Elsevier

Ecological Modelling

Volume 114, Issues 2–3, 1 January 1999, Pages 287-304
Ecological Modelling

Habitat and population modelling of roe deer using an interactive geographic information system

https://doi.org/10.1016/S0304-3800(98)00164-1Get rights and content

Abstract

Management of German roe deer (Capreolus capreolus) populations is a challenge for wildlife managers and foresters because population densities are difficult to estimate in forests and forest regeneration can be negatively affected when roe deer density is high. We describe a model to determine deer population densities compatible with forest management goals, and to assess harvest rates necessary to maintain desired deer densities. A geographic information system (GIS) was used to model wildlife habitat and population dynamics over time. Our model interactively incorporates knowledge of field biologists and foresters via a graphical user interface (GUI). Calibration of the model with deer damage maps allowed us to evaluate density dependence of a roe deer population. Incorporation of local knowledge into temporally dynamic and spatial models increases understanding of population dynamics and improves wildlife management.

Introduction

The purpose of this study is to develop a tool for managing roe deer in Germany. Management of roe deer is controversial because foresters claim that high roe deer populations inhibit natural tree regeneration. In contrast, hunting organizations oppose reducing deer populations, and perceive environmental factors (e.g. acid rain) as the main constraint on forest regeneration. Sound management of roe deer populations is complicated by the difficulty of assessing the population density of roe deer in forested areas (Strandgaard, 1972, Vincent et al., 1991).

Roe deer are common throughout Europe and are the most important game species in Germany. The species is managed in administrative units usually not larger than 10 km2. For each hunting unit, one hunter leases hunting privileges for 9–12 years. Under the current management system in Germany, the hunting administration prescribes the annual harvest for each hunting unit. The harvest plan specifies the number of roe deer to be harvested in three age classes for each sex. The hunting leaseholder must report every deer killed and is legally responsible for fulfilling the harvest plan (Ueckermann, 1988). The number of deer to be harvested is calculated from a spring population estimate by the hunting leaseholder under the assumption of a recruitment rate of 1.0 fawns per breeding female (fawn sex ratio 1:1). This procedure is not based on current scientific knowledge about roe deer population dynamics, such as density dependence. It does not capture the dynamics of roe deer populations, changes in habitat suitability, or previous hunting success. These shortcomings in the assessment of the harvest plans result in their low acceptance and poor fulfilment by many hunting leaseholders.

We present an improved approach to roe deer management, incorporating current knowledge of population dynamics and modelling techniques, to assess roe deer densities and harvest plans. Our goals are: (i) to develop a roe deer management model that links habitat suitability and population dynamics; (ii) to make the model adaptable to local conditions via a graphical user interface (GUI); and (iii) to assess whether harvest rates are adequate to prevent forest damage by roe deer browsing.

Roe deer prefer to forage near protective cover and are often found in early successional habitat and forest plantations. The presence of spatial structures (e.g. forest/field edges) determines habitat suitability of a management unit for roe deer. Thus a geographic information system (GIS) -based approach for assessing habitat for roe deer is appropriate.

GIS is often used to derive habitat suitability models from a set of GIS layers (Donovan et al., 1987, Pearce, 1987, Aspinall, 1991, Aspinall and Veith, 1993), that describe large areas on a relatively coarse scale (Heinen and Mead, 1984, Scott et al., 1993, Mladenoff et al., 1995). Most GIS models of wildlife habitat capture only one point in time rather than habitat dynamics over time (Ormsby and Lunetta, 1987, Aspinall, 1991). Dynamic models of wildlife habitat interactions often do not incorporate GIS functionality (Bhat et al., 1996, Bettinger et al., 1997, Stankovski et al., 1998). Dynamic GIS based models have only recently been developed (Ozesmi and Mitsch, 1997), and used to optimize wildlife habitat spatially (Nevo and Garcia, 1996, Garcia and Armbruster, 1997).

