � 05 0 � 082 0 �0025March 5, 2010 (64) 579 0 � 06 0 � 134 0 �0060April 13, 2010 (103) 532 0 � 05 0 � 022 0 �0004May 30, 2010 (150) 619 0 � 06 0 � 057 0 �0047July 8, 2010 (189) 781 0 � 08 0 � 150 0 �0165August 14, 2010 (226) 483 0 � 05 0 � 127 0 �0156Note: Bare sand surface area of the study site (10 320 m2) remained approximately constant during the period of study.

Figure 4. Enlarged aerial photograph of the 10 320 m2 study siteseparated into discrete geomorphic units (beach, foredune and trans-gressive dune complex) and showing the location of the extractedcross-shore profiles shown in Figure 6. Letters a–e indicate locationsof vantage photographs shown in Figure 8. This figure is available incolour online at wileyonlinelibrary.com/journal/espl

1152 J. B. R. EAMER ET AL.

As such, the landward portion of the beach unit included anephemeral incipient dune that formedwithin seasonal vegetationin the backshore (Figure 4). The foredune unit extended landwardfrom the beach unit either to the landward extent of foredunevegetation or to the edge of depositional lobes extending fromthe foredune crest as identified in the original LiDAR survey.The transgressive dune complex was defined landward of theforedune unit and its remaining perimeter was defined bythe extent of active sand surface in the initial survey. As earlier,the total study site area (10 320 m2) and outer perimeter (924 m)remained essentially constant over the period of study.

Geostatistical modeling

The spatial structure of survey datasets must be consideredwhen converting field data into modeled continuous surfaces.To do so, variogram results are used to quantify scales of spatialautocorrelation that, in turn, can be incorporated into modeledDEM surfaces. LiDAR and topographic survey DEM datasetswere imported into QGIS© and separated into geomorphicunits using a spatial intersect join, which crops the datasetusing a defined polygon for each geomorphic unit. These data

Copyright © 2013 John Wiley & Sons, Ltd.

were then exported to the R statistical software package andanalyzed with two modules. First, the geostatistical package,geoR (Ribeiro and Diggle, 2001) was used to analyze the spa-tial structure of surface elevation, produce observed and modelvariograms, and to fill a DEM with gridded model results.Second, RGDAL (R Geospatial Data Abstraction Library) wasused to export the DEMs back into QGIS©.

Observed variograms were produced in geoR for each unitand these revealed that the spatial structure of surface elevationdata was found to be anisotropic (i.e. distinct directionality wasevident in the spatial structure of surface elevation), whichis consistent with other studies in coastal and aeolian settings(e.g. Swales, 2002; Chappell et al., 2003). However, as the orien-tation of the geomorphic units was elongated in the alongshoredirection, there were often insufficient data (i.e. n=30–50)available in the direction of anisotropy to produce directionalmodel variograms. Thus, non-directional model variograms weregenerated, as is recommended for studies with lower-resolutiondata (Chappell et al., 2003), using initial, researcher-derivedbest-fit model parameters from an interactive tool in geoR.Gaussian or spherical models best represented the semi-variancein all cases (temporally and across all three units). Model param-eters were then refined in geoR using ordinary least squares toprovide a best fit for the models to the observed variograms.Model performance was assessed using cross-validation in geoR,a method that removes one datapoint from the set, predicts a sur-facemodel based on the remaining datapoints, and computes thedifference between the measured and modeled point. This wasrepeated for each datapoint and accuracy statistics were gener-ated for the model.

Results of the cross-validation analysis (Table II) show arange in combined error of 0 �0064 to 0 �057 m and, generally,that the most accurate interpolation (lowest combined errorvalues) was within the transgressive dune unit, which is mostlikely due to higher point densities that allow for improvedmodel specification and refinement. The most varianceoccurred within the beach unit (resulting from lower pointdensities) and the most consistency between interpolations(based on mean cross-validation and combined errors) waswithin the foredune unit. Combined uncertainty in each datasetwas calculated as the sum of the 95% confidence intervalsurrounding the mean cross-validation error (based on meanand standard deviation of the cross-validation results) addedto the stated instrument precision (0 �005 m) (Table II).

Following model variogram development and cross-validation, geoR was used to interpolate DEM surfaces usingordinary Kriging with weights derived from each of the modelvariograms and a grid density of 1 m (4285, 1565, and 4470 gridcells in transgressive dune, foredune, and beach units, respec-tively). This spacing was chosen to approximate the samplingdensity of the baseline LiDAR data (Table I) and a 1–2 m spacing

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

Table II. Results of error assessment for this study showing mean cross-validation (c-v) error, standard deviation of the cross-validation error (s c-v)across all sampled locations, and combined error (mean+95% confidence interval + instrument precision).

