The effect of terrain structures and environmental factors ... Doccuments/2004/P/P1304.pdf · The effect of terrain structures and environmental factors on the distribution marine

Not to be cited without prior reference to the author

CM 2004/P13: Physical-biological Interactions: Experiments, Models and Observations

The effect of terrain structures and environmental factors on the distribution marine habitats

T. Bekkby, E. Rinde, R. Rosenberg, V. Bakkestuen, L. Erikstad. Trine Bekkby. Norwegian Institute for Nature Research. Dronningensgate 13, 0105 Oslo, Norway [tel: +47 23 35 51 03, fax: +47 23 35 51 01, e-mail: [email protected], URL: www.nina.no]. E. Rinde. Norwegian Institute for Nature Research. Dronningensgate 13, 0105 Oslo, Norway [tel: +47 23 35 51 17, fax: +47 23 35 51 01, e-mail: [email protected], URL: www.nina.no]. R. Rosenberg. Institution for Marin Ecology/Kristineberg Marine Research Station, 450 34 Fiskebäckskil, Sweden. [tel: +46 0523 185 29, fax: +46 0523 185 02, e-mail: [email protected], URL: www.kmf.gu.se]. V. Bakkestuen. Norwegian Institute for Nature Research. Dronningensgate 13, 0105 Oslo, Norway [tel: +47 23 35 51 23, fax: +47 23 35 51 01, e-mail: [email protected], URL: www.nina.no]. L. Erikstad. Norwegian Institute for Nature Research. Dronningensgate 13, 0105 Oslo, Norway [tel: +47 23 35 51 08, fax: +47 23 35 51 01, e-mail: [email protected], URL: www.nina.no] Abstract The coastal areas of the West coast of both Norway and Sweden are characterised by complex relief and environmental and habitat variability, both in time and space. To get a better understanding of ecosystems, ecological functions and the physical-biological interactions, we need to get basic information on the habitats and the factors determining their distribution and abundance. Several studies, mostly terrestrial, have documented that terrain structures and environmental factors influence the distribution of habitats. Hence, we developed integrated GIS-models for the West coast of Norway and for the Gullmarsfjord at the West coast of Sweden as a basis for spatial analyses of the distribution of habitats. Using these models, we analysed the relationships between depth, slope, terrain variability, sea-bed substrate, the degree of wind exposure and the distribution of biotope building species, such as macroalgaes and eelgrass meadows. The results show the usefulness of non-biological factors as indicators of habitat distributions. Knowledge of the relationship between physical factors and habitat distribution may be used to develop predictive models for use in habitat mapping. Background Mapping marine habitats is often complicated and costly, due to factor such as wind, waves, depth and lack of data. Hence, we do not have as good knowledge on habitat distribution in the marine environment as we have on land. We therefore need indicators based on available parameters in areas where data are missing. Several studies have documented that terrain structures (such as depth) and environmental factors (such as sea-bed sediment and degree of wave exposure) influence the distribution of habitats (Kain 1971, Lein et al. 1987, Pasquallini et al. 2001). Predictive models may therefore be a useful tool in mapping marine habitats. The aim The Norwegian Institute for Nature Research, in cooperation with the Institution for Marin Ecology/Kristineberg Marine Research Station, the Geological Survey of Norway and the Norwegian Institute for Water Research, has developed procedures for modelling non-biological factors and developing predictive habitat distribution models. The project presented here is a part of this work, and will provide baseline information on the factors determining the structure and distribution of habitats. To be able to complete this task, we analysed the relationships between different environmental factors (depth, derived terrain structures, exposure models, tidal models and sea-bed substrate) and the distribution of marine coastal sea-bed habitats. This paper presents some results based on two projectr, one in Norway and one in Sweden. More information can be found at the web sites http://www.nina.no/marmodell/ (in English) and http://www.kmf.kva.se/project/modellering/GullmarModell.htm (in Norwegian).

