Development of a GIS model to enhance macrophyte re-establishment projects

Applied Geography 32 (2012) 629e635

Contents lists available at ScienceDirect

Applied Geography journal homepage: www.elsevier.com/locate/apgeog

Development of a GIS model to enhance macrophyte re-establishment projects Jonathan P. Fleming a, *, John D. Madsen b, Eric D. Dibble a a b

Department of Wildlife, Fisheries and Aquaculture, Mississippi State University, PO Box 9690, Mississippi State, MS 39762, USA Geosystems Research Institute, Mississippi State University, PO Box 9627, Mississippi State, MS 39762, USA

a b s t r a c t Keywords: GIS Suitability modeling Aquatic plants Habitat

Diverse native aquatic macrophytes serve a number of physical and biological functions in the aquatic environment and provide essential habitat for several fish species. In systems that lack submersed macrophytes, native macrophyte re-establishment can be used to revitalize the aquatic community. Planning re-establishment projects requires knowledge of the system along with the growth requirements of macrophytes. Prior studies have identified factors that are important for macrophyte colonization, persistence, and dispersal. However, deductive approaches to identify macrophyte habitat that is suitable for management application have not been developed. A potential solution to this problem is the incorporation of waterscape-wide variables into a Geographic Information System (GIS) and the use of spatial modeling techniques to identify suitable macrophyte habitat. This provides a scientifically based approach to macrophyte re-establishment planning to make efforts more efficient and to recognize potential coverage. The flexibility, scalability, and topological advantages of using a GIS to identify and visualize habitat allow integration with other spatial ecological variables to improve the management of aquatic resources from plants to fish, including invasive species mitigation. Using Little Bear Creek Reservoir, Alabama, as an example system, we illustrate a GIS modeling process that can be applied to any system where the identification of macrophyte habitat is relevant to aquatic resource management goals. Ó 2011 Elsevier Ltd. All rights reserved.

Introduction Diverse native aquatic macrophyte communities serve a number of physical and biological functions in the aquatic environment (Carpenter & Lodge, 1986). In some freshwater ecosystems native macrophyte communities are absent even when trophic status and light availability are conducive to rooted submersed macrophyte growth, especially in young systems that may not have a macrophyte seed bank or are too distant for dispersal of native species (Smart, Dick & Snow, 2005; Smart, Doyle, Madsen, & Dick, 1996). Native aquatic macrophyte establishment is a tool that may be used to enhance littoral freshwater communities (Smiley & Dibble, 2006). In addition, re-establishing plants has numerous advantages such as improving water quality (Scheffer, 1998), providing essential habitat and relatively safe foraging areas for juvenile fishes (Dibble, Killgore, & Harrel, 1996) and refugia for phytoplanktivores (Diehl & Kornijow, 1998). Re-establishing native submersed aquatic plant communities for fish habitat is of prime interest to fisheries managers (Miranda et al., 2010) but quantified research on the propagation and establishment of these * Corresponding author. Tel.: þ1 662 325 0954. E-mail address: jfl[email protected] (J.P. Fleming). 0143-6228/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.apgeog.2011.07.013

communities is limited. Studies regarding the restoration of submersed aquatic plant communities generally refer to the natural re-colonization of macrophytes following some environmental manipulation resulting in plant growth (Sondergaard et al., 2007), not specifically human mediated propagation. However, there have been a number of trials dealing specifically with native plant reestablishment, and techniques for propagation and establishment have been suggested in recent years (Smart, Dick & Doyle, 1998; Smart et al., 2005). One important consideration that is currently a hindrance to re-establishment is the selection of suitable sites to focus planting efforts (Grodowitz, Smart, Dick, Stokes, & Snow, 2009). Macrophyte re-establishment can be an expensive and labor intensive endeavor and identifying sites that have a potentially higher probability of supporting macrophyte growth provides an advantage for managers (Smart et al., 1996). In addition, locating suitable habitat conditions has been highlighted as one of the most important considerations when planning ecological reintroduction projects (Armstrong & Seddon, 2007). Factors affecting macrophyte colonization and growth have received considerable attention in the literature. It is commonly noted that the most limiting factor for freshwater macrophytes is light availability which directly impacts their maximum colonization depth (Case & Madsen, 2004; Van Duin et al., 2001; Wersal,

