Methods

We will start by visualizing the aforementioned datasets to obtain an understanding of the data as a whole. General trends and first observations about spatial distribution of each dataset will be noted and we will then try to find relevant relations between them. In partical we will try to evaluate the correlation between vegetation and ground temperature.

Results

The  Fig \ref{931769}. shows a map of NDVI spatial distribution according to natural breaks classes. It shows a red zone in the city center at the lake's end which is probably completely built-up and contains no green areas at all. Now, the question raised is : does this area show particularly high values for the thermal data ? This was precisely the case: all the areas which have values of NDVI inferior to 0.180 (two lowest classes) correspond to the category of highest themal values for the same classification (Natural breaks map with 4 classes). In order to know if this relation between the two datasats could be generalized and remains in non extreme values, we made a LISA scatter plot (visible in the Fig \ref{237263}). which allows us to analyze the spatial correlation between the variables of interest. The negative slope (Moran's I equal to -0.56 approximately) expresses the negative correlation between NDVI and thermal data. Regarding the contiguity, we used a queen weighting scheme of order 1. Moreover a randomization of 999 permtutations gave a p-value of 0.001 so the results can be considered really significant.