Date of survey (Julian day) Geomorphic unit Mean c-v error (m) s c-v (m) Combined error (m)

September 24 (267) Transgressive Dune �0 �0025 0 � 21 0 � 0084Foredune �0 �0086 0 � 19 0 � 0160Beach �0 �0007 0 � 31 0 � 0570

October 23 (296) Transgressive dune �0 �0001 0 � 25 0 � 0070Foredune �0 �0160 0 � 28 0 � 0280Beach �0 �0260 0 � 27 0 � 0410

December 8 (342) Transgressive dune 0 � 0024 0 � 26 0 � 0100Foredune �0 �0088 0 � 27 0 � 0180Beach �0 �0014 0 � 18 0 � 0110

January 15 (15) Transgressive dune �0 �0018 0 � 22 0 � 0084Foredune �0 �0066 0 � 24 0 � 0160Beach �0 �0230 0 � 29 0 � 0390

March 5 (64) Transgressive dune �0 �0006 0 � 23 0 � 0073Foredune �0 �0067 0 � 28 0 � 0160Beach �0 �0016 0 � 28 0 � 0130

April 13 (103) Transgressive dune 0 � 0007 0 � 33 0 � 0083Foredune �0 �0092 0 � 26 0 � 0180Beach �0 �0007 0 � 24 0 � 0120

May 30 (150) Transgressive dune 0 � 0007 0 � 32 0 � 0080Foredune �0 �0093 0 � 19 0 � 0160Beach �0 �0005 0 � 16 0 � 0080

July 8 (189) Transgressive dune 0 � 0004 0 � 19 0 � 0064Foredune �0 �0110 0 � 17 0 � 0170Beach �0 �0042 0 � 16 0 � 0120

August 14 (226) Transgressive dune 0 � 0015 0 � 29 0 � 0088Foredune �0 �0130 0 � 21 0 � 0220Beach �0 �0100 0 � 22 0 � 0220

Average �0 �0058 0 � 27 0 � 0170

1153GEOMORPHIC RESPONSES TO INVASIVE VEGETATION REMOVAL

is suggested as sufficient to capture morphological and volumet-ric changes within coastal dune systems on a monthly timescale(Woolard and Colby, 2002).

Volumetric calculations and geomorphic changemap generation

DEMs were imported to the Geomorphic Change Detection(GCD) package (Wheaton et al., 2010), which was developedprimarily for sediment budgeting and morphological changedetection in river systems. The software calculates volumetricchanges in sediment storage from repeat topographic surveys,quantifies associated errors, and identifies statistically signifi-cant changes. In GCD there are several different functions foraccounting for error, and this study employed a relativelysimple, computationally lightweight method that involvesusing a probability distribution to model error for each surface.An important part of the functionality of this specific method isthe ability to account for type 1 error, where a true null hypoth-esis (i.e. no surface change) would be incorrectly rejected, byremoving survey and interpolation noise from the results. Thecombined error (described earlier, Table II) was used in theGCD software to assess significant changes based on the stu-dent’s t distribution and a test statistic (Wheaton et al., 2010)based on changes in elevation relative to the uncertainties de-scribed earlier. Thus, the errors reported in Table II represent aminimum uncertainty with a larger minimum change thresholdresulting from t-tests based on these uncertainties and the distri-bution of elevation data for each survey. Furthermore, as all sur-vey datasets are normalized against the initial LiDAR baselineDEM, inherent uncertainties in the LiDAR dataset (�0 �15m)are not included in these calculations. It would be redundantto include a constant uncertainty in relative temporal change

Copyright © 2013 John Wiley & Sons, Ltd.

surfaces (i.e. LiDAR elevation uncertainties are constant and itis the subsequent survey uncertainties that change). Significantvolumetric change estimates were then derived by essentiallysubtracting grid cell elevations of the baseline LiDAR DEM fromthe respective survey DEM at successive time intervals (Table III).

Difference maps for all units and survey dates were producedand imported into QGIS© for overlay onto site orthophotographsand graphical preparations (e.g. Figure 5). The difference mapsvisualize a progression of seasonal, statistically significant surfacechanges within the beach–dune system following restorationfrom which volumetric calculations and geomorphic interpreta-tions were derived. To complement the volumetric estimatesand surface change maps and to provide additional two-dimensional visualization of changes across the units, cross-shore profiles were extracted from the interpolated DEMs alongthe coordinates of an existing coastal erosion monitoringprofile (dashed transect in Figure 4). A detailed subset of theseprofiles showing changes across the restored foredune region ispresented in Figure 6.

Results

Surface change detection maps (Figure 5) visualize statisticallysignificant, seasonal responses of the beach–dune system follow-ing vegetation removal. For clarity and ease of interpretation, aselection of five of the nine interval datasets (October 23, March 5,April 13,May 30, August 14) is presented. Corresponding volumet-ric change values for these intervals, and those intervening, areprovided in Table III and extracted cross-shore topographic profilesare shown in Figure 6. Figure 7 shows sediment budget responsesas normalized volumetric changes within each unit over time.

