University of Twente Student Theses
Hyper temporal NDVI images for modelling and prediction the habitat distribution of Balkan green lizard (Lacerta trilineata): Case study: Crete (Greece)
Taheri, Shirin (2010) Hyper temporal NDVI images for modelling and prediction the habitat distribution of Balkan green lizard (Lacerta trilineata): Case study: Crete (Greece).
|
PDF
10MB |
| Abstract: | In the study of species environment relationship, it would be desirable to find predictors those can better explain the habitat characteristics of the species. One of the main problems in habitat distribution modelling is missing covariates (i.e. limiting predictors), reflecting lack of knowledge which environmental factor constrain the distribution of species. This study aims to model the temporal pattern of vegetation dynamic, to explain the geographical distribution of the reptile species. The Balkan green lizard (Lacerta trilineata) in Crete Island, Greece was selected as the case study. The ISODATA clustering algorithm was used to recognize the temporal vegetation dynamic pattern. The separability divergence method was employed to identify the appropriate number of classes based on spatio-temporal patterns of vegetation. The 81 observation points of the species occurrences were provided from Natural History Museum of Crete and field work data collection. Classified hyper-temporal NDVI in addition with other environmental variables (e.g. climate data, soil, etc.) were used as biophysical features of the ecosystem. The predictability of classified hyper-temporal NDVI images was investigated by comparing the performance of models based on different combination of variables. For each predictor variable alone, regularize training gain calculated to see the drop in gain when the variable is omitted from the full model. The results show that the classified hyper temporal NDVI can significantly improve the predictability of the model. In all developed models, hyper-temporal NDVI emerged as the most important predictor. The result indicates when NDVI is omitted from the model, the training gain significantly decrease, which suggests it contained the most useful information that are not presented in the other variables. The result indicates strong relationship between some NDVI classes and high probability occurrence of the species. Considering these results the probability of species occurrence was highest in sites, where shrubs and rocks were dominant, or in old olive plantations and abandoned agriculture. |
| Item Type: | Essay (Master) |
| Faculty: | ITC: Faculty of Geo-information Science and Earth Observation |
| Programme: | Geoinformation Science and Earth Observation MSc (75014) |
| Link to this item: | https://purl.utwente.nl/essays/90730 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page