My purpose is to map hotspots using LISA tools (or in general local statistics). By developing GIS application using available spatial statistic libraries (i.e. PySAL - ESDA), I have still not found documented methods (or suitable literature) to include in the computation more than two variables.
Have you any suggestions on how to map simultaneously hotspot or significant spatial clusters of more than two variables?
In the following link one of the papers (Anselin et al. 2002) I have considered as reference
http://europa.nvc.cs.vt.edu/~ctlu/Project/SoftwareTools/multi_lisa.pdf
The usdm R packkage has the LISA function. This function generates a map in which the locations (points) show the magnitude of LISA from variables.
Reference:
Naimi, B., Hamm, N.A.S., Groen, T.A., Skidmore, A.K., and Toxopeus, A.G. 2014. Where is positional uncertainty a problem for species distribution modelling, Ecography. 37(2), 191-203.
Very nice library, thank you Oliver!
The hybrid use of bot raster and vectors objects to include different variables is really interesting. I will strongly consider it.
Daniele
You could use other clustering alghoritms which actually are multivariables, just like AZP or its variations. I hardly recommend you clusterPy (http://www.rise-group.org/risem/clusterpy/).
Here is a roughly review of regionalization and clustering methods:
Duque, J. C.; Ramos, R. & Suriñach, J. Supervised regionalization methods: A survey International Regional Science Review, Sage Publications, 2007, 30, 195-220
To maps hotspots considering more than two variables, you could to use Antonio Velázquez, Luis Manuel Martínez and Fátima Maciel Carrillo 'Caracterización climática para la región de Bahía de Banderas mediante el sistema de Köppen, modifiado por García, y técnicas de sistemas de información geográfia' method. Miquel Ninyerola, Xavier Pons and Joan M. Roure ' A methodological approach of climatological modelling of air temperature and precipitation through GIS techniques' method.
Dear Antonio;
Really interesting work, even if what I was looking for a way to escape from the use of regression. In any case this is a valuable starting point to further the hotspot analysis toward the multivariate case!
Regards
Daniele
To whom it may concern,
I have found a recent and interesting preliminary work on the extension of the Geary's C statistic to the multivariate case. I am working on this as it promises to be the closest solution to the traditional univariate clusters/outliers mapping.
Working Paper A Local Indicator of Multivariate Spatial Association
- Badges
- Science topic
- Similar topics
- Geoscience
- Geoinformatics
- Geographic Information System