4th International Conference on Integrating GIS and Environmental Modeling (GIS/EM4):
Problems, Prospects and Research Needs. Banff, Alberta, Canada, September 2 - 8, 2000.

Modeling submerged macrophytes in fluvial environment using 2D physical habitat simulation

GIS/EM4 No. 33

Jean Morin
José Bechara
Michel Leclerc

Abstract

Submerged macrophytes are abundant in the Saint-Lawrence River (Canada). These plants change drastically currents distribution, wind waves growth and sedimentation processes. Abiotic factors control their habitat and determine the spatial distribution of plant species and biomass. The local fluvial habitat of submerged vegetation is described using a method derived from microhabitat modeling in order to establish the basis for a quantitative and predictive model of submerged vegetation. The abiotic factors controlling the distribution of plants have either been measured directly or result from numerical simulations, and they are known for the whole Lake Saint-François over a grid mesh of the order of 50 m in diameter. Currents, waves, nutrients associated with the substrate, and light reaching the bottom are simulated. Relationships between plants sampling sites and the local abiotic factors are analyzed in order to produce a matrix of habitat preferences.

Keywords

Habitat modeling, Submerged plants, Fluvial ecosystem, Hydrodynamic, Waves, Suspended load, St. Lawrence River

Introduction

Submerged aquatic plants are present all along the St. Lawrence River corridor and for such a large river, they are surprisingly abundant. This is mainly related to the relatively low concentration of suspended load in the river. Submerged plants are an important component of the river's productivity. In the context of global climatic changes and discharge regulation, it is imperative to understand and quantify the impact of water level fluctuations on the distribution and biomass of submerged plants. To this end, efforts are currently underway at Environment Canada under the St. Lawrence Action Plan 3 to integrate biotic and abiotic information on the St. Lawrence River within a georeferenced decision-support system that combines existing information with various 2D mathematical models.

Lake Saint-François is a fluvial lake within the St. Lawrence system; it is 62 km long and 7 km wide with a mean flow of 7 500 m³/s. The bottom of Lake Saint-François, except for zones exceeding 6 to 8 meters in depth, is covered with plants during the summer (Morin and Leclerc 2000). Macrophytes are so abundant that they drastically modify the physics of their environment during the growth season. Their presence : 1) modifies the currents, reduces the flow in their habitat and thus increases the flow in the main channels (Petticrew and Kalff 1992, Morin and Leclerc 2000) ; 2) absorbs wave energy (Kobayashi et al. 1993, Camfield 1977); 3) increases the sedimentation of fine material rich in nutrients by filtering and by reducing the current velocity (Dale 1986, Spence 1982,Chambers et al 1991); 4) increases light penetration by increasing sedimentation of suspended load.

Figure 1: Geographic setting of the Lake Saint-François, Saint-Lawrence River.

Submerged plants, because of their ability to improve their own habitat, are a very complex component to model. Relationships between abiotic factors and plants are non-linear and retroaction loops are present during the growth season as plants improve their habitat quality downstream and downwind. However, field data suggest that within the Saint-Lawrence River and particularly in Lake Saint-François (Morin and Leclerc 2000), abiotic factors control macrophyte habitat and determine a large portion of their spatial distribution and biomass. The adaptation of plants to specific environmental conditions translates into various mechanical and physiological aspects, such as growth phase, stem flexibility or reproduction mode (Chambers and Kalff 1987, Chambers 1987 ).

The modeling effort aims to map the species distribution over the entire lake, 1) model and map abiotic variables on the entire domain with precision, 23) establish the relationships between the occurrence of plants and their habitat preferences, 3) interpret ecological adaptation of plants and prepare the modeling of plant species distribution and biomass.

Habitat modeling

The fluvial habitat of submerged vegetation is described using a method derived from microhabitat methodologies developed for salmonid species (Leclerc et al. 1994 : 1995; Stalnaker 1994; Bovee 1996). In general, these studies use only 3 variables as description factors : currents, depth and substrate, that are known over the entire river reach of interest. The combination of fish preferences and data distribution are integrated into mapable preferences and suitability indexes.