Roe deer of both sexes occupy small home ranges and adult males defend territories. In central Europe, yearling roe deer of both sexes disperse only short distances (Wahlström and Liberg, 1995). In several European studies, the majority of marked fawns were recaptured or harvested as adults within 1 km of the capture site. Dispersal distances of more than 10 km are rarely observed (Ellenberg, 1978, Danilkin and Hewison, 1996). The small hunting units and the small home ranges of roe deer require a management tool that operates at the local scale (1–10 km2). Previous GIS-based studies on deer species have operated on much broader scales, making their approach poorly suited for managing a species with small home ranges within small harvest units. (Tomlin et al., 1983, Leckenby et al., 1985, Milne et al., 1989, Huber and Casler, 1990, Wright and Boag, 1994). None of the existing roe deer habitat assessment tools for Central Europe are GIS based (Müller, 1964, Bobek, 1980, Ueckermann, 1988).

Early research on roe deer population dynamics by Strandgaard (1972) and Bobek, 1977, Bobek, 1980 neglected density dependence and assumed dispersal to be the primary regulator of population density. Recent evidence indicates that roe deer populations are not regulated by dispersal (Vincent et al., 1995, Wahlström and Kjellander, 1995, Wahlström and Liberg, 1995). The high emigration rates in Strandgaards’ study appear to be due to the juxtaposition of the study area, where no harvest occurred, with the surrounding area, where harvest was high (Gaillard et al., 1993).

During the last decade, research groups in France (e.g. Gaillard et al., 1992, Gaillard et al., 1993, Vincent et al., 1995, Gaillard et al., 1996), Scandinavia (e.g. Wahlström and Liberg, 1995) and the UK (e.g. Hewison, 1996, Putman et al., 1996) have presented evidence for two density dependent factors influencing roe deer population dynamics. First, in regions of mild climate, there is an inverse correlation between body mass and population density (Blant, 1991, Gaillard et al., 1992, Gaillard et al., 1993, Vincent et al., 1995, Gaillard et al., 1996). Body mass is positively correlated with the probability of pregnancy in roe deer does younger than 20 months (Gaillard et al., 1992, Hewison 1996). Second, juvenile survivorship increases with decreasing population density (Fruzinski and Labudzki, 1982, Gaillard et al., 1992).

The main density independent factor in roe deer population dynamics is climate (Gaillard et al., 1993). Winter snow depth is negatively correlated with survival rates in all age classes (Fruzinski et al., 1983, Gaillard et al., 1993). In most German hunting grounds, winter mortality related to snow depth is reduced by supplemental feeding. Precipitation in April and May is negatively correlated with male fawn body weight (Gaillard et al., 1996) but precipitation in May and June is positively correlated with fawn survival rates for both sexes (Gaillard et al., 1997). These complex responses of roe deer populations to weather make their modelling difficult.

In summary, a management tool for roe deer useful at the local scale requires use of a dynamic habitat suitability model in a GIS that can incorporate habitat change. The model should provide a GUI to make it easily accessible and flexible for wildlife managers. Finally, the habitat model must be linked to a population model that incorporates density dependence. We describe such a model.

Section snippets

Study area

The study site, Holzerath, is located in western Germany, 50 km Southwest of Luxembourg city. It covers 673 ha, of which 513 ha are forested, 135 ha are agriculture, and 25 ha are covered by settlements (Fig. 1). The potential natural vegetation type is beech (Fagus silvatica) forest. Forests are managed in even-aged stands with rotation cycles ranging from 30 years for firewood up to 150 years for timber. Half of the forested area on the study site is dominated by conifers (primarily spruce (

Methods

Our roe deer model is structured in two major parts (Fig. 2). The first part is a spatially explicit habitat model, based on a GIS land cover data set. This part of the model includes an interactive GUI. The second part is an iterative population model that calculates population levels for single hunting units. The model output for our study area was validated by deer browse maps made by the Ministerium für Landwirtschaft, Weinbau, Umwelt und Forstwirtschaft Rheinland-Pfalz (i.e. the State