630

J.P. Fleming et al. / Applied Geography 32 (2012) 629e635

Madsen, McMillan, & Gerard, 2006). Light is important to all aquatic organisms because it is the energy source that fuels primary production and consequently all trophic levels in freshwater aquatic ecosystems (Scheffer, 1998). Sediment texture also is important because it directly relates to benthic stability as well as biogeochemical variables important to macrophytes such as nutrient availability and presence of organic materials (Barko & Smart, 1986). Temperature is important in life-cycles of macrophytes, but water column profile and seasonal variability as well as latitude may be of greater importance than its areal distribution in lentic systems (Barko, Adams, & Clesceri, 1986; Madsen & Adams, 1989). Other physical factors that may significantly contribute to presence of macrophytes are fetch and littoral slope (Bini & Thomaz, 2005; Doyle, 2001; Duarte & Kalff, 1986; Koch, 2001). Water level fluctuations in reservoirs also create hydrologic conditions that can enhance or reduce macrophyte growth (Johnson, Allen, & Havens, 2007) depending on temporal periodicity and variability. Considered together, these variables will likely produce the most powerful predictive models. Geographic information system (GIS) technology provides a useful tool for producing spatially explicit models of macrophyte habitat suitability. This technology provides a coherent set of tools and procedures to relate spatial data with elaborate habitat models (Lehmann, Jaquet, & Lachavanne, 1997). Although GIS has traditionally been used more in the terrestrial realm, it has a great deal of applicability in aquatic systems as well (e.g., Touchart & Bouny, 2008). Spatial technology is now prolific in universities and management agencies, presenting a unique opportunity for researchers to apply the current state of knowledge regarding the fundamental niche of macrophytes to the development of spatially explicit tools that can actually be applied by management personnel to enhance re-establishment efforts. Most studies related to using GIS in aquatic vegetation science thus far have applied an inductive approach by explaining macrophyte distributions based on empirical data specific to a particular study area (e.g., Lehmann, 1998; Narumalani, Jensen, Althausen, Burkhalter, & Makey, 1997; Rea, Karapatakis, Guy, Pinder, & Mackey, 1998; Remillard & Welch, 1993). While this approach is useful for identifying spatially explicit factors that are important to macrophyte distributions, they rarely produce predictive models or processes that are general enough to be applied elsewhere (but see Narumalani, Jensen, Althausen, Burkhalter, & Makey, 1997 and Lehmann, 1998). This makes their management application impractical. To maximize effectiveness of aquatic plant re-establishment in areas devoid of vegetation, predictive models of suitable locations should be investigated. Data regarding the important variables for macrophyte occurrence and distributions should be the basis for these predictive models. Incorporating these data into GIS provides a framework for organizing spatial data and modeling ecological processes and has the potential to be a powerful tool for enhancing macrophyte re-establishment efforts. The deductive nature of this type of tool makes validation difficult but when based on scientific information it provides a starting point for science-based management. We acknowledge that more and continuing research is needed to identify important factors for macrophyte growth and especially validation of predictive management models. The objectives of this paper are to synthesize information from literature related to the prediction of macrophyte occurrence in lentic littoral habitats, collect appropriate data, and create a GIS model, based on this information, to identify and locate suitable areas for native aquatic plant re-establishment in Little Bear Creek Reservoir, Alabama. This is accomplished mainly through development of a conceptual spatial model and a methodological

framework applied with GIS. Although not tested here, our general working hypothesis is that native aquatic macrophytes will have greater survival rates if areas conducive to plant growth are selected for re-establishment sites which will save costs and make efforts more efficient and management more effective. Methods Study area description Little Bear Creek Reservoir (LBCR) is located in Franklin County in Northwest Alabama. The reservoir was impounded in 1975 for flood control and has since become an important driving force for the local economy by providing opportunities for fishing, camping, boating, and other outdoor recreational activities. It has a total impounded surface area (at full pool) of approximately 650 ha, extends 13 km upstream from the dam and has a fluctuating water level of approximately 3.6 m each year with full pool occurring from mid-April until late October (TVA, 2009). After impoundment, LBCR was reported to have contained submersed (e.g., Potamogeton spp.) and emergent (e.g., waterwillow Justicia americana (L.) Vahl) aquatic plant species (P. Cooper & G.D. Fleming, Bear Creek Millennium Project, personal communication). However, in recent years, there has been no sign of submersed aquatic plant species, with the exception of muskgrass (Chara spp.) a non-vascular macro-alga. Additionally, there are local reports of a reduction in emergent plant assemblages. Since 2007, macrophyte re-establishment trials have taken place in the reservoir in TVA approved pre-selected areas, but other potential suitable areas for growth have not been identified (Fleming, Madsen, & Dibble, 2011). Parameter selection, data acquisition, and calculations Defining and justifying the necessary variables are an essential first step in the development of a generalized predictive GIS model. Literature regarding factors important for macrophyte occurrence was collected and summarized and the following parameters and methods describe the most commonly identified factors for determining the occurrence and distribution of aquatic macrophytes, the reasons these factors are important, and methods used to collect and apply data to the GIS model. All map algebra calculations were performed using raster math geoprocessing with ArcGIS 9.3.1 Spatial Analyst and Model Builder (ESRI, 2009). Light Availability of light is used to define littoral zones in freshwater lentic systems because it provides the energy source that fuels photosynthesis. For this reason, light has been noted as the most limiting factor to macrophyte occurrence (Barko et al., 1986). Photosynthetically active radiation (PAR) was collected using a cosine corrected Li-Cor LI-1000 light meter with surface and underwater quantum sensors at 50-cm increments in the water column in the middle of the reservoir in June 2008 and from June to September 2009. A commonly used metric for assessing light availability for submersed macrophytes that was used for this model is the light extinction coefficient (Kd) which is calculated as:


    Kd ¼ ln LZ1  ln LZ2 ðZ2  Z1 Þ where, LZn is the light intensity at depth interval n, and Zn is the depth (in meters) of depth interval n (Madsen et al., 1999). Coefficients are generated for each depth interval except for the final measurement. The light extinction coefficient for a particular day was estimated by

