Dynamic floodplain vegetation model development for the Kootenai River, USA

Dynamic floodplain vegetation model development for the Kootenai River, USA
of 13
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
  Author's personal copy Dynamic  fl oodplain vegetation model development for the Kootenai River, USA Rohan Benjankar a , * , Gregory Egger b , Klaus Jorde a , Peter Goodwin a , Nancy F. Glenn c a Center for Ecohydraulics Research, University of Idaho, 322 E. Front Street, Boise, ID 83702, United States b Umweltbuero Klagenfurt, Bahnhofstrasse 39, 9020 Klagenfurt, Austria c Department of Geosciences, Idaho State University, 322 E. Front St., Suite 240, Boise, ID 83702, United States a r t i c l e i n f o  Article history: Received 1 February 2011Received in revised form12 July 2011Accepted 19 July 2011Available online 17 August 2011 Keywords: Spatially distributedDynamic vegetation modelFloodplain physical processesError matrixHydrological alterationHydrodynamic model a b s t r a c t The Kootenai River  fl oodplain in Idaho, USA, is nearly disconnected from its main channel due to leveeconstruction and the operation of Libby Dam since 1972. The decreases in fl ood frequencyand magnitudecombined with the river modi fi cation have changed the physical processes and the dynamics of   fl ood-plain vegetation. This research describes the concept, methodologies and simulated results of the rule-based dynamic  fl oodplain vegetation model  “ CASiMiR-vegetation ”  that is used to simulate the effect of hydrological alteration on vegetation dynamics. The vegetation dynamics are simulated based on existingtheory but adapted to observed  fi eld data on the Kootenai River. The model simulates the changingvegetation patterns on an annual basis from an initial condition based on spatially distributed physicalparameters such as shear stress,  fl ood duration and height-over-base  fl ow level. The model was cali-brated and the robustness of the model was analyzed.The hydrodynamic (HD) models were used to simulate relevant physical processes representinghistoric, pre-dam, and post-dam conditions from different representative hydrographs. The generalconcept of the vegetation model is that a vegetation community will be recycled if the magnitude of a relevant physical parameter is greater than the threshold value for speci fi c vegetation; otherwise,succession will take place toward maturation stage. The overall accuracy and agreement Kappa betweensimulated and  fi eld observed maps were low considering individual vegetation types in both calibrationand validation areas. Overall accuracy (42% and 58%) and agreement between maps (0.18 and 0.27)increased notably when individual vegetation types were merged into vegetation phases in both cali-bration and validation areas, respectively. The area balance approach was used to analyze the proportionof area occupied by different vegetation phases in the simulated and observed map. The result showedthe impact of the river modi fi cation and hydrological alteration on the  fl oodplain vegetation. Thespatially distributed vegetation model developed in this study is a step forward in modeling riparianvegetation succession and can be used for operational loss assessment, and river and  fl oodplain resto-ration projects.   2011 Elsevier Ltd. All rights reserved. 1. Introduction Vegetation models have been developed to simulate vegetationdynamics considering factors such as competition, shade tolerance,and hydrologic variables including  fl ood frequency,  fl ood duration,and depth to ground water table to study the impact of an alteredhydrologic regime (e.g., Pearlstine et al.,1985). During recent years,a variety of ecological models have evolved to address changes invegetation species as consequences of modi fi cation in environ-mental variables and hydrological alterations (e.g., Braatne et al.,2007, 2002; Carmel et al., 2001). Different models weredeveloped for analytical and management tools to assess thepotential effects of human-induced disturbances upon wetlandsand vegetation (e.g., Baptist et al., 2004; Glenz, 2005; Poiani and Johnson, 1993). These vegetation models are developed based onthe concept that vegetation establishes on an elevation gradientand the  fl oodplain physical processes such as  fl ood frequency,inundation duration, shear stress, and velocity are driving forces tocreate suitable habitat. Most models are based on two majorconcepts: static equilibrium and dynamic-transient (Korzukhinet al., 1996). First, one assumes that the state of vegetation is instatic equilibrium with environmental conditions. Later, oneassumes that the vegetation is in a