Habitat suitability

The habitat suitability index is comprised of four components: geology, tree composition, grassland, and forest/plowed field boundary (Ueckermann, 1988). The geology value of the study site Holzerath is 20 (sandstone). The tree species composition was relatively stable during the study period; the calculated value was 15 (oak <30% and no other species >50%). The total length of the forest boundary was 6.72 km, with 0.17 km of forest/plowed field boundary. These values also remained stable; the

Discussion

We chose the study site Holzerath partly because of the habitat changes due to windthrow which resulted in an increase of the culling plan by 30% in anticipation of a major increase in the roe deer population. We assumed the habitat value would change substantially. Contrary to this assumption, changes in the habitat value, as assessed from the Ueckermann, 1951, Ueckermann, 1957, Ueckermann, 1988 habitat model, were small. Whether Ueckermann’s model is not sensitive enough or vegetation changes

Conclusion

We developed a GIS model at the local scale for managing roe deer populations in Germany. The model offers a tool for resolving conflicts about roe deer densities, harvesting levels, and browse damage, between forest managers, wildlife managers, and hunters. It translates current scientific understanding into a management tool suitable for every-day use by wildlife managers. Compared to the current management practice, our model improves the assessment of roe deer population densities by

Acknowledgements

We are thankful to Holzerath hunting leaseholder H. Schulten and forester G. Franzen who provided us with both input data and insight into our study site. W. Mackaness provided guidance for developing the interactive habitat model. J.-M. Gaillard, M. Hewison, S. Kohlmann, S. Lutz, N. Mathews, and T. Wiegand reviewed drafts of our manuscript and gave most valuable suggestions.

References (62)

  • R. Aspinall

    Use of an inductive modelling procedure based on Bayes theorem for analysis of pattern in spatial data

  • R. Aspinall et al.

    Habitat mapping from satellite imagery and wildlife survey data using a Bayesian modelling procedure in a GIS

    Photogr. Eng. Rem. Sens.

    (1993)
  • Begon, M., Mortimer, M., Thompson, D.J., 1996. Population Ecology: a Unified Study of Animals and Plants, 3rd ed....
  • E. Bideau et al.

    Effects of age and sex on space occupation by European roe deer

    J. Mamm.

    (1993)
  • M. Blant

    Reproduction in roe deer populations in Western Switzerland

  • B. Bobek

    Summer food as the factor limiting roe deer population size

    Nature

    (1977)
  • B. Bobek

    A model for optimization of roe deer management in Central Europe

    J. Wildl. Manag.

    (1980)
  • Buss, H. 1994. Gutachten zum Einfluß des Schalenwildes auf das waldbauliche Betriebsziel. Revier Holzerath, Unpublished...
  • A.M.H. Byron

    Distribution patterns of roe deer (Capreolus capreolus) related to the availability of food and cover

    Notes Mamm. Soc.

    (1981)
  • A. Danilkin et al.

    Behavioural Ecology of Siberian and European Roe Deer

    (1996)
  • M.L. Donovan et al.

    Use of geographic information systems to develop habitat suitability models

    Wildl. Soc. Bull.

    (1987)
  • H. Ellenberg

    Zur Populationsökologie des Rehes (Capreolus capreolus) in Mitteleuropa

    Spixiana

    (1978)
  • B. Fruzinski et al.

    Demographic processes in a forest roe deer population

    Acta Theriol.

    (1982)
  • B. Fruzinski et al.

    Habitat, density, and spatial structure of the forest roe deer population

    Acta Theriol.

    (1983)
  • J.M. Gaillard et al.

    Early survival in roe deer: causes and consequences of cohort variation in two contrasted populations

    Oecologia

    (1997)
  • J.M. Gaillard et al.

    Body mass of roe deer fawns during winter in two contrasting populations

    J. Wildl. Manag.

    (1996)
  • J.M. Gaillard et al.

    Roe deer survival patterns: a comparative analysis of contrasting populations

    J. Anim. Ecol.

    (1993)
  • J.M. Gaillard et al.

    Effects of age and body weight on the proportion of females breeding in a population of roe deer (Capreolus capreolus)

    Can. J. Zool.

