Comparing Discriminant Analysis, Ecological Niche Factor Analysis and Logistic Regression Methods for Geographic Distribution Modelling of Eurotia ceratoides (L.) C. A. Mey

Document Type: Research and Full Length Article

Authors

1 Department of Rangeland Management, Rangeland and Watershed Management Faculty, Gorgan University of Agricultural Sciences & Natural Resources, Gorgan,

2 Department of Rehabilitation of Arid and Mountainous Regions, University of Tehran

3 Department of GIS and Remote Sensing, Geography faculty, Tarbiat Modares university,

Abstract

Eurotia ceratoides (L.) C. A. Mey is an important plant species in semi-arid lands
in Iran. New approaches are required to determine the distribution of this plant species. For
this reason, geographical distributions of Eurotia ceratoides were assessed using three
different models including: Multiple Discriminant Analysis (MDA), Ecological Niche Factor
Analysis (ENFA) and Logistic Regression (LR). The study area was located in northeast
rangelands of Semnan, Iran. Sampling was performed in each vegetation type using
randomized-systematic method. Vegetation data in addition to environmental factors' data
such as topography and soil were prepared. The MDA and LR methods were performed with
SPSS software as predictive modelling methods based on presence and absence data. The
ENFA model was performed by the means of necessary statistical analysis in Biomapper
(Version 4.0) software only by presence data. The plant predictive mapping needs the maps of
all effective factors based on model parameters. Mapping of soil characteristics was done by
geo-statistical method. The accuracy of the predicted map was tested with the actual
vegetation map. Predictive maps of E. ceratoides (based on the LR and MDA methods) with
Kappa coefficients as 0.56 and 0.64 had a good accordance with actual vegetation map
prepared for the study area. Kappa coefficient of potential habitat map (based on ENFA
method) of E. ceratoides was 0.85; hence, it had a very good accordance. The results obtained
by all methods showed that this species is distributed in the rangeland with pH as 7.8-8, EC as
0.17-0.26 dc/m and silty-sandy texture in 1600-2200 m elevation. Organic matter in the depth
of 20-80 cm and pH in the depth of 0-20 cm did not significantly influence the differences.
Minimum sampling is needed using these methods which provide worth while data about the
presence of the plant species in the other places.

Keywords