Methods Study and modelling areas The studies (still ongoing) take place in the Sandøy area at the West coast of Norway and in the Gullmarsfjord at the West coast of Sweden. The Sandøy area is typical for the outer coast of Mid Norway. It is part of coastal archipelago, and contains a very large and productive strandflat (Holtedahl, 1959). The highest density of seaweed (Jensen 1998) and the largest standing crop of the ecologically important kelp species Laminaria hyperborea (Sjøtun et al. 1995) are found here. Kelp grows on rock and stones, and the kelp forest is intercepted by channels of sand and shell sand. In the less exposed areas, e.g. the inshore side of the islands, the substrate is dominated by sand and silt. Eelgrass (Zostera marina) is often found. Sand, silt and clay are also covering larger parts of the deeper areas. The Gullmarsfjord has an estuary in the inner part, and the fjord is characterised by steep and rocky slopes intercepted by sandy bays. The sill is 52 m deep and separates the fjord from the outer archipelago. The Gullmarsfjord has one of the highest numbers and diversity of species found in Sweden. The shallow bays with eelgrass meadows are reproduction sites for fish and crustaceans. The area is also an important reproduction site for seals. Input data and the modelling of the non-biological factors The terrain data included in the analyses contains data on • Depth: Information on depth is based on digital data purchased from the Norwegian

Hydrographical Service (NHS) and the Swedish company “Marine Mätteknik”. For Norway, the resolution of the digital depth model (i.e. the size of the grid cells) was 50 m. For Sweden, the resolution was 10 m.

• Slope: The ArcView 3.3 “slope” function was used on the digital depth model, giving the value for the change in depth between each cell and the neighbouring cells. The values are between 0 and 90 degrees.

• Convexity/concavity: The algorithm is based on the fact that a cell with a depth value less (here in the meaning of more negative) than the mean value for the surrounding cells will indicate a basin, and that a cell with a depth value higher than the mean value for the surrounding cells will indicate a top. Estimating terrain characters in this way will highlight different terrain structures depending on the scale that is chosen. We have chosen to estimate the mean within a 1 km x 1 km square, showing large scale structures. A basin is defined as the difference between the cell and the mean of the neighbouring cells of <-1 m, a top is defined as a difference of >+1 m.

• ”Terrain Ruggedness Index”, TRI: TRI is calculated by finding the differences between the elevation of a central pixel and its eight neighbouring pixels in the digital depth model; by squaring each of these differences; by averaging these squared values; and by finding the square root of this average (Blaszczynski 2003). ArcGis, version 8.3, was used to estimate TRI.

The environmental data included in the analyses contains data on • Wind exposure: Degree of exposure is one of the most important factors determining the

distribution of habitats. Our grid-based approach (Rinde & al. 2004) uses an algorithm based on the formula of Oug & al. (1985), and is the Norwegian standard for estimating exposure. The method includes information on fetch (the distance for the wind to build waves) and the wind strength, frequency and direction. Long distance fetch (> 100 km) is weighted 10 times more than medium distance fetch 7.5-100 km), which is again weighted 10 times more than short distance fetch 500 m-7.5 km).

• Sea-bed substrate: Information on sea-bed substrate was only available for the Gullmarsfjord (Sweden). These data were based on backscatter measured by the company “Marin Mätteknik”. For Norway, information on convexity/concavity and TRI was used as indicators of sea-bed substrate, as e.g. soft sediment is rarely is found in steep slopes and very rugged terrain.

Data and analyses Data on presence and percent coverage of habitats were collected using an echo sounder with GPS and an underwater video camera connected to a digital video recorder through a 100 m cable. Data were collected in 2003 and 2004. Transects were not chosen at random, but to cover the different combinations of depth, slope, terrain variability and degree of exposure. A transect started close to the