dynamic-transient stage withchangesinbioticandabioticconditions(MerrittandCooper,2000).A species- or community-based approach is typically used inpredictive vegetation modeling (Zimmermann and Kienast, 1999). *  Corresponding author. Tel.:  þ 1 208 2848472; fax:  þ 1 208 332 4425. E-mail address: (R. Benjankar). Contents lists available at ScienceDirect  Journal of Environmental Management journal homepage: 0301-4797/$  e  see front matter    2011 Elsevier Ltd. All rights reserved.doi:10.1016/j.jenvman.2011.07.017  Journal of Environmental Management 92 (2011) 3058 e 3070  Author's personal copy Vegetation models are developed based on ecological processesof riparian and wetland systems. Riparian ecosystems and theirdynamicsaremainlygoverned byhydrology, although disturbance,climatic condition, moisture condition, soil types, and nutrientsplay important roles. Riparian systems are transitional semi-terrestrial areas (Naiman et al., 2005) regularly in fl uenced byfresh water and usually extend from the edge of the water to theedge of the upland community. Periodic disturbances in riparianzonesprovidespatialandtemporalheterogeneity,lifehistorytraits,and the availability of regeneration habitat (Nakamura et al.,1997).Intensity, timing, duration, and frequency of disturbances governthe erosion and depositionprocesses that create open channel barsand islands where riparian vegetation colonize (Dykaar andWigington, 2000). Generally, riparian zones are classi fi ed basedon a variety of environmental characteristics that strongly in fl u-enceplantcommunitiessuchasdifferentrecurrenceinterval fl oods(Goodwin et al., 1997).Wetlands are also considered a component of riparian ecosys-tems, even though they differ signi fi cantly in spatial context,disturbance regime, hydrology, and ecology (Naiman et al., 2005).Wetlandsareinundatedorsaturatedbysurfaceorgroundwaterfora long duration of the year, and develop when inundation of wetland plants produce anaerobic processes and force rootedplants to adapt to  fl ooding (Keddy, 2000). The wetland representsboth aquatic and terrestrial environments; hence, they are dif  fi cultto delineate precisely (Keddy, 2000).Succession and retrogression, also termed as recycle, arecomplex and key processes in riparian ecosystems (Formann et al.,in preparation; Naiman et al., 2005). Recycle takes place due todifferent levels of disturbances (e.g.,  fl ood,  fi re, grazing, etc) and isin fl uenced by many factors (e.g., nutrient, soil, moisture, climate,etc). Succession is a systematic unidirectional process of vegetationchange, in which vegetation types replace each other sequentiallyuntil stable vegetation is reached ( Johnson and Miyanishi, 2007).Retrogression occurs when the effects of disturbances, also knownas environmental stress, destroy or reduce the community struc-ture or change species composition in reverse succession. One of the common disturbances in riparian vegetation dynamics is fl ooding on the river  fl oodplain, which is key for creating newhabitats and destroying existing vegetation by changing morpho-dynamics, erosion, and deposition (Bendix, 1998).Various ecological models have been developed to assess theecologicallosses attributed todams, levees,and rivermanagement,but mostof them have been based on transectorcross-section data(e.g., Braatne et al., 2002) and the static equilibrium concept (e.g.,Auble et al.,1994). The development of vegetation and topographycan take up to several decades to reach a state of dynamic equi-librium and thus the static equilibrium concept may not be a validassumption (cf. Formann et al., in preparation). Dynamic-transitionmodels based on temporal and spatial scales are important toolsand are often used to predict potential future vegetation, sinceecological systems are dynamic, changing over a variety of spatialand temporal scales due to disturbance.The current approach toanalyze loss of riparian ecosystems dueto humanin fl uences,suchasdamoperations isto measureriparianvegetation loss at a speci fi c location (transects) and time, and laterextrapolate to a larger area (e.g., Braatne et al., 2007; Jamieson andBraatne, 2001; Merritt and Cooper, 2000; Polzin and Rood, 2000).The method that considers spatial and temporal distribution of vegetation may be suitable to systems that are diverse and differsigni fi cantly in spatial and temporal scale. A dynamic vegetationmodel may be a proper tool to assess ecological losses and tomanage or restore a riparian ecosystem in the future.The primary objective of this study is to develop a rule-basedspatially distributed vegetation model to predict dynamic succes-sionandretrogressionoftheplantspeciesbasedprimarilyonscourdisturbance, shear stress as an indicator of mechanical stress and fl ood duration as an indicator of physiological stress. A dynamicrule-based spatially distributed vegetation model  “ CASiMiR-vege-tation ”  was developed in this research to simulate vegetationdynamics on the  fl oodplain based on simulated physical parame-ters, literature and  fi eld observed data.  “ CASiMiR  ”  stands forComputer Aided Simulation Model for Instream Flow Require-ments. This paper describes the structure, development and modeltest at a braided reach of the Kootenai River in the Northern Idaho(Fig.1), USA. 2. Study area The Kootenai Basin is an international watershed (41,910 km 2 )srcinating in southeastern British Columbia, Canada (Fig. 1). TheKootenaiRiverfromLibbyDamtoKootenaiLakecanbedividedinto Fig. 1.  Study site including calibration and the validation area at the braided reach, which is located just upstream of the town of Bonners Ferry, Idaho. R. Benjankar et al. / Journal of Environmental Management 92 (2011) 3058 e  3070  3059  Author's personal copy three geomorphic reaches, which include a canyon reach, a braidedreach, and a meandering reach. The Kootenai River has a meanannual  fl ow of 390 m 3 /s at the Leonia gage station. The canyonreach is characterized by alternate con fi ned and semi-con fi nedsections. The braided reach is located just upstream of the townof Bonners Ferry and considered to be a transitional sectionbetween the canyon and meandering reaches. The reach consists of a wide  fl oodplain and complex channels that are seasonally active.Bed surface materials in this reach are gravels and cobbles and thechannel slope is about 0.019%. The meandering reach is heavilyin fl uenced by the hydraulic backwater from Kootenay Lake. Koo-tenai Basin has been intensively managed since the late 1800s. Thehistory of basin management includes:  fl oodplain diking, drainage,channel dredging and construction of 130 m high Libby Dambetween 1968 and 1974 impacting hydrology, sediment transport,water quality, riparian  fl oodplain vegetation and aquatic species. 3. Methodology  The  “ CASiMiR-vegetation ”  model was developed in the ArcGISenvironment using Model Builder and is a grid-based (raster)model that simulates vegetation succession or retrogression inannual time steps within 10 m by 10 m grid cells (Benjankar, 2009;Politti, 2008). The model functions with a Boolean logic and relieson hard thresholds provided by users. The model uses differentphysical parameters as input (Fig. 2) i.e., shear stress,  fl ood dura-tion, height-over-base  fl ow level (HBFL), height-over-mean waterlevel (HML), topography, and model zones (Fig. 3), as describedbelow. The outputs of the model are different vegetation types(Table 1).The fundamental concept for the vegetation model develop-ment is the functional relationship between hydrology, physicalprocesses, riparian ecosystems, and vegetation types. The ideabehind the model development is that the distribution of thehabitat patches changes spatially over time due to the primarydrivers, particularly  fl ooding, erosion, and deposition of sedimentand regeneration of riparian vegetation. Generally, the  fl ooddisturbances (i.e., erosion and deposition) create a barren site forthe seedling colonization. The seedlings can be removed byerosion/deposition or by  fl ood stress. The shear stress and  fl oodduration is an indicator of erosion and  fl ood stress, respectively.HBFL is used as a surrogatefor erosion and deposition disturbances(hereaftertermedscourdisturbance)thatdestroyvegetationatlowlevel from mean base  fl ow stage (Mahoney and Rood,1998). Whilenot considered in this study, other physical parameters such as ice fl ow (Smith and Pearce, 2000),  fl ood frequency (Perona et al.,2009), depth of ground water table (Rood et al., 2003) and sedi- ment transport can be major driving forces for mortality of vegetation.Three different functional scenarios are developed for themodeling processes based on regulation of Kootenay lake levelsthrough operation of Corra Linn Dam and upstream dischargesfrom the Libby Dam operation. The  “ Historic ”  scenario (pre 1939)represents apristineconditioneventhoughcertaindegreesoflocalmodi fi cation (levees) were made to the landscape. The  “ Pre-dam ” period (1939 e 1972) incorporates all developments prior toconstruction of Libby Dam, including levees, and modi fi ed down-stream boundary conditions (effect of Kootenay Lake levels). The “ Post-dam ”  scenario (1973 e 2007) incorporates operations of LibbyDam to the pre-dam scenario.  