J.P. Fleming et al. / Applied Geography 32 (2012) 629e635

calculating the arithmetic average of Kd. Furthermore, the seasonal arithmetic average extinction coefficient (Kds) was generated by averaging the mean coefficient for each sampling date (Table 1). Light extinction coefficients have been used in several models as an explanatory variable for macrophyte distribution or growth. However, the coefficient itself is only useful for suitability selection when incorporated with bathymetry (see below). The extinction coefficient is the most useful (as compared to turbidity or Secchi disk depth) because it takes into account attenuation from particulate and dissolved material whereas turbidity only measures particulate material and Secchi disk is subject to visual errors from the observer (Bini & Thomaz, 2005). We note, however, that if PAR measurements are unattainable then the use of a Secchi disk may be a suitable alternative. Bathymetry Light availability measurements are only useful in this model when incorporated with depth. Prior to the start of this study no bathymetric data existed for LBCR. In June 2008 a pointeintercept survey (Madsen, 1999) was performed with a 100 m separated point grid (Fleming, 2010). Each point was precisely located using a Panasonic Toughbook equipped with a Trimble AgGPS 106 GPS receiver and Farmworks Site Mate geospatial software. At each point, depth was recorded to the nearest meter. Data points were exported as a shapefile and added to ArcMap GIS software (ESRI, 2009). Elevation adjustments were made to convert depth measurements to mean sea level values and the points were interpolated into a 10  10 m continuous bathymetric surface (ArcGIS Spatial Analyst: Topo To Raster; Fleming, 2010). Topo To Raster is specifically designed to generate hydrologically correct elevation surfaces and if needed allows finer control of input variables. Maximum colonization depth (Zc; m) was estimated from the mean seasonal light extinction coefficient (Kds) using the Lambert-Beer equation:

ZC ¼ ln ðIZ =I0 Þ=Kds where, Iz/I0 is the percentage of light required by the species under consideration or the percentage of light at the maximum depth distribution of the plants. Vant, Davies-Colley, Clayton, and Coffey (1986) found that the value of ln (Iz/I0) could be set at a constant coefficient equal to 4.34 to explain most (93%) of the variability of maximum colonization in North Island, New Zealand lakes of differing clarity. This location has similar latitude, albeit in the southern hemisphere, as northern Alabama where LBCR is located. Middelboe and Markager (1997) indicated that maximum colonization depth for a species should be constant in lakes at the same latitude if the true maximum colonization depth is determined exclusively by light availability. Therefore, Zc is estimated by using the constant coefficient (4.34) reported by Vant et al. (1986). Table 1 Mean light extinction coefficients (Kd), maximum colonization depth (Zc), and Secchi disk depth (Zsd) for each sampling date in Little Bear Creek Reservoir, Alabama in 2008 and 2009. Date

Kd

Zc (m1)

Zsd (m1)

6/5/2008 6/12/2008 5/18/2009 6/2/2009 6/15/2009 6/29/2009 7/27/2009 8/22/2009 9/7/2009 Mean

* 0.935 0.843 0.734 0.931 0.922 0.952 0.580 0.552 0.806

* 4.64 5.15 5.91 4.66 4.71 4.56 7.48 7.86 5.62

2.18 2.29 2.20 2.07 1.43 2.12 2.25 2.05 * 2.07

*Data not available.

631

Mean maximum colonization depth was used as the depth barrier in the GIS model by first subtracting Zc from MSL of LBCR. The GIS layer for Zc used in the model is binary because plants either can or cannot survive at a given depth based on light availability. Therefore, the bathymetry raster was queried using raster math (Spatial Analyst: Greater Than Equal logical math geoprocessing operator). This operation returns a value of 1 for pixels where the statement is true and 0 for pixels where the statement is false. The resulting layer (Fig. 1A) indicates areas suitable for macrophyte growth based on light and depth. It should be noted that the bathymetry layer could first be queried and corrected if the water level was not at full pool for the time period of interest. This also allows for future simulations of environmental change. The model presented here, however, uses the full pool MSL value because it corresponds to the growing season of macrophytes at this geographic location. Slope Duarte and Kalff (1986) found significant differences between maximum submerged macrophyte biomass at steep and gentle slopes with greater biomass occurring at gentle slopes. This difference indicates that slope may be a significant factor not only in submersed plant biomass but also for optimal growth. Rea et al. (1998) found that slope was a significant factor in some years but not others indicating collinearity with other, possibly unmeasured, variables. Others have recommended aquatic macrophyte reestablishment should only be performed at locations with gentle slopes (e.g. Smart et al., 2005). Quantitative estimates of suitable slope have been reported (Duarte & Kalff, 1986), however, Rea et al. (1998) only classified slope into steep and less steep. Using GIS, slope estimates can be made rather easily from the bathymetric data generated in the previous step. Percentage slope was calculated from the bathymetric data layer with ArcGIS (Fleming, 2010). The slope layer was classified into two classes using a geometric interval classification method so that values represent relative suitability as opposed to pre-determined ranges. The geometric interval algorithm was designed specifically for continuous data. It minimizes variance within classes and works well with data that may be non-normal (ESRI, 2009). The slope layer was then queried with geoprocessing tools (as described above). Unlike bathymetry, it was necessary to retain the ability to rank slope values to incorporate into the model because most studies have indicated that less-steep slopes are generally more conducive to plant growth than steeper slopes. Because zero percentage slope is considered an acceptable value the Set Null conditional operator was used to create a raster with the original slope values for each pixel within the suitable range and sets all non-suitable pixels to a null value instead of zero. To incorporate it into the model the data were standardized to a scale of 0.0 to 1.0 using the formula:

Vi ¼ 1  ½ðSi  Smin Þ=ðSmax  Smin Þ where, Vi is the standardized value for the original slope percentage Si at a particular pixel, Smin is the least original slope percentage, and Smax is the greatest original slope percentage (modified from Chang, 2010), shown in Fig. 1B. Fetch Effective fetch length is the weighted distance around a wind direction that wind can travel over the water surface without obstruction from land (Lehmann, 1998). This is important in aquatic plant models because the longer the length the greater the probability of generating larger waves that can dislodge plants or increase suspended sediments which Doyle (2001) demonstrated as a significant factor limiting total plant mass accumulation. The

632

J.P. Fleming et al. / Applied Geography 32 (2012) 629e635

USGS developed a model (python script) that runs in ArcGIS to calculate effective fetch length (with no correction for wind speed) for a water body. The model requires a binary raster of the water body (water body ¼ 1, land ¼ 0) and an angular degree measurement of wind direction (direction of origin). The existing vector polygon of LBCR was converted to a 10  10 m pixel raster grid for use as the input. Wind direction was estimated from weather data collected from a weather station in Belgreen, Alabama (approximately 8 km from LBCR). Wind direction has the potential to change throughout the year but most wind direction is from due west (270 ), especially during spring colonization periods. Wind direction is user defined and may be adjusted for simulation. The effective fetch model calculates distances to land for nine radials spaced every three degrees on each side of the input degree of wind direction. Each measurement (n ¼ 9) is then weighted by the cosine of the angle deviation. Effective fetch for each pixel is calculated in the USGS model with the equation:

Lf ¼

X

X xi *cos gi = cos gi

where, Lf is the effective fetch, xi is the distance to land for a given angle, and gi is the deviation angle. A suitable value for effective fetch is rarely reported in the literature even though it is often used

as an explanatory factor for macrophyte occurrence. Because there is no suitable evidence to justify selecting prime areas based on a set range of values, defining this as a binary layer would be subjective. Therefore, to incorporate it into the model, data were standardized to a scale of 0.0 to 1.0 using the same linear transformation equation used above for slope, shown in Fig. 1C. Sediment characteristics Several studies have indicated that sediment texture and organic matter content can impact the distribution of submersed aquatic macrophytes (Barko & Smart, 1983, 1986; Barko et al., 1986; Koch, 2001; Lehmann et al., 1997; Madsen et al., 2006). General conclusions have been that fine sediment texture increases macrophyte productivity by increasing nutrient availability and cation exchange capacity whereas increasing organic matter decreases occurrence because of unfavorable chemical accumulation from decomposition (Koch, 2001). Several studies have noted differences in growth when different sediment types were analyzed in the laboratory (Madsen et al., 1996). Given its relative importance in the distribution of macrophytes, sediment cores were collected in LBCR for the analysis of texture and organic matter content to be included in the habitat suitability model. Although it is included here, sediment analysis is often tedious and

Fig. 1. Maps of Little Bear Creek Reservoir depicting results for different model parameters estimated from data collected in 2008. (A) Bathymetry with suitable depth in solid black. (B) Suitable standardized slope with darker values indicating more suitable areas. (C) Standardized fetch with darker values indicating larger fetch lengths. (D) Final suitability index with darker values indicating more suitable areas.

J.P. Fleming et al. / Applied Geography 32 (2012) 629e635

unfeasible. Therefore, we believe that this model can be applied without incorporating sediment data for general management purposes although this would be goal and site-specific. Texture. After surveying the depth of LBCR with a 100 m point grid in 2008, points less than 2 Secchi disk depth were selected for the sediment survey. This method of selection is sometimes used as a rough estimate of maximum macrophyte colonization depth and to define the littoral zone of a lentic system. Surveying all points of the 100 m grid (N ¼ 566) was not feasible and unnecessary considering that light is the main limiting factor on colonization depth (Barko et al., 1986). The points that were left (N ¼ 213) were defined as sample sites in the littoral zone and surveyed in July 2008 (Fleming, 2010). Using a modified 5 cm diameter by 12 cm length tip biomass sampler made of PVC pipe (after Madsen et al., 2007), each point was located with a HP Ipaq pocket PC equipped with a Holux GPS ultra receiver (GR-271) and Farmworks Site Mate geospatial software and sediment cores were collected and stored in whirl-packs. Texture was measured as the particle size distribution of sand, silt and clay in each sample using the Bouyoucos hydrometer method (Ashworth, Keyes, Kirk, & Lessard, 2001; Bouyoucos, 1962). However, point interpolation algorithms do not accept multivariate inputs. Therefore, to get accurate sediment surface predictions for the littoral zone of LBCR each particle class was interpolated separately using the ArcGIS Spatial Analyst kriging interpolator. Verification for the use of various interpolation methods for hydric sediments was sought out but materials that were relevant for this model were not found. A likely reason for this is the limited effort that has been put forth in mapping sediment in shallow-water habitats (Demas, Rabenhorst, & Stevenson, 1996). However, Valley, Drake, and Anderson. (2005) reviewed interpolation techniques for mapping submersed vegetation and found that kriging provided the best results. Therefore, we decided that if macrophyte distribution is correlated with sediment (as is an assumption of this model) then sediment interpolation surfaces should resemble those of macrophyte distribution surfaces. Depending on the study area, other interpolation techniques (i.e., Inverse Distance Weighting or Spline) may be adequate or more suitable. The interpolation resulted in three raster layers (sand, silt, and clay) containing the interpolated percentages. The pixels in each layer were normalized to 100% by dividing by the total combination of the three layers using raster math. Texture preferences for freshwater macrophytes vary in the literature, but Koch (2001) used the percentage of silt plus clay to define fine sediments for classifying macrophyte preferences. For the GIS model, suitable fine sediments were defined as such if the silt plus clay content were greater than or equal to 50%. The normalized silt layer and the normalized clay layer were combined and queried using raster math, creating a binary raster with preferable sediment texture (Fleming, 2010). There were no assumptions made that an increase from 50 to 100% would increase preference so no standardization was performed on suitable fine sediment values. Points marked as bedrock, gravel, or cobble were considered unsuitable for macrophyte colonization. These points were queried in ArcMap and converted to a 100  100 m pixel binary raster and used as a constraint in the final model. Organic matter content. The percentage of organic matter content generally has a negative correlation to macrophyte growth (Barko & Smart, 1983, 1986) although the limiting mechanism is not well understood (Koch, 2001). It is likely that this is due to nutrient limitations or the generation of unfavorable chemicals through decomposition (Barko & Smart, 1986). However, it is hypothesized that oxygen released in the rhizosphere may buffer these effects on