Over the year following vegetation removal, all threegeomorphic units experienced net positive sediment budgetsas indicated by increases in normalized sediment volumes

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

Figure 5. Significant geomorphic change maps of the study site for selected dates (Julian days) to show seasonal and annual (lower map) statisticallysignificant spatial trends in surface responses (erosion= lighter shades, deposition =darker shades). This figure is available in colour online atwileyonlinelibrary.com/journal/espl

Table III. Estimates of statistically significant sediment volume changes within geomorphic units.

Volumetric change from LiDAR baseline (m3) and area-normalized (m3 m–2) in brackets

Survey Beach: 4470 m2 Foredune: 1565 m2 Transgressive dune: 4285 m2

September 24 +715 (+0 �16) +18 (+0 �01) +201 (+0 �05)October 23 +977 (+0 �22) +101 (+0 �06) +123 (+0 �03)December 8 +373 (+0 �08) +137 (+0 �09) +179 (+0 �04)January 15 �101 (�0 � 02) +91 (+0 �06) +126 (+0 �03)March 5 �679 (�0 � 15) +113 (+0 �07) +118 (+0 �03)April 13 +721 (+0 �16) +97 (+0 �06) +345 (+0 �08)May 30 +410 (+0 �09) +129 (+0 �08) +373 (+0 �09)July 8 +917 (+0 �21) +142 (+0 �09) +204 (+0 �05)August 14 (annual total) +834 (+0 �19) +200 (+0 �13) +284 (+0 �07)Note: Area-normalized values (in m3 m–2) provide an effective depth of average sediment accretion (+) or erosion (�) within each unit. Area of eachunit is indicated at the top of each column.

1154 J. B. R. EAMER ET AL.

(Table III, Figure 7). The beach unit received+834m3 (0 � 19m3m–2),the foredune unit gained +200 m3 (0 �13 m3 m–2), and the trans-gressive dune increased in volume by +284 m3 (0 �07 m3 m–2).The beach was also the only geomorphic unit to experience

Copyright © 2013 John Wiley & Sons, Ltd.

erosion relative to the baseline LiDAR survey and had the highestvariance in volumetric change (+1656 m3). Significant accretionoccurred within the beach unit in the fall (September to October)totaling +977 m3 (0 �22 m3 m–2) resulting from bar welding and

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

Figure 7. Sediment budget responses of geomorphic units through the year of observation following restoration treatment shown as area-normalizedvolumetric changes (in m3 m–2), which provide an indication of effective depth of sediment accretion (+) or erosion (�).

Figure 6. Cross-shore topographic profiles extracted from themodeledDEMsurfaces shown in Figure 5.Geomorphic unit boundaries indicatedby the dashed linesand restored zone where vegetation was removed in September 2009 is also indicated. This figure is available in colour online at wileyonlinelibrary.com/journal/espl

1155GEOMORPHIC RESPONSES TO INVASIVE VEGETATION REMOVAL

berm development, followed by beach steepening and significanterosion during the winter months (�1656 m3 or �0 �37 m3 m–2

from October through March) (Table III, Figures 6 and 7). Mostof the accretion on the beach occurred within the incipient duneand upper (supratidal) beach (Figure 5) although the incipientdune was then eroded between October and March and thelower seaward slope of the foredune was scarped and retreatedby approximately 2 m (Figure 6). Erosion of the upper beachand lower foredune at this site is driven principally by enhancedwave runup and elevated surges during winter storms (Heathfieldet al., 2012). Beach rebuilding was rapid between March andApril (approximately 36 m3 day–1, Figure 7) likely due toincreased occurrence of seasonal transporting wind events(Figure 2) and lower energy wave conditions conducive toonshore movement of sediment in the nearshore, as evidencedby vertical accretion and lowering of the slope on the beachprofile (Figure 6). In early spring (betweenApril andMay surveys),the accretion volume in the beach unit declined in response tonotable erosion on the intertidal beach (Figure 5) whilst the upperbeach accreted and steepened by way of aeolian sand rampdevelopment and incipient dune re-establishement in seasonalvegetation (Figure 6). Through the spring and summer months(April through August), the beach re-established a positive sedi-ment budget (Table III) and aeolian delivery to the upper beachresulted in as much as 0 �4 m of accretion within the incipient

Copyright © 2013 John Wiley & Sons, Ltd.

dune and foredune toe region (Figure 6). Overall, the beach unitexhibited a positive sediment budget +834 m3 (0 �19 m3 m–2) ayear following vegetation removal, which indicates that anappreciable supply of sediment was available for landwarddelivery to the dune systems.

