Predicting species occurrence using a modelling approach based on a geographic information system (GIS) represents a new methodological tool which can be used to endorse conservation policies, on condition that models are tested for reliability. Habitat suitability models are often used to predict species occurrence through the modelling of proper environmental variables. A major constraint in building large-scale models of species distribution is the availability of data and therefore the deductive approach is adopted. This approach is suitable for bio-diversity assessment as it can be applied to a great number of species, it does however require statistical validation, both to test the accuracy of the input information and to deal with presence data (often affected by high spatial and temporal variability). Available data are, in the large majority, represented by a small sample collected without an ad hoc sampling design and without information on species absence. Reliable information on absence is not easily obtained with animal species due to elusive behaviour, poorly accessible habitats and their activity patterns. Therefore, absence data are to be considered ambiguous. Hence the need to develop statistical tools to deal with presence-only data. Moreover, presence data are rarely geo-referenced with high precision. Often information on species presence is spatially degraded as referred to polygon within which the exact species location cannot be recovered. Consequently, the validation procedure should statistically consider this uncertainty. In this paper we suggest that a compositional procedure be used when presence-only data are affected by strong spatial and temporal variability, and that a multinomial (based) procedure be used when presence-only data are geo-referenced with high precision during an ad hoc field survey. The two procedures have been tested on a large number of deductive models built to represent the distribution of vertebrate species in Italy. Results show that both procedures allow the assessment of the calibration properties of the models. The reliability of the two methods is confirmed by a model rejection pattern, determined by ecological factors, scale issues, spatial accuracy and sample size. Both procedures have a wide range of applications being threshold independent. (C) 2004 Elsevier B.V. All rights reserved.

Two statistical methods to validate habitat suitability models using presence-only data / Daniela, Ottaviani; JONA LASINIO, Giovanna; Boitani, Luigi. - In: ECOLOGICAL MODELLING. - ISSN 0304-3800. - STAMPA. - 179:4(2004), pp. 417-443. [10.1016/j.ecolmodel.2004.05.016]

Two statistical methods to validate habitat suitability models using presence-only data

JONA LASINIO, Giovanna;BOITANI, Luigi
2004

Abstract

Predicting species occurrence using a modelling approach based on a geographic information system (GIS) represents a new methodological tool which can be used to endorse conservation policies, on condition that models are tested for reliability. Habitat suitability models are often used to predict species occurrence through the modelling of proper environmental variables. A major constraint in building large-scale models of species distribution is the availability of data and therefore the deductive approach is adopted. This approach is suitable for bio-diversity assessment as it can be applied to a great number of species, it does however require statistical validation, both to test the accuracy of the input information and to deal with presence data (often affected by high spatial and temporal variability). Available data are, in the large majority, represented by a small sample collected without an ad hoc sampling design and without information on species absence. Reliable information on absence is not easily obtained with animal species due to elusive behaviour, poorly accessible habitats and their activity patterns. Therefore, absence data are to be considered ambiguous. Hence the need to develop statistical tools to deal with presence-only data. Moreover, presence data are rarely geo-referenced with high precision. Often information on species presence is spatially degraded as referred to polygon within which the exact species location cannot be recovered. Consequently, the validation procedure should statistically consider this uncertainty. In this paper we suggest that a compositional procedure be used when presence-only data are affected by strong spatial and temporal variability, and that a multinomial (based) procedure be used when presence-only data are geo-referenced with high precision during an ad hoc field survey. The two procedures have been tested on a large number of deductive models built to represent the distribution of vertebrate species in Italy. Results show that both procedures allow the assessment of the calibration properties of the models. The reliability of the two methods is confirmed by a model rejection pattern, determined by ecological factors, scale issues, spatial accuracy and sample size. Both procedures have a wide range of applications being threshold independent. (C) 2004 Elsevier B.V. All rights reserved.
2004
distribution modelling; distributions; ecology; habitat suitability models; landscapes; logistic-regression; maps; null model; populations; prediction; predictions; presence-only data; spatial accuracy; species distribution; validation
01 Pubblicazione su rivista::01a Articolo in rivista
Two statistical methods to validate habitat suitability models using presence-only data / Daniela, Ottaviani; JONA LASINIO, Giovanna; Boitani, Luigi. - In: ECOLOGICAL MODELLING. - ISSN 0304-3800. - STAMPA. - 179:4(2004), pp. 417-443. [10.1016/j.ecolmodel.2004.05.016]
File allegati a questo prodotto
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/238243
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 81
  • ???jsp.display-item.citation.isi??? 74
social impact