shore, continued towards the deeper areas and ended when 3-5 subsequent recordings contained homogenous soft sea-bed sediment. Data was integrated in a geographical information system. The analyses of the relationship between non-biological factors and habitat distribution were made in SPSS 12.0.1 for Windows and the models were developed in ArcView, version 3.3. Results - the habitat distribution models This work is ongoing, and several of the analyses are yet to be finished. Some preliminary results are presented here. Modelling sea-bed substrate Sea-bed substrate is one of the most important factors in the EUNIS habitat classification system and is important for the distribution of other habitats. Sea-bed substrate may partly be modelled by using different indices for terrain variability. For the Sandøy area, information on sea-bed substrate was not available. Hence, we developed an approach for modelling potential soft sea-bed substrate (defined as concave areas with low terrain variability) and potential rocky sea-bed (defined as areas of very high terrain variability). The figure below shows modelled rocky and soft sea-bed sediment from the West coast of Norway. Note that our algorithm only chooses the areas with high probability of finding rocky and soft sea-bed substrate, respectively. Hence, our model is an underestimate.This model was therefore not used as input data in the kelp forest and eelgrass modelling presented later. Including information on currents will most likely improve the substrate model significantly.

Modelled rocky sea-bed substrate and

Modelled soft sea-bed substrate on the West coast of Norway

Modelling potential kelp (Laminaria hyperborea) forest areas in Norway Kelp forests have a particular significance for marine biodiversity along the Norwegian coast. They have different depth distribution depending on the degree of exposure. The more exposed the area, the deeper we find the kelp. Based on field data, we modelled kelp forest distribution using information on depth and degree of wave exposure. In the most exposed sites on the west coast of Norway, dense kelp forests will reach a depth of 25 m. In areas with medium exposure, dense kelp forests may go down to about 20 m. The figure below shows modelled kelp forest (Laminaria hyperborea) from the West coast of Norway. As long as substrate is not included in the model, the distribution of kelp forests will be overestimated, as the sea-bed will consist of sediment not suitable for kelp. Including data, or even coarse models, of sea-bed substrate will most likely improve the model. However, till sediment is difficult to model, particularly based on marine depth data only. Till sediment will consist of fine grained sediment, rocks and boulders in a mixture, and kelp occurrences will depend on the amount of rocks and boulder. It is important to note that this is a model of the kelp forest, i.e. areas with high and medium density of kelp. Areas of low kelp density were not included in the analyses.

Modelled kelp forest (Laminaria hyperborea) on the West coast of Norway

Modelling sublittoral sandbanks in Sweden Sublittorale sandbanks belong to the Natura 2000 habitat 1110 and the EUNIS habitat classes A4.1, A4.2, A4.5 and A4.6). They are defined as being covered by sea water all the time. It’s most shallow point is at 20 m depth, and the lower limit is defined by lack of primary production at the sediment surface, reductions in wave or current exposure or lack of typical sandbank communities. The sandbanks may be barren or with vegetation (e.g. eelgrass, Zostera marina, modelled distribution of eelgrass is presented later). Based on these definitions and data obtained in the field, we modelled the distribution on sublittoral sandbanks from • Backscatter values>6.9, i.e. substrate other than rock and large stones • Convexity/concavity values <-1 at a 50 m x 50 m scale, i.e. a basin • Areas down to 30 m depth containing a convex area shallower than 20 m

Modelling potential areas of eelgrass (Zostera marina) in Norway and Sweden Our analyses showed that eelgrass meadows should be expected in sheltered or medium exposed, shallow and relatively flat areas. All densities of eelgrass, even very scattered, were included in the analyses. This is due to observations of rapid changes in eelgrass distribution. Hence, we chose to defined an area as a potential eelgrass habitat as long as the area proved itself to be suitable for eelgrass growth (i.e. eelgrass was present independently of density). The criteria for the distribution of eelgrass varied between the observed locations in Sweden and Norway, especially with regard to wave exposure. Backscatter was only available for the Swedish sites. In Norway, eelgrass was found in areas of • medium exposure • moderate slope (<7 degrees) • shallow depths (<7 m) In Sweden, eelgrass was found in areas of • backscatter values>6.9, i.e. substrate other than rock and large stones • depth down to 8.5 m • slope<10 degrees • In very sheltered areas The figures below show modelled distribution of eelgrass at the West coast of Norway and the West coast of Sweden (the outer Gullmarsfjord area), respectively. For the West coast of Norway, the model is most likely to be an overestimate, as we did not have data on substrate. In some areas, flat terrain indicates soft sea-bed substrate. However, in this area, moraine deposits will contain large rocks and will function as a rocky substrate, even though it appears, at our scale, flat.