    (1992)
  • N.J. Gotelli

    A Primer of Ecology

    (1995)
  • J.T. Heinen et al.

    Simulating the effects of clearcuts on deer habitat in the San Juan National Forest

    Can. J. Rem. Sens.

    (1984)
  • H. Herbold

    Reactions of roe deer to human disturbance (in German)

  • Cited by (46)

    • Investigating spatial non-stationary environmental effects on the distribution of giant pandas in the Qinling Mountains, China

      2020, Global Ecology and Conservation
      Citation Excerpt :

      Understanding of species-environment relationships in heterogeneous landscapes is essential for biodiversity conservation and landscape management (Wu et al., 2000; Turner and Tjørve, 2005). Conventional regression models, such as multivariate linear regression (Radeloff et al., 1999), ordinary least-squares regression (Coppolillo, 2000), and logistic regression (Pereira and Itami, 1991; Augustin et al., 1996), are the most common statistical methods used for estimating the ecological relationships between organism and environmental covariates (Windle et al., 2009; Tavernia and Reed, 2012). These models postulate that the relationships between species’ measures and environmental covariates are spatially stationary over space, and produce constant regression coefficients through the studied area (Fotheringham et al., 1996, 2002; Mcnew et al., 2013).

    • Habitat mapping as a tool for water birds conservation planning in an arid zone wetland: The case study Hamun wetland

      2016, Ecological Engineering
      Citation Excerpt :

      In this study, application of time series data to investigate suitable habitat change was introduced as a tool for water birds habitat management, in a wetland with water limitation and high risk of degradation located in a semi-arid area. Previous studies have identified habitat assessment as an indispensable component of conservation programs (Radeloff et al., 1999; Reza et al., 2013; Yamada et al., 2003). Ma et al. (2010) highlighted the significance of waterbird habitat evaluation in wetland ecosystem management.

    • Population control based on abundance estimates: Frequency does not compensate for uncertainty

      2014, Ecological Complexity
      Citation Excerpt :

      Model calculations and results used density values expressed as the number of individuals per km2. Initial variables and parameters were set according to published data (Andersen, 1953; Heurich et al., 2012; Melis et al., 2009; Pegel and Thor, 2000; Radeloff et al., 1999 [Table 1]; Apollonio et al., 2010b; Borkowski et al., 2011; Eiberle, 1972; Flajšman et al., 2013; Gaillard et al., 1992, 1997; Heurich et al., 2012; Hewison et al., 1999; Jędrzejewska et al., 1997; Pegel and Thor, 2000; Simard et al., 2013; Smart et al., 2004; Stubbe, 1963; Vreugdenhil et al., 2007 [Table 2]). We assume that migration and emigration are in a steady state as the area of concern is neither a source nor a sink habitat patch.

    • Bundling ecosystem services in Denmark: Trade-offs and synergies in a cultural landscape

      2014, Landscape and Urban Planning
      Citation Excerpt :

      Hunting was the service least correlated with other services; thus it did not seem to be in conflict with any other services we measured and could coexist with other land uses in the area. The weak positive correlation to forest carbon storage was logical since forests and small wooded biotopes are important refugees for roe deer (Radeloff, Pidgeon, & Hostert, 1999). The indicator for sense of place was positively correlated with those for nature appreciation and tourism, exemplifying a win–win situation for landscape development with economic incentives for protecting localities important for cultural and natural heritage (Schroth & McNeely, 2011).

    • Integrating GIS and expert judgment in a multi-criteria analysis to map and develop a habitat suitability index: A case study of large mammals on the Malayan Peninsula

      2013, Ecological Indicators
      Citation Excerpt :

      These are the constraints faced by conservationists, planners and even scientists or researchers working in relation to the conservation planning of wildlife and systems of protected areas. In conservation planning and biodiversity management research, the development of wildlife habitat suitability indices and maps is considered a crucial tool to address biodiversity loss and habitat degradation (Radeloff et al., 1999; Yamada et al., 2003). To produce such indices and maps one needs to depend on the availability of ecological and biological data of wildlife.

    View all citing articles on Scopus
    View full text