For submerged aquatic plants, the habitat is controlled by a relatively large number of abiotic factors: currents, wind waves, light penetration (suspended load), sediment composition (nutrients and physical characteristics), bottom slope and water depth. All these abiotic factors have complex interactive relations: they vary with time (f(t)) depending on the season, discharge and wind and vary with space (f(x,y)) depending on the topography and geology. Abiotic factors vary in space and time, and a combination of these variations at any given point creates the environmental conditions of a specific Eabitat.

Bidimensional models developed for river and coastal hydraulics calculate the abiotic factors with sufficient precision at all points of the area modeled, even under transitory conditions. Current velocity, wave energy, fine material deposition and light penetration can all be simulated for various conditions judged critical for aquatic plants. The abiotic factors being estimated for the whole domain, at each point (calculation node) a data set is obtained (vector, matrix) representing the local habitat conditions. These values at each node and associated biological data can easily be addressed with statistical tools developed for ecosystems analysis. The resulting statistical models which describe habitat preferences can then be applied to the entire domain since the abiotic variables are known everywhere and eventually be used as a predictive model for aquatic plants distribution.

Submerged plants data

Data on species classification were collected in 1995 during an echosounding survey. The characterization used an echosounding technique derived from Fortin et al. (1993) to which a dGPS positioning system was coupled; a data logger also recorded the course and plotted regular positions on the sounder. In addition, an underwater video camera allowed species identification and determination of their relative proportion. The field survey was conducted in October 1995 during the maximum submerged macrophyte growth for that year; 35 echosounder transects were recorded. Soundings were then described in terms of species composition, height of each species, their proportion and their relative abundance. From this field survey, plant distribution and physical characteristics were mapped to simulate abiotic factors in the presence of plants.

To build the Habitat Data Base (HDB), a series of 580 points with known dGPS positions were assembled. Sampling points correspond to fixes taken at every 50 m in a straight line to position the sounder paper rolls. The 580 points are distributed throughout the whole Lake Saint-François (Figure 3). These points were described in terms of the proportion of species present, plant height and local density to provide an estimate of biomass. The HDB includes 11 classes of plants and growth forms (Table 1).

Biomass is estimated by a method derived from the concept of "biomass density" developed by Duarte and Kalff (1990). "Biomass density" is characteristic of the species and its growth form and represents the mass of living tissue per volume occupied by the species in the field.

Data management system : modeleur

Data must all be georeferenced and distributed over the whole study area. Digital data management is realized by MODELEUR (Secretan and Leclerc 1998). MODELEUR is a computer software similar to a GIS, which uses finite elements as discretization and interpolation tools. The software was developed specifically for river applications and has an excellent modeling capacity based on interactions between layers of data. With MODELEUR, information layers can be projected on various types of mesh grids : triangular, square or linear.

Abiotic data : modeling and processing

The habitat model is built from abiotic factors controlling plant distribution. The value of these factors must be known at plant sampling points and the majority of these variables are obtained by numerical simulation. A total of 24 layers of distributed abiotic factors influencing habitats are used to represent most of the abiotic conditions present in the system during a normal spring, summer and fall, with and without the influence of submerged plants.

Water depth is the local difference between topometry and water level (Figure 3). Topometry is the key variable of the system; therefore, it must be known as accurately as possible. It is the density of these data that determines the precision limits of the whole model. Topometry also has a broad influence on plant habitats, not only in terms of "depth", i.e. the vertical space available, but also in a lateral way by conditioning the directional parameters such as waves and currents and by determining the bottom slope. Basic topometry data are derived from soundings by the CHS (Canadian Hydrographic Survey); a total of 292 270 soundings, with a precision of ±0.1 m, were integrated in the numerical terrain model. These measures have a mean density of about 1 point at each 10 m.