Modelled eelgrass (Zostera marina) on the West coast of Norway

Modelling potential areas of shell sand in Norway We have modelled the potential distribution of shell sand at two levels. First we have modelled the potential primary accumulation sites, i.e. the locations where shells are most likely to be found after being transported from the site of the living organisms. We modelled the accumulation sites as the transition zones between wave exposed and sheltered areas in the outer coast. Second we modelled the secondary location of shell sand as concave areas in sheltered and medium exposed areas within 1.5 km from the primary accumulation site. The secondary locations will to a large extent depend on the strength of currents, but such data are not yet available.

Discussion Improving the models Modelling the marine habitats depends on correct information on sea-bed substrate. E.g. improving the Zostera model includes finding areas of soft sea-bed. Similarly, predicting kelp forest areas depends on the prediction of rocky substrate. Substrate type is most likely determined by factors such as terrain structures (e.g. slope) and wind, current and tidal exposure. Many of these factors have not been available for this project and are, in many areas, not available at all. However, by combining data on depth, terrain structures (such as slope and terrain ruggedness) and wave exposure models, we have been able to model sea-bed substrate at the higher level in the EUNIS classification system. Even though the substrate models still are coarse and inaccurate, they are getting better. Our next step will be to include information on tidal currents. The challenge with till deposits it still unsolved. This is a challenge regarding its definitions (till contains both rocks and soft sediment) and how to predict their distribution (determined not by exposure, but by historic events, and the terrain structures often disappear at the scale that digital data often are available at).

Modelled shell sand on the West coast of Norway

Need for management and planning across borders - relating to European classification systems In Norway, extensive efforts have been put into developing infrastructures for classifying habitat types. Habitat type classifications are frequently used in terrestrial (e.g. Fremstad 1997, DN 1999), aquatic and shoreline (DN 1999) management and planning. However, no comprehensive and unified habitat type classification system has been applied to the coastal zone of Norway. In Sweden, both the EU Water Framework and the Habitat Directive apply. Hence, Sweden has implemented a system for classifying Natura 200-habitats. However, this system has its weaknesses, and a more hierarchical system is needed. The European Environment Agency (EEA, see Davies and Moss 1999) has developed a classification system, EUNIS, in collaboration with the Council of Europe. The EUNIS Habitat classification has been developed to facilitate harmonised description and collection of data across Europe through the use of criteria for habitat identification. EUNIS is a comprehensive pan-European system, covering all types of habitats. EUNIS builds on the CORINE and Palaearctic Habitat classifications, and is built to link to and correspond with other major habitat systems in Europe, e.g. it cross-references to all EU Habitat Directive habitat types used for EU Member States and it contains and will continue to include relevant marine habitat types as they are developed in collaboration with the OSPARCOM marine work. Hence, using the EUNIS system will make it easier to communicate across borders. An important aspect of the implementation of EUNIS will be to identify the level in the EUNIS hierarchy to which habitat types may be modelled using regionally available data, and whether or not additional and more detailed information is needed. Our studies aim at testing and evaluating the usefulness of this system using data at different scales and in different kinds of coastal seascapes. The effect of scale on the implementation of a classification system Evaluation of how environmental patterns change with focal scale is a fundamental step in understanding ecosystems (e.g. Wiens 1989, Holling 1992, Levin 1992). There is no optimal scale at which ecosystems should be described, classified, modelled and mapped, and a landscape might appear heterogeneous at one scale, but quite homogeneous at another (Forman and Godron 1986). A researcher choosing a scale for investigation must understand and consider how a change of scale affects the level of detail obtained from the results (Meentemeyer and Box 1987). The challenge is to determine the resolution required to study the ecological properties at different levels of a habitat type classification system (Jorgenson 2000). However, the choice of scale will also have an economic aspect as an increase in resolution most often will involve higher costs. In order to develop a predictive model for the distribution of habitat type classes that may be regionally applied, we will in the future study the possibilities for classifying habitat types according to the EUNIS system for input data of different scales. Literature • Blaszczynski, J. 2003. Geomorphometric analysis of surface landscape features. The web site of

the National Science & Soil Resource Center, Denver, Colorado, USA. http://www.blm.gov/nstc/ecosysmod/surfland.html