3.1. Model variables Floodplain physical processes and topography are considereddriving forces in simulating the initial vegetation types in the startmodule, and dynamic module. These relevant physical processes Fig. 2.  Model structure. R. Benjankar et al. / Journal of Environmental Management 92 (2011) 3058 e  3070 3060  Author's personal copy were simulated with a combined one-dimensional (1D river) andtwo-dimensional (2D  fl oodplain) hydrodynamic model using riverhydrology, cross sections, and a digital elevation model (DEM) of the  fl oodplain (Benjankar, 2009). To simulate relevant physicalprocesses, all annual hydrographs at Libby Dam gage station from1911 to 2007 were classi fi ed into eight classes in each scenariobased on hydrologic conditions, i.e., dry, average and wet(Benjankar, 2009). For each class, a representative hydrograph wasdeveloped and all considered physical processes were simulatedwith this hydrograph.Landscape topography was represented in the form of a DEM inthe model. In the current study,1928 topography was used for thehistoric scenario (1911 e 1938), whereas 2005 topography was usedfor pre-dam and post-dam conditions (1939 e 2007). Light Detec-tion and Ranging (LiDAR) technology was used to survey existingtopography in 2005, while historic topography was generated bydigitization of 1928 topographic maps. HML is the distancebetween topography and mean water level. HML was simulatedfrom mean discharges during the vegetation growing period (Aprilto September of each year) for each scenario. HML was used topredict the speci fi c vegetation type based on elevation (Benjankar,2009). HBFL is the distance between topography and base  fl owlevelsintheriver.Theaverage fl owfromAugust15toSeptember15of each representative year was used as a late summer base  fl ow(Mahoney and Rood, 1998) to calculate the water stage using thehydraulic model MIKE11. Depth-averaged shear stress was simu-lated using the MIKEFLOOD HD model (DHI, 2004) consideringpeak  fl ow of representative hydrographs. Flood duration wascalculated using MIKE11GIS (DHI, 2005) as the number of days the fl oodplain was inundated during the vegetation growing period.The model areawas classi fi ed into three different zones (Fig. 3a)i.e., aquatic zone (AZ), bank zone (BZ) and  fl oodplain zone (FZ)based on the concept that magnitude and frequency of   fl oodinggoverns the presence or absence, and the structure of riparianvegetation. The AZ is typically covered with water the entire year.ThespatialextentofBZisbetweentheAZandbankfull fl ood(1.5 e 2yearrecurrenceinterval).ThespatialextentofFZisbetweenBZandapproximately 100 year recurrence interval  fl oodplain in historictopography. Modeling processes are classi fi ed into three differentmodules, including: start, dynamic and visualization modules asshown in Fig. 2.The vegetation model was developed using literature informa-tion and  fi eld observed data. The basic information about thenatural systemwas important in order to model historic vegetationand habitats in the Lower Kootenai River. Therefore, a detailedstudy of   fl oristical composition, habitat condition, succession Fig. 3.  a. Model zones, b. different recruitment bands based on HBFL, and c. conceptual model for recruitment sub-module. Upward and downward arrow direction indicate a newvegetation type after succession and retrogression, respectively. R. Benjankar et al. / Journal of Environmental Management 92 (2011) 3058 e  3070  3061  Author's personal copy velocity, and ecology of the riparian ecosystem was performed atthe adjacent natural river reaches such as Elk River, Flathead River,Upper Kootenai River  e  at Fenwick and Wasa, all similar to theKootenai River. A vegetation survey was conducted at the calibra-tion and validation areas in summer 2006. Fourteen vegetationcover types (Table 1) were identi fi ed based on the area coverage byspeci fi c vegetation type according to visual judgment. The vege-tation types were grouped into different stages i.e., colonization,transition andmaturephases(Table1).Primarysuccessionstartsinthe colonization stage when bare sites are available for seedlingcolonization. Aftera few years, the vegetation cover will rise due tohigh biomass production in the transition stage. The vegetation isdominated by competitive woody and long-lived species in themature stage. All the vegetation types or phases considered in thisstudy are composed of different herbaceous and woodland species(Table 1).  3.2. Model conceptualization and structure The time-for-space-hypothesis approach (cf. Johnson andMiyanishi, 2007) was adopted in developing the vegetationmodel. The model assumes that vegetation will either developfollowing succession towards a maturation stage, or it will beremoved if the magnitude of certain physical parameters is greaterthan the threshold value.  