The foredune unit maintained a net positive sediment budgetof +200 m3 (0 �13 m3 m–2) over the year, which is about 24% ofthat experienced on the adjoining beach. Only slight accretion(+18 m3 or 0 �01 m3 m–2) occurred following the vegetationremoval in September but deposition increases by an order ofmagnitude by December (+137 m3 or 0 �09 m3 m–2). Cross-shore profiles show a distinct pivot point near the beach–dunejunction elevation at 5 �5 maCD that marks a notable changefrom a depositional setting on the beach below to an erosionalone on the foredune slope (Figure 6) resulting from somecombination of wind erosion and scarping by high wave runupevents, mostly between October and May. As mentionedearlier, the foredune toe retreated by about 2 m during thistime, however, by early summer increased accretion rates(+0 �5 cm day–1 between April and May, Figure 7) rebuilt theeroded scarp (Figuers 8a and 8b). The crest region showedappreciable aeolian activity during the year as evidenced bysporadic patches of significant erosional and depositionalsurface change (Figure 5). Erosion of approximately 0 �25 moccurred at the crest in October (Figure 6), which may reflect

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

Figure 8. (a) and (b) Vantage photographs looking landward toward the foredune from March and July, respectively. The foredune toe region isscarped in March and rebuilds by July with accretion and re-establishment of the incipient dune zone on the supratidal beach. (c) and (d) Vantagephotographs looking seaward on the lee side of the foredune crest in December and April, respectively, showing growth and extension of a deposi-tional lobe extending from the crest that buries established vegetation and delivers sediment to the transgressive dune. (e) Growth and transgression ofthe precipitation ridge into forest along the southeastern edge of the transgressive dune complex. This figure is available in colour online atwileyonlinelibrary.com/journal/espl

1156 J. B. R. EAMER ET AL.

sediment losses during the mechanical removal process inSeptember. Following this, the crest accreted 0 �5 m quickly(+0 � 23 cm day–1) by the May survey, followed by erosion of0 �2 m by August (�0 � 26 cm day–1). Most of the accretion inthe foredune unit occurs within depositional lobes that extendleeward from the crest (Figures 5, 6, 8c and 8d) toward thesoutheastern end of the unit. Overall, the foredune profilereturns to a very similar form and height to that of the baselineLiDAR profile with additional accretion and resultingprogradation of approximately 1 to 2 m in the toe region andextension of depositional lobes in the lee (Figures 5, 6, 8cand 8d). The mid-stoss slope (between 7 and 7 � 5 maCD,Figure 6) appears to be a location of sediment bypassing andminimal morphological change throughout the year while sur-plus sediment from the beach and lower stoss slope are movedlandward by aeolian transport to rebuild the crest and feed theexpanding depositional lobes in the lee. This rebuilding andlandward extension of the foredune occurred in response toincreased sediment supply to the upper beach, mostly throughto the May survey (Figures 5–7).The transgressive dune complex is perhaps the only unit within

which geomorphic and sediment volume responses arecontrolled almost entirely by aeolian activity. Sediment influx tothe transgressive dune complex is high immediately followingvegetation removal in September (+201 m3 or 0 �05 m3 m–2)and an order of magnitude greater than that on the foreduneduring this time. The surface experienced denudation during

Copyright © 2013 John Wiley & Sons, Ltd.

the winter months (Table III, Figure 7) then increased rapidly(+5 �68 m3 day–1) between March and April whilst the foreduneunit experienced a subtle rate of deflation (�0 �4 m3 day–1) duringthe same period. Accretion in the transgressive dune complex roseto a peak of +373 m3 (0 �09 m3 m–2) in May, mostly in response tohigh inputs from the extending depositional lobes in the lee of theforedune, where asmuch as 0 �4m of sandwas deposited over theyear (Figures 5 and 6) and bymigration of small, discrete dunes thateventually fed into the precipitation ridge on the eastern edge ofthe complex (Figures 5 and 8d). This increase in accretion volumesin early summer reflects some combination of increased and moreonshore oriented sand transport potential (Figure 2) as well asimproved sand transport pathways over the foredune followingrebuilding of the scarped seaward slope. Over the year followingvegetation removal, the transgressive dune accreted 1 �42 timesmore sediment than that of the foredune (+284 versus +200 m3,respectively), yet, distributed over almost three times the surfacearea, this translates to just over a half of the normalized accretiondepth (0 � 07 versus 0 �13 m3 m–2, Table III).