633

plants but at the same time facilitating the need for greater light availability for the production of this oxygen through photosynthetic processes (Koch, 2001). Organic matter content was analyzed using the same sediment samples that were used for texture analysis. Organic matter content was determined using standard protocols for the combustion (also called loss on ignition) method (Jones, 2001; Storer, 1984). Percentage organic matter was calculated by the difference in preand post-combustion sediment weights and was interpolated using the kriging method in ArcGIS Spatial Analyst (same as texture, described above). Organic matter preferences are reported to be below 5% (Barko & Smart, 1983). Therefore, raster math was used to create a binary raster with preferable organic matter content (Fleming, 2010). There were no assumptions made that a decrease in value from 5 to 0% would increase suitability so no standardization was performed on suitable organic matter content values. GIS model development All suitable layers were combined to create an overall suitability map. If all layers were binary and each had equal value then multiplying the raster datasets together would result in a binary suitability raster dataset where pixels with a value of 1 are suitable and areas with a value of 0 are not suitable. However, in reality some variables are more important than others and there may be more preferable ranges within variables that should be included. Of the layers that were generated; depth, texture, bedrock (consisting of bedrock, gravel or cobble presence) and organic matter content were binary. Fetch and suitable slope was standardized to a value between 0 and 1. As already mentioned, light is considered the most important factor in determining macrophyte distribution because if light is not suitable then macrophytes simply cannot survive. Of the substrate types, bedrock was considered completely unsuitable. However, among the other variables it would be subjective to rank them in importance based on the data currently available. Therefore, an index of suitability was calculated using all of the variables except depth and bedrock. The other four layers were weighted and summed using a weighted linear combination method to create an index between 0 and 1 using the formula:

I ¼

X

X wi xi = wi

where, I is the index value, wi is the weight for layer i, and xi is the standardized (or binary) value of criterion i for all layers i ¼ 1en (modified for these data from Malczewski (2000). All layers were given the equal weight of 0.25. This index does not include depth or bedrock because even if these layers were given a greater weighted value, all other criteria summed together could still be greater than 0 which in reality should not occur. Therefore, to insure that no depths greater than the suitable range and no areas with bedrock as the substratum were selected the binary raster pixel values of 0 for both layers individually were set to null and all pixel values of 1 were replaced with the composite index layer pixel values to yield a final indexed suitability layer (Fig. 1D). Results and discussion The purpose for developing this GIS model was to have a tool that used scientifically based information to select suitable sites to focus aquatic macrophyte re-establishment efforts. In LBCR the mean maximum colonization depth (Zc) was 5.2 m. This covered 293 ha which constitutes approximately 50% of the reservoir and is a good estimate of littoral zone area, which can be used in comparing lakes of various morphologies. When other variables

634

J.P. Fleming et al. / Applied Geography 32 (2012) 629e635

Table 2 Results of macrophyte growth GIS suitability model for Little Bear Creek Reservoir, Alabama from data collected in 2008. Parameters

Suitability criteria Original pixel type New pixel type Layer type Cell size (m2) Original range Suitable range Suitable area (h) Percent area Index weight Total suitable area Percentage area a

Maximum colonization depth (m)

Slope (%)

Fetch (m)

Sediment particle size (%)

Organic matter (%)

Substrate


Std. (0e1)a Floating point Floating point Continuous 10 0e98 0e16.2 273 46.5 0.25

Std. (0e1) Integer Floating point Continuous 10 10e1916 m NA NA NA 0.25

>50% Clay þ Silt Floating point Floating point Binary (within range) 10 4e94.0% Clay þ Silt >60% Clay þ Silt 383 65.2 0.25

<5% Floating point Floating point Binary (within range) 10 1.3e29.1% <5% 466 79.3 0.25