3.2.1. Start module The start module is a static module that predicts differentvegetation types (Table 1) based on HML, topographic elevation,and three different zones (Benjankar, 2009). The main purpose of this module is to predict Potential Natural Vegetation (PNV) to useas starting conditions for the dynamic module. PNV is a steadyplant community that is present in an area as the result of progressive succession without human in fl uences.  3.2.2. Dynamic module The dynamic module evaluates dynamics and the spatialdistribution of   fl oodplain vegetation based on initial vegetation(startcondition)orpreviousyearvegetation(Benjankar,2009),and fi ve physical parameters including topography, HBFL, differentzones,shearstress,and fl oodduration(Fig.2).Thedynamicmoduleincludes three succession series: cottonwood, reed, and wetland(Fig. 4). The dynamic module is divided into three sub-modulesbased on their functions. Recruitment sub-module.  The recruitment sub-module(Fig. 3c) de fi nes the areas suitable for cottonwood and willowshrub, reed and forbs, deep marsh, and the possibility of distur-bances through scouring based on in fl uences of hydraulic andmorphologic factors (Benjankar, 2009). The sub-module predictsvegetation type (outcome) based on different rules (Table 2). HBFL was used as an indicator of hydraulic factors, which act as refer-ence elevations to establish different bands (Fig. 3b). Band 1 is anarea having the possibility of scour disturbance, and Band 2 is thearea where cottonwood and willow seedling can survivesuccessfully (Fig. 3b). Similarly, Band 3 is an area where wetlandsuccession takes place. In certain circumstances, wetland vegeta-tion type (deep marsh), can follow succession towards pioneer  Table 1 List of different vegetation types considered in the study.Succession stages and phases Vegetation types Primary dominant plant species (common name) Plant speciesCommon name Scienti fi c name Colonization stage  Cattail species  Typha spp. Deep wetland phase Deep marsh (DM) c Open water dominant wetland Fowl blue grass  Poa palustris Initial phase Gravel and sand bar (GS) Open gravel or sand bar without vegetation Reed canary grass  Phalaris arundinacea Pioneer phase Pioneer vegetation (PV) Red top, Sand bar willow, Black cottonwood,Reed canary grassRed top  Agrostis stoloniferaTransition stage  Quack grass  Agropyron repens Shallow wetland phase Shallow marsh and wetmeadow (SMM) c Shallow marsh and wet meadow dominantwetland with Cattail speciesCommon tancySpotted knapweed Tanacetum vulgareCentaurea maculosa Herb phase Reed and forbs (RF) d Reed canary grass, Red top, Quack grass Kentucky blue  Poa pratensis Shrub phase Reed-shrub a,d Reed dominant shrub vegetation (50-75% reed) Yellow willow  Salix lutea Cottonwood and willowshrub (CWS) e Sandbar willow, Black cottonwood,Drummond willow, Yellow willow,Reed canary grass, Red topDrummond willow  Salix drummondiana Sandbar willow  Salix exigua Shrub-reed a,e Shrub-dominant reed vegetation (50-75% shrub) Subalpine  fi r  Abies lasiocarpa Successional wetland phase Wet forbs and shrubs (WFS) b,c Wet forbs and shrubs dominant wetland aftersuccession from shallow marsh and wet meadowBlack cottonwood  Populus trichocarpa Successional reed phase Reed, forbs and shrub (RFS) b,d Reed canary grass, Red top, Commontansy, Quack grass, Spotted knapweed, Kentuckyblue grass, forbs with other shrubsGreat plains cottonwood  Populus deltoides Early successionalwoodland phaseYoung cottonwood forest (YCF) e Black cottonwood, Great plains cottonwood,Common snowberry, Red osier dogwood,Red top, Quack grass, Reed canary grassCommon snowberry  Symphoricarpos albus Established forest phase Old cottonwood forest (OCF) e Black cottonwood, Spruce, Red top,Fowl blue grass, Quack grass, Fowl meadow grassRed osier dogwood Spruce  Cornus sericea Picea spp. Beaver grassland a,e Old cottonwood habitat, but beaver affected Western red cedar  Thuja plicataMature stage  Paper birch  Betula papyrifera Mature mixed forest phase Mature mixed forest (MMF) e Western red cedar, Black cottonwood,Great plains cottonwood, Subalpine  fi r,Douglas  fi r, Paper birch, Pine species,Common snowberry, Quack grass,Kentucky blue grass, Red topDouglas  fi r  Pseudotsuga menziesii Pine species  Pinus contorta a Mapped in the  fi eld, absent in the simulation. b Considered in simulation, but not mapped in the  fi eld. c Wetland series. d Reed series. e Cottonwood series. R. Benjankar et al. / Journal of Environmental Management 92 (2011) 3058 e  3070 3062
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!