The Case of Arthropods on Terceira Island
Total Page:16
File Type:pdf, Size:1020Kb
Ann. Zool. Fennici 46: 451–464 ISSN 0003-455X (print), ISSN 1797-2450 (online) Helsinki 18 December 2009 © Finnish Zoological and Botanical Publishing Board 2009 Species distribution models do not account for abundance: the case of arthropods on Terceira Island Alberto Jiménez-Valverde1,2,*, Francisco Diniz1, Eduardo B. de Azevedo3 & Paulo A. V. Borges1 1) Azorean Biodiversity Group (CITA-A), Universidade dos Açores, Dep. de Ciências Agrárias, Terra-Chã, 9700-851 Angra do Heroísmo, Terceira, Açores, Portugal (*corresponding author’s e-mail: [email protected]) 2) Present address: Natural History Museum and Biodiversity Research Center, The University of Kansas, Lawrence, Kansas 66045, USA 3) CMMG (Centro de Estudos do Clima, Meteorologia e Mudanças Globais), Universidade dos Açores, Dep. de Ciências Agrárias, Terra-Chã, 9700-851 Angra do Heroísmo, Terceira, Açores, Portugal Received 27 Feb. 2009, revised version received 10 May 2009, accepted 22 May 2009 Jiménez-Valverde, A., Diniz, F., de Azeved, E. B. & Borges, P. A. V. 2009: Species distribution models do not account for abundance: the case of arthropods on Terceira Island. — Ann. Zool. Fennici 46: 451–464. The objective of this study is to investigate whether presence/absence models can be used as surrogates of arthropod abundance, and eventually under which circumstances such surrogacy is guaranteed. Presence/absence data for 48 arthropod species from Terceira Island were modelled using artificial neural networks. Probabilities of occur- rence were correlated with abundance data from a standardized arthropod survey pro- gramme. Although a tendency was found for vagile species to show relationships, only nine species showed significant positive correlations between probability of presence and abundance. Five of these were exotic spider species with high abundances and wide distributions in several human-modified habitats. The patchy distribution of pris- tine habitats, the capacity to reach them and the probable low dependence on limiting resources, other than food, enhance the relationship. A lack of significant correlations for the majority of the species may be due to historical events, inappropriate scale, demographic controls of density, or the incapacity of presence/absence models to account for environmental suitability. The difficulty to identify a priori the species for which the relationship will hold advises against the use of species distribution models as surrogates of arthropod abundance. Introduction spatial and temporal patterns of distribution and abundance is critical to assess the threat status of Species occurrence and abundances are key vari- species (Regan et al. 2000, Conrad et al. 2006). ables in modern ecological and conservation The abundance of a species is ultimately a func- sciences (see Gaston & Blackburn 2000 and tion of its birth and death rates, which depend references therein). Moreover, understanding on its fitness in different environments (Boyce 452 Jiménez-Valverde et al. • ANN. ZOOL. FENNIcI Vol. 46 & McDonald 1999, Pearce & Ferrier 2001, Tyre ship will be enhanced if resources are limiting et al. 2001, but see Van Horne 1983). Besides, and patchy and if the densities of the populations abundance is closely related to persistence, a are at their upper limit, as at low densities, other property of the population that is desirable to factors apart from environment may be determin- maximize in reserve selection schemes (Mangel ing the size of the populations (Mitchell 2005, & Tier 1994, Araújo & Williams 2000). It fol- Nielsen et al. 2005). However, factors other than lows that predicting population densities across environment will make species reach high densi- its distributional range is of great value for con- ties in places with low probability of occupancy servation and management purposes, in order to (Van Horne 1983, Tyre et al. 2001, Nielsen et al. assess the impact of environmental changes on 2005). For example, biotic interactions such as organisms (Boyce & McDonald 1999, Joseph et aggregation, intraspecific competition or preda- al. 2006, Smith et al. 2006). tor pressure can force individuals to concentrate However, the logistical difficulties of obtain- in unsuitable sink marginal habitats. Thus, the ing abundance data in many locations are partic- relationship between probability of occurrence ularly relevant for arthropods. Many projects are and local abundance is a complex phenomenon currently compiling information on species dis- that depends on many different factors acting in tributions (e.g. GBIF at www.gbif.org, ATLAN- isolation or in synergy. However, little empirical TIS www.azoresbioportal.angra.uac.pt; Borges work has been undertaken to study the capac- 2005, Hortal