Substrate mapping proceeded from data collected during different field surveys. The map is the result of an interpolation " by expert opinion " of grain size data and visual observations; its precision is a function of the density of sampling. Mean grain size appears to be the best index of the physical properties of the substrate.

Currents were simulated using hydrodynamic modeling with the HYDROSIM model (Heniche et al. 1999). This model uses a discretization of the shallow water equations solved by the finite elements method. The high density grid of 55 997 elements (114 562 nodes) was developed to increase the precision of the hydrodynamic calculations at an appropriate scale for the analysis of plant communities, i.e. an average mesh of about 40 m (Figure 2). Two average events of 7500 m³/s were simulated on the high density mesh grid. These two simulations represent conditions of currents in the absence and in the presence of plants (Figure 3), also they provide the water levels required to calculate water depth; they also provide values of diffusion potential and shear stress for the transport-diffusion model.

Figure 2: Refined finite element mesh used for simulation of abiotic factors.

Point Source-Grid Cell Intersection Algorithm

Modeling waves and integrating the values in the HDB is a fairly complex procedure which must take into account friction on submerged plants and the broad variability of wind direction and strength (Figure 3). The Delft University HISWA model (Booij et al.1993) simulates the growth and the behavior of waves in shallow water. This shallow water wave model calculates various parameters such as wave energy, frequency, height and direction on a regular square grid. It also simulates wave build-up in shallow water, and the propagation and dissipation of waves by surf (surf breaking and white-capping), bottom, vegetation and currents. The model is particularly well adapted to the context of Lake Saint-François, where topometry is complex, currents are strong and vegetation is abundant. Waves are calculated on a square grid with 277 302 elements of 50 m sides, covering the whole of Lake Saint-François. The orbital "near-bottom" speed generated by waves appears to be the best index of stress to plants. However, the model does not take into account the impact of submerged plants on wave energy absorption. This was solved by integrating a modification to a function designed for emerging plants.

Light penetration to the bottom is a function of water clearness and depth. The attenuation coefficient varies mostly as a function of the quantity of suspended matter. In order to simulate the distribution of the concentrations of suspended matter in the lake, the DISPERSIM transport-diffusion model adapted for sedimentation simulations was used (Padilla et al. 1997). DISPERSIM is an eulerian transport-diffusion model developed at INRS-Eau (Université du Québec, Sainte-Foy, Canada), in which concentration is the state value to be solved.

Nutrient deposition on the bottom is simulated in a similar manner DISPERSIM allows to simulate the deposition of suspended matter. The same simulations as used for light penetration produce data on the amount of suspended matter deposited at the bottom. This value is obtained by post-processing of the simulation results, by multiplying the local concentration by the local deposition velocity.

Figure 3: Examples of distributed abiotic factors, comprising measured data like water depth, simulated data like currents and waves (wind 35 km/h integrated average directions) and submerged plants sampling location.

Data base and analysis

The Habitat Data Base (HDB) containing the information from 24 abiotic factors and 11 species (and growth form) biomass on 580 different locations is used to analyze habitat preferences. The abiotic conditions without plants are the initial conditions of a predictive model for plants, while the conditions in the presence of plants are its final conditions. In order to fine tune the analysis and to avoid possible biotic interactions, the analysis is conducted with the data on monospecific communities (100%) and on significant incidental species (>30%). Results were analyzed with monovariate distributions of biomass as a function of each of the 24 abiotic variables in the database, pointing towards interesting conclusions about plants ecology.

Analysis of univariate relations between biomass data for every species (and growth form) and abiotic factors show sufficient segregation for modeling purposes. The Table 2 shows habitat preferences as roughly determined with local abiotic factors. Every species has preferences that can discriminate its habitat.