Discussion

Methodological implications

The methodology employed in this study is derived from previ-ous research using geostatistical analysis of DEMs to identify

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

1157GEOMORPHIC RESPONSES TO INVASIVE VEGETATION REMOVAL

and quantify geomorphic and volumetric changes in sedimen-tary landscapes. It is often difficult, however, to compare resultsbetween such studies largely due to inconsistencies in DEMgeneration and varying or unspecified handling of error (e.g.Andrews et al., 2002; Chappell et al., 2003; Swales, 2002;Mitasova et al., 2005; Anthony et al., 2006, 2007; Wheatonet al., 2010). This study uses an easily implemented and cost ef-fective methodology that employs common topographic surveydatasets and the open-source GCD package (Wheaton et al.,2010), which provides a statistically robust means to quantifyand incorporate uncertainty resulting from both instrumentand interpolation error. This approach is highly useful formodeling dynamic and sometimes subtly changing coastaldune landscapes as it provides a systematic and accurate pro-cess for distinguishing between simply small, yet observable,changes (i.e. noise) from those that are truly statistically signifi-cant and it avoids the use of arbitrary thresholds. Although thismethod of probabilistic threshold generation is not new to geo-morphological research (e.g. Lane et al., 2003), its applicationin coastal geomorphology is limited as is its use for detectingchange within linked geomorphic units (see later). Wheatonet al. (2010) caution that, in cases with lower point densities,greater areas within the DEM used for change detection analy-sis may yield insignificant results. Further integration of thismethodology is recommended for geomorphic applicationswhere lower point densities are common or logistically de-fined, higher resolution aerial or ground-based LiDAR surveysare available, or more advanced fuzzy inference systems canbe incorporated to utilize the full functionality of the GCDmethod (Wheaton et al., 2010).A key distinction of this study involves delineation of the

landscape into discrete geomorphic units, which provides threekey advantages. First, delineation allows for identification ofunits whose process-response dynamics are controlled by vary-ing factors including slope and aspect, effective fetch (sedimentsupply), surface roughness, grain size, vegetation cover,moisture content, etc. This recognizes that different processes(i.e. swash zone dynamics, aeolian transport, vegetation–trans-port interactions, etc.) and other factors control observedresponses at the sub-landscape scale. In turn, this allows foranalysis and interpretation of sedimentary exchanges andresulting morphodynamics within and between linked units inthe landscape, as opposed to considering only one larger land-scape unit wherein process-response dynamics are controlledby a broader suite of processes and morphological controls.The second advantage of this approach is that it captures and

models inherent spatial structure in measured datasets at thesub-landscape scale as controlled by the aforementioned pro-cess-response dynamics. This allows for refinement of the inter-polation method and provides some geomorphic rationale forrepresentative geostatistical modeling. For instance, Krigingconsiders directionality and variation distance (or orientationand elongation of features, respectively) and unit delineationallows for modeling to be compartmentalized into regionswhere processes and features are mostly related. For example,the spatial patterns in many coastal dune systems typicallyshow morphological alignment (i.e. anisotropy in the spatialstructure of elevation data) controlled by dominant sand trans-port pathways, such as alongshore orientation of foredunes(aligned essentially normal to onshore transport pathways fromthe beach) (e.g. Hesp, 2002; Walker et al., 2006) and obliqueorientation of blowouts and their depositional lobes (oftenaligned parallel to a dominant mode in the transportingwind regime) (e.g. Hesp, 2002; Anderson and Walker, 2006).Observed variogram results can be used to quantify scales ofspatial autocorrelation that, in turn, can inform a Krigingapproach to produce accurate and systematically repeatable

Copyright © 2013 John Wiley & Sons, Ltd.

modeled DEM surfaces from floating point cloud data, typicalof field surveys. This requires, however, that the size of, and/or sampling density within, a geomorphic unit is sufficient(i.e. n= 30–50 observations) to capture the spatial structureand produce a directional model variogram. If not (as was thecase in this study), non-directional model variograms should beused (Chappell et al., 2003) with best-fit refinement and cross-validation to observed variograms as described earlier.

Third, separation of this portion of the Wickaninnish Dunecomplex into distinct geomorphic units allowed for increasedaccuracy in volumetric and geomorphic change detection forthe purposes of assessing sediment budget and landscaperesponses to vegetation removal from the foredune (expandedfurther later). Methodologically, this allows for comparison ofsignificant sediment exchanges and related geomorphicresponses within and between the beach (sediment source),the foredune (source and sink), and transgressive dune system(sink) to a pre-restoration control surface. While simple com-parison of cross-validation results for the units with interpola-tions of the entire study site revealed that the methodologyshows small increases in interpolation accuracy, a highersampling density within units could have enabled utilizationof variogram directionality (cf. Swales, 2002) and may haveyielded improved results.