Bedrock, gravel, cobble String Integer Binary (discrete) 100 N ¼ 72 Absent NA 1.2 NA

First classified into two classes using a geometric interval algorithm. The lesser class was selected and classified.

were considered, the total suitable area for macrophyte growth was approximately 124 ha, 21% of the reservoir. This information is valuable for fisheries managers because the total potential coverage of macrophytes gives insight to the effect re-establishment could have on the overall productivity of the fishery. The specific results from each process of the model for Little Bear Creek Reservoir are presented in Table 2 but could be adjusted to any water body where the model is applied. In the past, only subjective methods have been used in site selection and many have been based on convenience rather than environmental suitability for macrophytes (e.g., Grodowitz et al., 2009). Best, Teeter, Landwehr, James, & Nair (2008) proposed a combination of models to maximize submerged macrophyte restoration success. Their approach involved complex hydrodynamic and sediment transport models and predicts macrophyte growth using carbon flow models. In addition, they predicted biomass of two species (Vallisneria americana and Stuckenia pectinata) with the primary goal of deciding which species had a greater likelihood of survival given a set of environmental conditions. Although these approaches may be used to generate hypotheses (e.g., van Nes, Scheffer, van den Berg, & Coops, 2003), they are not practical for management application because they require high amounts of location-specific fieldwork and data collection. In addition, rarely do these models use technology that is readily available for management or easily understood without specialized training. Finally, although data input for these models is often spatially explicit (e.g., van Nes et al., 2003), there is rarely an output provided that allows managers to assess the spatial distribution of the results. Geographic Information System technology provides a solution to many of these problems. The application of GIS to the field of aquatic botany has been warranted for over a decade (Lachavanne et al., 1997). However, GIS modeling approaches for macrophytes thus far have relied on inductive, site-specific applications. Similar to models discussed earlier, these approaches are suitable for generating hypotheses and determining site-specific environmental influences on aquatic plant assemblages but lack the necessary next step for management application (i.e., the creation of a deductive model based on scientific data that can be widely applied by managers with minimal tweaks). The model presented here is a deductive approach using technology and methodology that can be easily used to help make management decisions. Each process was performed using technology that is readily available to management personnel. Given the flexible nature of GIS software, this model also allows simulations of different environmental conditions with only minimal adjustments or data manipulations. The model proposed by Grodowitz et al. (2009) has the potential to be applied by managers

but the GIS model presented here should be the first step in deciding the preliminary locations necessary for their model. Future adjustments could be made with the GIS model to simplify data collection and processing even further. One consideration would be to develop a different way for calculating maximum colonization depth. The Secchi disk is sometimes used to define the photic zone of a lentic system and studies have shown a negative correlation between Secchi disk depth and the light extinction coefficient in moderately productive waters. Using a Secchi disk depth would eliminate the need for the use of a PAR meter which may not be available. Sediment analysis is the most time-consuming process of data collection for the model input. One solution for this problem is to identify soils by sight or texture in the field. Data collectors trained in soil analysis might be able to simply take sediment cores in the field and record the texture. This also might be useful in areas where the within lake sediment has high variability and therefore some sediments may be easily ranked for their macrophyte suitability. Several studies have used aerial photography to digitize soils (e.g., Lehmann et al., 1997; Narumalani et al., 1997) but this requires accurate photography of a site before impoundment. Another issue with using historic aerial photography is that soils may change because of erosion, sedimentation or anoxic conditions depending on inundation time (Wetzel, 2001) which could compromise information derived from pre-impoundment data. Sediment interpolation also is questionable. Kriging is the most complicated statistical operation in the model. It is likely that different geographic locations may have sediments with spatial distributions more or less feasible for the use of this interpolation method. However, this is a consideration that needs to be made on a location by location basis and cannot be automated by the model. Conclusion As with all models, this one represents a process and not a solution. Validation of the model is limited because it is used to predict suitable areas for macrophyte re-establishment efforts based on data obtained in other studies. Furthermore, the study area it has been applied to is lacking in aquatic plant communities and until extensive re-establishment projects are carried out, rigorous validation will be lacking. As an initial positive note, Fleming et al. (2011) performed a small re-establishment study in LBCR and found that American pondweed could survive for multiple years in this reservoir. The model described in this paper identified areas where this re-establishment project was conducted as suitable, which gives some insight into its predictive ability. This model represents an important step in the management