Predictive use of these habitat preferences is planned and will allow the possibility of simple mapping with rough parameterization. Fine tune multivariate analysis of the relations between species and their habitat characteristics is under process. Preliminary results show an R² of 0.75 for the biomass determination and several impressive results for species determination with success varying from 85 to 95% of validity. Again, complexity is expected for the predictive use of this work because of the non-linear relationships between the plants and their environment.

References used

Booij, N., L.H. Holthuijsen, J. Dekker and R. Shoonbeek 1993. Standard tests for the shallow water wave model HISWA version 100.21 Delft University of Technology, Department of Civil Engineering, Group of hydraulic and Geotechnical Engineering, 43 p.

Bovee, K.D. 1996. Perspectives on two-dimensional river habitat models : the Phabsim experience. Compte-rendus du 2e symposium International sur l'écohydraulique. Ecohydraulique 2000. Québec, 16 juin 1996. B149-162.

Camfield, F.E. 1977. Wind-wave propagation over flooded, vegetated land. Tech. paper No. 77-12, Coast Engrg Res. Ctr., U.S. Army Engineer Waterways Experiment Station, Vicksburg, Miss.

Chambers, P.A. 1987. Nearshore occurrence of submersed aquatic macrophytes in relation to wave action. Can. J. Fish. Aquat. Sci. 44: 1666-1669.

Chambers, P.A. and J. Kalff 1987. Light and nutrients in the control of aquatic plant community structure. II. In situ observation. Journal of Ecology 75 : 621-628.

Duarte, C.M. and J. Kalff 1990. Biomass density and the relationship between submersed macrophyte biomass and plant growth form. Hydrobiologia 196 : 17-23.

Heniche, M. Y. Secretan, P. Boudreau, and M. Leclerc. 1999. A new 2-D finite element drying-wetting model for rivers and estuaries. Accepted for publication in Advances in water ressources

Leclerc, M., A. Boudreault, J. Bechara and G. Corfa 1995. Two-dimensional hydrodynamic modeling : a neglected tool in the instream flow incremental methodology. Transaction of the american Fisheries Society 124 :645-662

Leclerc, M., P. Boudreau, J. Bechara, L. Belzile and D. Villeneuve 1994. Modélisation de la dynamique de l?habitat des jeunes stades de saumon atlantique (salmo salar) de la rivière Ashuapmushuan (Québec, Canada). Bull. Fr. Pêche Piscic. 332 :11-32.Morin and Leclerc 2000

Padilla, F., Y. Secretan and M. Leclerc 1997. On open boundaries in the Finite Element Approximation of two-dimensional Advection-Diffusion Flows. Int. J. Num. Meth. in Engineering 40:

Petticrew, E.L. and J. Kalff 1992. Water flow and clay retention in submerged macrophyte beds. Can. J. Fish. Aquat. Sci., 49: 2483-2489.

Secretan Y. and M. Leclerc 1998. A 2D hydrodynamic GIS and simulation software. Proceeding of the third international conference on Hydroinformatics 98. IAHR. Copennagen Danmark. August 24-26 1998.

Stalnaker, C.B. 1994. Evolution of instream flow habitat modeling. in: The rivers handbook: hydrological and ecological principles. P. Calow and G.E. Petts (ed). Blackwell scientific publications, Oxford.

Authors

Jean Morin, Reseacher, Environment Canada, Meteorological Service of Canada
1141 rte de l'Église, Ste-Foy QC, Canada
E-mail: jean.morin@ec.gc.ca

José A. Bechara, Professor, Instituto de Ictiología del Nordeste (INICNE)
Facultad de Ciencias Veterinarias. UNNE
S. Cabral 2139. -3400- Corrientes. Argentina
E-mail: jbechara@vet.unne.edu.ar

Michel Leclerc, Professor, Institut national de la recherche scientifique-Eau, Université du Québec
2800 Einstein, Ste-Foy, QC, Canada
E-mail: Michel_Leclerc@inrs-eau.uquebec.ca