Evolution of the beach–dune system one year aftervegetation removal

Following vegetation removal, the foredune–transgressive dunecomplex at Wickaninnish Beach showed significant increasesin sediment volumes (as described earlier) in response to thecombined effects of high sand supply to the beach (source),and reduced vegetation cover on the foredune, which im-proved sand delivery to the transgressive dune complex (sink).The seaward portion of the foredune profile returned to a formvery similar to that of the baseline following: (i) erosion and re-building of the crest by approximately 0 �5 m following vegeta-tion removal (Figure 6), and (ii) re-establishment of a gradualstoss slope profile following erosion of the foredune toe and re-moval of the incipient dune by high wave runup events duringthe winter. New and extending lobes in the foredune crest re-gion and the re-activation and growth of the precipitation ridgeon the downwind (southeastern) edge of the transgressive dunecomplex were observed (Figure 8) as well as growing lobesand deflation basins within the transgressive dune complex(Figure 5). Finally, deflation was observed in the transgressivedune complex in the winter months and also between Mayand July (Figure 7). Both periods of deflation can be explainedby the nature of inputs to the transgressive dune complex beingpurely aeolian. In the winter months, more frequent, higher el-evation surge events (and generally wetter surface conditions)combined with a steeper, shorter beach (Figure 5) and a morealong-shore competent wind regime (Figure 2) promotereduced fetch (supply) conditions on the beach (Davidson-Arnott et al., 2005). Thus, it is very likely that competent windsfor aeolian transport would have been limited in their ability totransport sediment landward from the beach into the transgres-sive dune complex but instead transport existing sediments inthe transgressive dune complex out of the study area into thesurrounding forest. Combined, this resulted in a stable to negativenet sediment volume change. Sediment volume losses in thetransgressive dune complex between May and July were in re-sponse to the migration of a megacusp embayment into the studyarea (Figure 5) which also led to a reduced beach fetch (due to ashorter, steeper beach as above) and less available sediments.

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

1158 J. B. R. EAMER ET AL.

These responses indicate that, in the first year followingvegetation removal, the study site experienced positive sedi-ment budgets and relatively rapid geomorphic recovery withinthe beach and foredune units. Prior to vegetation removal(between 1973 and 2007), Heathfield and Walker (2011) docu-mented a 28% decline in active sand surface area within thelarger foredune–transgressive dune complex and suggested thatcolonization by Ammophila spp. on the foredune andencroachment by native Kinnikinnik (Arctostaphylos uva-ursi)and Sitka spruce (Picea sitchensis) on the transgressive duneperimeter were responsible. It has been documented elsewherethat foredunes colonized by invasive Ammophila experiencealterations in their sediment budgets that result in reduced land-ward sand transfers and dune steepening (e.g. Wiedemann andPickart, 1996; Hesp, 2002). In contrast, foredunes in the studyregion vegetated with the native dune grass (Leymus mollis),which has a lower density growth pattern, tend to be less steep,more hummocky, and more dynamic in their behavior (Pageet al., 2011). Observed historical dune stabilization rates atthe site and regional climate trends toward wetter, warmer con-ditions in coastal British Columbia (Walker and Sydneysmith,2008) suggest that, without restoration intervention to maintainsand supply and/or enhance aeolian activity within this dunesystem, complete stabilization of the transgressive dunes couldoccur within a few decades (Heathfield and Walker, 2011).

Conclusions

This paper documents the application of an established, statis-tically grounded method for geomorphic change detection inriver floodplains (Wheaton et al., 2010) to quantify sedimentvolume changes and geomorphic response patterns in a coastaldune setting following mechanical removal of invasive vegeta-tion. Rationale and key goals of the broader restoration projectare discussed in Darke et al. (2013) and include enhancingsediment exchanges and morphodynamics within and betweenthe foredune and transgressive dune system. A geostatisticalapproach using Kriging was used to model statistically signifi-cant changes within discrete, linked geomorphic units (beach,foredune, and transgressive dune), which captures and incorpo-rates inherent spatial structure in measured elevation datasets atthe sub-landscape scale as controlled by distinctly differentprocess-response dynamics. The method produced low averagecross-validation error (0 �0058 m), a robust accounting ofuncertainty in DEM generation, statistically significant volumet-ric changes, and accurate surface change (erosion/deposition)maps. Further integration of this methodology into geomorphicapplications where lower point densities are common or logisti-cally defined is recommended. In this study, a slightly highersampling density could have enabled use of directionalvariograms (cf. Swales, 2002) that may have yielded improvedresults.Seasonal and annual responses within the beach–dune system

are interpreted as follows:

1. The beach was the only geomorphic unit to experience aninterval of net erosion during the period of observationand had the highest total (differential) volumetric change(+1656 m3) relative to the pre-restoration surface. Overall,the beach maintained a positive sediment budget (+834 m3

or 0 �19 m3 m–2), largely due to a net supply of sand to thesupratidal beach and incipient dune zone. This store of sedi-ment was available both as a buffer against wave erosion andas supply to the foredune and transgressive dune systems,

2. The foredune experienced a net positive sediment budget(+200 m3 or 0 �13 m3 m–2) in the year following vegetation

Copyright © 2013 John Wiley & Sons, Ltd.

removal. The disturbed seaward portion of its profilereturned to a similar form following significant erosion atthe crest related to vegetation removal and scarping of thelower seaward slope by high water events. During thisrebuilding, the mid-stoss slope was a location of sedimentbypassing and minimal morphological change whilesurplus sediment from the beach was moved landward torebuild the crest and feed the expanding depositional lobesin the lee.