J.P. Fleming et al. / Applied Geography 32 (2012) 629e635

application of spatially explicit predictive models for aquatic systems. At the very least, the model provides a framework for the development of future macrophyte prediction models and can still be used for generating hypotheses. Aquatic resource managers can use this model as a tool to make projects for enhancing fish habitat with macrophytes more efficient and successful. Acknowledgments This research was supported by the Alabama Department of Conservation and Natural Resources, and the Little Bear Creek Millennium Group. We thank Gary Don Fleming, Cindy Fleming, B.J. Horton, Kyle Mangum, Matthew Spickard, Thad Huenemann, Eric Whittner, Elizabeth Flanary, and the Bear Creek Development Authority for field and technical assistance. References Armstrong, D. P., & Seddon, P. J. (2007). Directions in reintroduction biology. Trends in Ecology and Evolution, 23, 20e25. Ashworth, J., Keyes, D., Kirk, R., & Lessard, R. (2001). Standard procedure in the hydrometer method for particle size analysis. Communications in Soil Science and Plant Analysis, 32, 633e642. Barko, J. W., & Smart, R. M. (1983). Effects of organic matter additions to sediment on the growth of aquatic plants. Journal of Ecology, 71, 161e175. Barko, J. W., & Smart, R. M. (1986). Sediment-related mechanisms of growth limitation in submersed macrophytes. Ecology, 67, 1328e1340. Barko, J. W., Adams, M. S., & Clesceri, N. L. (1986). Environmental factors and their consideration in the management of submersed aquatic vegetation: a review. Journal of Aquatic Plant Management, 24, 1e10. Best, E. P. H., Teeter, A. M., Landwehr, K. J., James, W. F., & Nair, S. K. (2008). Restoration options for potential persistence of submersed aquatic vegetation: combining ecological, hydrodynamic and sediment transport modeling. Freshwater Biology, 53, 814e826. Bini, L. M., & Thomaz, S. M. (2005). Prediction of Egeria najas and Egeria densa occurrence in a large subtropical reservoir (Itaipu Reservoir, Brazil-Paraguay). Aquatic Botany, 83, 227e238. Bouyoucos, G. J. (1962). Hydrometer method improved for making particle size analysis of soils. Agronomy Journal, 54, 464e465. Carpenter, S. R., & Lodge, D. M. (1986). Effects of submersed macrophytes on ecosystem processes. Aquatic Botany, 26, 341e370. Case, M. L., & Madsen, J. D. (2004). Factors limiting the growth of Stuckenia pectinata (sago pondweed) in Heron Lake, Minnesota. Journal of Freshwater Ecology,19,17e23. Chang, K. (2010). Introduction to geographic information systems (5th ed.). New York, NY: McGraw-Hill. Demas, G. P., Rabenhorst, M. C., & Stevenson, J. C. (1996). Subaqueous soils: a pedological approach to the study of shallow-water habitats. Estuaries, 19, 229e237. Dibble, E. D., Killgore, K. J., & Harrel, S. L. (1996). Assessment of fish-plant interactions: multidimensional approaches to reservoir fisheries management. American Fisheries Society Symposium, 16, 357e372. Diehl, S., & Kornijow, R. (1998). Influence of submerged macrophytes on trophic interactions among fish and macroinvertebrates. In E. Jeppesen, Ma. Sondergaard, Mo. Sondergaard, & K. Christoffersen (Eds.), The structuring role of submerged macrophytes in lakes (pp. 24e46). New York: Springer Verlag. Doyle, R. D. (2001). Effects of waves on the early growth of Vallisneria americana. Freshwater Biology, 46, 389e397. Duarte, C. M., & Kalff, J. (1986). Littoral slope as a predictor of the maximum biomass of submerged macrophyte communities. Limnology and Oceanography, 31,1072e1080. ESRI.. (2009). ArcGIS desktop and spatial analyst. Redlands, CA: Environmental Systems Research Institute, Inc. Fleming, J. P., Madsen, J. D., & Dibble, E. D. (2011). Macrophyte re-establishment for fish habitat in Little Bear Creek Reservoir, Alabama, USA. Journal of Freshwater Ecology, 12, 105e114. Fleming, J.P. (2010). Macrophyte re-establishment and deductive GIS modeling to identify planting locations for fish habitat enhancement projects. Master’s Thesis. Department of Wildlife, Fisheries and Aquaculture, Mississippi State University, Mississippi State, Mississippi, USA. Grodowitz, M. J., Smart, M., Dick, G. O., Stokes, J. A., & Snow, J. (2009). Development of a multi-attribute utility analysis model for selecting aquatic plant restoration sites in reservoirs. APCRP Technical Notes Collection. ERDC/TN APCRP-EA-21. Vicksburg, MS: U.S. Army Engineer Research and Development Center. Johnson, K. G., Allen, M. S., & Havens, K. E. (2007). A review of littoral vegetation, fisheries, and wildlife responses to hydrologic variation at Lake Okeechobee. Wetlands, 27, 110e126. Jones, J. B. (2001). Laboratory tests for conducting soil tests and plant analysis. Florida: CRC Press LLC.