et al. 2007), but spatial data from ity of occurrence models to account for local these databases have no associated abundance abundance. To our knowledge, only two studies data, which makes them at least incomplete. In specifically address the question: (i) Pearce and fact, presence data are much easier to survey, and Ferrier (2001) tested the capacity of logistic pres- are available in numerous biological databases ence/absence models to predict population densi- (Soberón & Peterson 2004, Hortal et al. 2007). ties of several species of vertebrates and vascular This is one of the reasons why predictive distri- plants, but found little correlation between the bution models have developed rapidly in the last probability of presence and abundance among decade (Scott et al. 2002). By quantifying spe- occupied sites; (ii) likewise, Nielsen et al. (2005), cies–environment relationships and prediction of studying a fern and a moose, obtained disappoint- species’ geographic distributions from confirmed ing results regarding probability of presence as a occurrences, they assume a direct link between surrogate of abundance, attributing their negative the probability of presence and environmen- results to unmeasured factors, such as site history, tal suitability (Pearce & Ferrier 2001, Peterson that ultimately determine abundance. Recently, in 2006). Such models can be used to develop man- a study not focused on this question, Real et al. agement decisions and conservation strategies (2009) found significant relationships between (Schadt et al. 2002, Barbosa et al. 2003, Hirzel environmental favourability and abundance in et al. 2004, Russell et al. 2004), among other two vertebrates in Spain, the Iberian lynx and its practical and theoretical applications (see Peter- prey, the European rabbit. All these results are son 2006 for a review). Their usefulness would inconclusive and no inference about the circum- be greatly strengthened if a relationship between stances, if any, under which probability of occur- local probability of occurrence and abundance rence acts as a surrogate for local abundance can were to exist. Moreover, several authors have be deduced. recently pointed out that relating probability In this study, we use information derived of occurrence to abundance could be a power- from a replicated, intensive arthropod sampling ful way of validating presence/absence models scheme on Terceira Island (Azores) in several (Albert & Thuiller 2008, Lobo et al. 2008). habitats (see Borges et al. 2005a, 2006). We eval- A positive relationship between the prob- uated the degree to which environmental suit- ability of occurrence and local abundance is ability assessed with presence/absence models expected when there is a strong relation between account for abundance estimates. By comparing population densities and environmental variables numerous arthropod species with diverse bio- (Pearce & Ferrier 2001). Moreover, the relation- logical characteristics, we tried to assess the con- ANN. ZOOL. FENNIcI Vol. 46 • Probability of presence and abundance 453 ditions under which the relationship holds. We 2005a, 2006). Data were referred to 500 ¥ 500 m hypothesize that the wider the distributions and UTM grids. In cases where two or more transects the higher the global abundance, the stronger the fell in the same grid (8 cases, 67 final UTM grids relation will be. Thus, we expect the relationship with abundance data), the mean abundance for to be enhanced in the case of vagile species able each species was used. BALA transects provided to colonize all suitable habitats. As occurrence abundance data (Table 1) as well as presence/ models can have two different interpretations absence information. In addition, we collated (potential versus actual distributions), depending presences of each species in 500 ¥ 500 m grids on the kind of distribution data and predictors from the GIS-based ATLANTIS database (see used (Soberón & Peterson 2005, Jiménez-Val- Borges 2005 and www.azoresbioportal.angra. verde et al. 2008), we developed two types of uac.pt) to increase their number for occurrence models to assess if there are any differences in models. their relation with abundance. We chose a 500 ¥ 500 m spatial resolution as we consider it to represent an appropriate balance between the incidences of the fauna Data and methods on BALA transects (both presence/absence as well as relative abundances for each species Biological data among cells) and minimization of the effects of other variables (competition, stochasticity, etc.) For this study we used a large dataset of epigean on population abundances and species presences. arthropod species sampled with pitfall traps in A lower resolution (larger cell size) for such a a standardised, large-scale biodiversity study in small territory