3. The transgressive dune also experienced a positive annualsediment budget (+284 m3 or 0 �07 m3 m–2) and accretionwas high (+201 m3) immediately following vegetationremoval, then declined during the winter months, andincreased appreciably in the spring. This was attributed todepositional lobes extending leeward from the foredune,growth of a precipitation ridge along the southeasterly(downwind) edge of the dune complex, and growth ofdepositional lobes within the complex.

4. The foredune–transgressive dune complex experienced a netpositive sediment budget, increased geomorphic diversity byway of new and expanded features within the complex, andrecovery of the foredune profile to a form very similar to thatpre-restoration following erosion at the crest from vegetationremoval and scarping at the dune toe from high water events.Should these responses continue, it is anticipated that dunemobility and active sand surface area within the study sitecould increase. However, given decadal scale stabilizationtrends and rapid regrowth of vegetation at the site, continuedvegetation removal will be required.

Acknowledgements—The authors gratefully acknowledge field andresearch assistance from Hawley Beaugrand, Danielle Bellefleur, BarryCampbell, Connie Chapman, Mike Collyer, Derek Heathfield, andNicholas von Wittgenstein. LiDAR and digital orthophotography wasprocessed by Hawley Beaugrand in the Hyperspectral LiDAR ResearchLaboratory in the Department of Geography, University of Victoria. TheLiDAR dataset was collected by Terra Remote Sensing, Sidney, BC,Canada by way of a cost-share agreement between University of Victoriaand PCA. Partial funding and logistical support was also provided by PCAand fieldwork and data analyses were supported by a Natural Sciencesand Engineering Research Council (NSERC) Discovery Grant to I.J.Walker, a Geological Society of America research grant and NSERCAlexander Graham Bell Canadian Graduate Scholarship to J. Eamer,and a MITACS Accelerate BC Graduate Internship, partially sponsoredby PCA, to I. Darke.

ReferencesAnderson JL, Walker IJ. 2006. Airflow and sand transport variationswithin a backshore-parabolic dune plain complex: NE GrahamIsland, British Columbia, Canada. Geomorphology 77(1–2): 17–34.

Andrews BD, Gares PA, Colby JD. 2002. Techniques for GIS modelingof coastal dunes. Geomorphology 48: 289–308.

Anthony EJ, Vanhee S, Ruz M. 2006. Short-term beach–dune sand bud-gets on the North Sea coast of France: sand supply from shoreface todunes, and the role of wind and fetch.Geomorphology 81: 316–329.

Anthony EJ, Vanhee S, Ruz M. 2007. An assessment of the impact of ex-perimental brushwood fences on foredune sand accumulation basedon digital elevation models. Ecological Engineering 31: 41–46.

Arens SM, Geelen LHWT. 2006. Dune landscape rejuvenation byintended destabilization in the Amsterdamwater supply dunes. Journalof Coastal Research 22: 1094–1107.

Arens SM, Slings Q, de Vries CN. 2004. Mobility of a remobilized par-abolic dune in Kennemerland, the Netherlands. Geomorphology 59:175–188.

Beaugrand HER. 2010. Beach–dune Morphodynamics and ClimateVariability Impacts on Wickaninnish Beach, Pacific Rim NationalPark Reserve, British Columbia, Canada. Unpublished MSc Thesis,University of Victoria, Canada.

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)

1159GEOMORPHIC RESPONSES TO INVASIVE VEGETATION REMOVAL

Chappell A, McTainsh G, Leys J, Strong C. 2003. Using geostatistics toelucidate temporal change in the spatial variation of aeolian sedi-ment transport. Earth Surface Processes and Landforms 28: 567–585.

Darke IB, Walker IJ, Eamer JBR. 2013. Geomorphical considerations fordynamic dune restoration: Pacific Rim National Park Reserve, BritishColombia, Canada. Earth Surface Processes and Landforms. DOI:10.1002/esp.3380

Davidson-Arnott RGD, MacQuarrie K, Aagaard T. 2005. The effect ofwind gusts, moisture content and fetch length on sand transport ona beach. Geomorphology 68: 115–129. DOI. 10.1002/esp.3380

Desmet PJJ. 1997. Effects of interpolation errors on the analysis ofDEMs. Earth Surface Processes and Landforms 22: 563–580.

Eamer JBR, Walker IJ. 2010. Quantifying sand storage capacity oflarge woody debris on beaches using LIDAR. Geomorphology118: 33–47.

Heathfield DK, Walker IJ. 2011. Analysis of coastal dune dynamics,shoreline position, and large woody debris at Wickaninnish Bay,Pacific Rim National Park, British Columbia. Canadian Journal ofEarth Sciences 48: 1185–1198.

Heathfield DK, Walker IJ, Atkinson DE. 2012. Erosive water levelregime and climatic variability forcing of beach–dune systems onsouthwestern Vancouver Island, British Columbia, Canada. EarthSurface Processes and Landforms. DOI: 10.1002/esp.3350

Heritage GL, Milan DJ, Large ARG, Fuller IC. 2009. Influence of surveystrategy and interpolation model on DEM quality. Geomorphology112: 334–344.