635

Koch, E. W. (2001). Beyond light: physical, geological, and geochemical parameters as possible submersed aquatic vegetation habitat requirements. Estuaries, 24, 1e17. Lachavanne, J. B., Caloz, R., & Lehmann, A. (Eds.). (1997). Geographic information systems and remote sensing in aquatic botany. Aquatic Botany, vol. 58 (pp. 195e440). Lehmann, A., Jaquet, J. M., & Lachavanne, J. B. (1997). A GIS approach of aquatic plant spatial heterogeneity in relation to sediment and depth gradients, Lake Geneva, Switzerland. Aquatic Botany, 58, 347e361. Lehmann, A. (1998). GIS modeling of submerged macrophyte distribution using generalized additive models. Plant Ecology, 139, 113e124. Madsen, J. D., & Adams, M. S. (1989). The light and temperature dependence of photosynthesis and respiration in Potamogeton pectinatus L. Aquatic Botany, 36, 23e31. Madsen, J. D., Bloomfield, J. A., Sutherland, J. W., Eichler, L. W., & Boylen, C. W. (1996). The aquatic macrophyte community of Onondaga Lake: field survey and plant growth bioassays of lake sediments. Lake and Reservoir Management, 12, 73e79. Madsen, J. D., Getsinger, K. D., Smart, R. M., Skogerboe, J. G., Honnell, D. R., & Owens, C. S. (1999). Evaluation of transparency and light attenuation by Aquashade. Lake and Reservoir Management, 15, 142e147. Madsen, J. D., Wersal, R. M., Tyler, M., & Gerard, P. D. (2006). The distribution and abundance of aquatic macrophytes in Swan Lake and Middle Lake, Minnesota. Journal of Freshwater Ecology, 21, 421e429. Madsen, J. D., Wersal, R. M., & Woolf, T. E. (2007). A new core sampler for estimating biomass of submersed aquatic macrophytes. Journal of Aquatic Plant Management, 45, 31e34. Madsen, J. D. (1999). Point intercept and line intercept methods for aquatic plant management. APCRP Technical Notes Collection (TN APCRP-M1e02). Vicksburg, MS: U.S. Army Engineer Research and Development Center. Malczewski, J. (2000). On the use of weighted linear combination method in GIS: common and best practice approaches. Transactions in GIS, 4, 5e22. Middelboe, A. L., & Markager, S. (1997). Depth limits and minimum light requirements of freshwater macrophytes. Freshwater Biology, 37, 553e568. Miranda, L. E., Spickard, M., Dunn, T., Webb, K. M., Aycock, J. N., & Hunt, K. M. (2010). Fish habitat degradation in U.S. reservoirs. Fisheries, 35, 175e184. Narumalani, S., Jensen, J. R., Althausen, J. D., Burkhalter, S., & Makey, H. E., Jr. (1997). Aquatic macrophyte modeling using GIS and logistic multiple regression. Photogrammetric Engineering and Remote Sensing, 63, 41e49. van Nes, E. H., Scheffer, M., van den Berg, M. S., & Coops, H. (2003). Charisma: a spatial explicit simulation model of submerged macrophytes. Ecological Modelling, 159, 103e116. Rea, T. E., Karapatakis, D. J., Guy, K. K., Pinder, J. E., III, & Mackey, H. E., Jr. (1998). The relative effects of water depth, fetch and other physical factors on the development of macrophytes in a small southeastern US pond. Aquatic Botany, 61, 289e299. Remillard, M. M., & Welch, R. A. (1993). GIS technologies for aquatic macrophyte studies: modeling applications. Landscape Ecology, 8, 163e175. Scheffer, M. (1998). Ecology of shallow lakes. London, UK: Chapman and Hall. Smart, R. M., Doyle, R. D., Madsen, J. D., & Dick, G. O. (1996). Establishing native submersed aquatic plant communities for fish habitat. American Fisheries Society Symposium, 16, 347e356. Smart, R. M., Dick, G. O., & Doyle, R. D. (1998). Techniques for establishing native aquatic plants. Journal of Aquatic Plant Management, 36, 44e49. Smart, R. M., Dick, G. O., & Snow, J. R. (2005). Update to the propagation and establishment of aquatic plants handbook. U.S. Army Engineer Research and Development Center, Lewisville Aquatic Ecosystem Research Facility, Report ERDC/EL TR-05e4. 44 pp. Smiley, P. C., Jr., & Dibble, E. D. (2006). Evaluating the feasibility of planting aquatic plants in shallow lakes in the Mississippi Delta. Journal of Aquatic Plant Management, 44, 73e80. Sondergaard, M., Jeppesen, E., Lauridsen, T. L., Skov, C., van Nes, E. H., Roijackers, R., et al. (2007). Lake restoration: successes, failures and long-term effects. Journal of Applied Ecology, 44, 1095e1105. Storer, D. A. (1984). A simple high sample volume ashing procedure for determination of soil organic matter. Communications in Soil Science and Plant Analysis, 15, 759e772. Tennessee Valley Authority [TVA]. (December 2009). Little Bear Creek operating guide. http://www.tva.com/river/lakeinfo/op_guides/littlebear.htm. Touchart, L., & Bouny, J. (2008). Phytoplankton geographic spatialization in tow ponds in Limousin (France). Applied Geography, 28, 295e310. Valley, R. D., Drake, M. T., & Anderson, C. S. (2005). Evaluation of alternative interpolation techniques for the mapping of remotely-sensed submersed vegetation abundance. Aquatic Botany, 81, 13e25. Van Duin, E. H. S., Blom, G., Los, F. J., Maffione, R., Zimmerman, R., Cerco, C. F., et al. (2001). Modeling underwater light climate in relation to sedimentation, resuspension, water quality and autotrophic growth. Hydrobiologia, 444, 25e42. Vant, W. N., Davies-Colley, R. J., Clayton, J. S., & Coffey, B. T. (1986). Macrophyte depth limits in North Island (New Zealand) lakes of differing clarity. Hydrobiologia, 137, 55e60. Wersal, R. M., Madsen, J. D., McMillan, D. R., & Gerard, P. D. (2006). Environmental factors affecting biomass and distribution of Stuckenia pectinata in the Heron Lake system, Minnesota, USA. Wetlands, 26, 313e321. Wetzel, R. G. (2001). Limnology: lake and river ecosystems (3rd ed.). San Diego, CA: Academic Press.