3. Data analysis
After the data is collected it is analyzed to find possible spatial and temporal relationships. When these relationships are found they can be visualized to gain more insight in the relationship.
3.1. Spatial relationships
As a first indication of a possible (linear) relationship between two images, which contain different types of information, the correlation coefficient between these images can be calculated. First the images are resampled such that the image with the largest resolution is left intact. The other image is resampled such that the overlapping area of the 2 images becomes a grid which is equal to the corresponding grid of the image which is left intact. This resampling is performed using interpolation according to the nearest neighbour principal. This form of interpolation is used to keep the original measured quantities intact.
The correlation coefficient provides an indication of the amount of noise and the direction of a linear relationship between the two images, but it does not provide the slope of this relationship or any information on the possible existence of a non-linear relationship between these images. The correlation coefficient for all possible pairs of available images is calculated.
The pairs of images which have a high indication of the existence of a linear relationship can be investigated further by performing a simple linear regression analysis. This can be used as a preliminary quantization of the linear relationship between the two images, or equivalently between the two variables contained in these images.
For the images in figures 1a and 1b such a preliminary quantization of the linear relationship performed. The result of this quantization is shown in figure 1c. Figure 1a shows an image of the electric conductivity measurements carried out on the parcel. The measurements are interpolated to obtain the shown raster data. Figure 1b shows the height map of the parcel.
The height map shows that the north side of the parcel is higher than the south side. This can be traced back to the presence of an old river dune on the north side of the parcel. This old river dune consists of finely sand with clay on top. This layer of clay becomes increasingly thicker to the south side of the parcel until eventually the soil can be classified as heavy sandy clay. The lowest conductivity is found in the old river dune, as can be seen in figure 1a. Here, the texture is different (coarser) and the ground is relatively high. As a result of this relative high position, and the coarser texture, this part of the field is relatively dryer. To further investigate the relation between the height map and the electric conductivity a simple linear regression between the electric conductivity and the freeboard is performed. The freeboard is the difference between the height map and the local surface water level in the polder. This resulted in a quantization of the linear relationship, see figure 1c. This quantization can be used to further analyze the relationship between the two images.