Hesp P. 2002. Foredunes and blowouts: initiation, geomorphology anddynamics. Geomorphology 48: 245–268.

Hilton M, Woodley D, Sweeney C, Konlechner T. 2009. The develop-ment of a prograded foredune barrier following Ammophila arenariaeradication, Doughboy Bay, Stewart Island. Journal of CoastalResearch SI59: 317–321.

Houser C, Hamilton S. 2009. Sensitivity of post-hurricane beach anddune recovery to event frequency. Earth Surface Processes andLandforms 34: 613–628.

Lane SN, Westaway RM, Hicks DM. 2003. Estimation of erosion anddeposition volumes in a large, gravel-bed, braided river using synopticremote sensing. Earth Surface Processes and Landforms 28: 249–271.

Mazzotti S, Jones C, Thomson RE. 2008. Relative and absolute sea levelrise in western Canada and northwestern United States from a com-bined tide gauge-GPS analysis. Journal of Geophysical Research113: 1–19.

Milan DJ, Heritage GL, Large ARG, Fuller IC. 2011. Filtering spatialerror from DEMs: implications for morphological change estimation.Geomorphology 125: 160–171.

Mitasova H, Overton M, Harmon RS. 2005. Geospatial analysis of acoastal sand dune field evolution: Jockey’s Ridge, North Carolina.Geomorphology 72: 204–221.

Copyright © 2013 John Wiley & Sons, Ltd.

Page N, Lilley P, Walker IJ, Vennesland RG. 2011. Status Report onCoastal Sand Ecosystems in British Columbia, report prepared forthe Coastal Sand Ecosystems Recovery Team. Coastal Sand Ecosys-tems Recovery Team: Vancouver, BC; vii + 83 pp.

Ribeiro JR, Diggle PJ. 2001. geoR: A package for geostatistical analysis.R-NEWS 1: No. 2 ISSN 1609-3631. http://can.r-project.org/doc/Rnews.

Riksen M, Ketner-Oostra R, Van Turnohout C, Nijssen M, Goossens D,Jungerious PD, Spaan W. 2006. Will we lose the last active inlanddrift sands of Western Europe? The origin and development of theinland drift-sand ecotype in the Netherlands. Landscape Ecology21: 431–447.

Sallenger Jr AH, Krabill WB, Swift RN, Brock J, List J, Hansen M,Holman RA, Manizade S, Sontag J, Meredith A, Morgan K, YunkelJK, Frederick EB, Stockdon H. 2003. Evaluation of airborne topo-graphic LIDAR for quantifying beach changes. Journal of CoastalResearch 19: 125–133.

Saye S, Vanderwal D, Pye K, Blott S. 2005. Beach–dune morphologicalrelationships and erosion/accretion: an investigation at five sites inEngland and Wales using LIDAR data. Geomorphology 72: 128–155.

Swales A. 2002. Geostatistical estimation of short-term changes inbeach morphology and sand budget. Journal of Coastal Research18: 338–351.

Walker IJ, Beaugrand HER. 2008. Coastal Geoindicators Monitoring-Protocol for Climate Change & Coastal Erosion in Canada’s PacificCoastal National Parks. Unpublished report submitted to PacificRim National Park Reserve. Pacific Rim National Park Reserve:Ucluelet, BC; 39 pp.

Walker IJ, Hesp PA, Davidson-Arnott RGD, Ollerhead J. 2006. Topo-graphic steering of alongshore airflow over a vegetated foredune:Greenwich Dunes, Prince Edward Island, Canada. Journal of CoastalResearch 22: 1278–1291.

Walker IJ, Sydneysmith R. 2008. British Columbia. In From Impacts toAdaptation: Canada in a Changing Climate 2007, Lemmen DS,Warren FJ, Lacroix J, Bush E (eds). Government of Canada: Ottawa;329–386.

Wheaton JM, Brasington J, Darby SE, Sear DA. 2010. Accounting foruncertainty in DEMs from repeat topographic surveys: improved sed-iment budgets. Earth Surface Processes and Landforms 35: 136–156.

Wiedemann AM, Pickart A. 1996. The Ammophila problem on thenorthwest coast of North America. Landscape and Urban Planning34: 287–299.

Woolard JW, Colby JD. 2002. Spatial characterization, resolution, andvolumetric change of coastal dunes using airborne LIDAR: CapeHatteras, North Carolina. Geomorphology 48: 269–287.

Zernetske PL, Hacker SD, Seabloom EW, Ruggiero P, Killian JR,Maddux TB, Cox D. 2012. Biophysical feedback mediates effects ofinvasive grasses on coastal dune shape. Ecology 93: 1439–1450.

Earth Surf. Process. Landforms, Vol. 38, 1148–1159 (2013)