• Davies C. E. and Moss, D. 1999. EUNIS Habitat Classification. Final Report to the European Topic Centre on Nature Conservation, European Environment Agency. 256 p.

• DN 1999. Kartlegging av naturtyper - verdisetting av biologisk mangfold. Direktoratet for Naturforvaltning, håndbok nr 13. 244 p.

• Forman, R. T. T. & Godron, M. 1986. Landscape ecology. New York: Wiley. • Fremstad, E. 1997. Vegetasjonstyper i Norge. NINA Temahefte 12: 279 p. • Holling, C. S. 1992. Cross-scale morphology, geometry and dynamics of ecosystems. Ecological

Monographs 62: 447-452. • Holtedahl, H. 1959. Den norske strandflate. Norsk Geografisk Tidsskrift XVI: 285-303. (In

Norwegian) • Jensen A. 1998. The seaweed resources of Norway. In: Critchley ATl, Ohno M, editors. Seaweed

resources of the world. Kanagawa International Fisheries Training Centre, Japan International Cooperation Agency. p. 200-209.

• Jorgenson, M. T. 2000. Hierarchical organization of ecosystems at multiple spatial scales on the Yukon-Kuskokwin Delta, Alaska, U.S.A. Arctic, Antarctic and Alpine Research 32(3): 221-239.

• Kain JM. 1971. Synopsis of biological data on Laminaria hyperborea. FAO Fisheries Synopsis No 87. 68 p.

• Lein TE, Sivertsen K, Hansen JR, Sjøtun K. 1987. Tare- og tangforekonster i Finnmark. Del 1: hovedrapport. Rapport til Finnut A/S Finnmark utbygningsselskap. 118 p. (Norwegian).

• Levin, S. A. 1992. The problem of patterns and scale in ecology. Ecology 73: 1943-1967. • Meentemeyer, V., Box, E. O. 1987. Scale effects in land-scape studies. In Truner, M. G. (ed.):

Landscape heterogeneity and disturbance. New York: Springer Verlag. • Oug, E., Lein, T.E., Holthe, B., Ormerod, K. Næs, K. 1985. Basisundersøkelser i Tromsøsund og

Nordbotn 1983. Bløtbunnsundersøkelser, fjæreundersøkelser og bakteriologi. NIVA Report 173b/84.

• Pasqualini, V., Pergent-Martini, C., Clabaut, P., Marteel, H. and Pergent, G. 2001. Integration of aerial remote sensing photogrammetry, and GIS technologies in seagrass mapping. Photogrammetric Engineering and Remotre Sensing 67(1): 99-105.

• Rinde, E., Sloreid, S.-E., Bakkestuen, V., Bekkby, T., Erikstad, L. & Longva, O. 2004. Modellering av utvalgte marine naturtyper og EUNIS klasser. To delprosjekter under det nasjonale programmet for kartlegging og overvåking av biologisk mangfold. NINA Oppdragsmelding 807. 33 pp. (Norwegian)

• Sjøtun K, Fredriksen S, Rueness J, Lein TE. 1995. Ecological studies of kelp Laminaria hyperborea (Gunnerus) Fosslie in Norway. In: Skjoldal HR, Hopkins C, Erikstad KE, Leinaas HP, editors. Ecology of fjords and coastal waters. Elsevier Science. p. 525-536.

• Wiens, J. A. 1989. Spatial scaling in ecology. Ecology 3: 385-397.