Scaling Properties of the Spatial Heterogeneity of Leaf Area Index on Arable Land Derived fro Remote Sensing and Simulation

Abstract:

The leaf area index (LAI) strongly influences mass and energy exchange processes of vegetated surfaces. Therefore, LAI is a crucial input variable in many large scale land surface schemes. Nonlinear dependencies of processes upon LAI require to account for its spatial heterogeneity.

LAI data derived from remote sensing (RapidEye, resolution 5 m) and simulated with the ecohydrological simulation system DANUBIA (resolution 150 m) were analyzed for the fertile loess plain (1100 km²) of the Rur catchment in Germany for the year 2011. Spatial heterogeneity is described by relative frequency distributions (spatial variability) and ranges of semivariograms (spatial structure).

Despite the different resolutions, model results of LAI are generally very similar to remote sensing data. Relative frequency distributions (RFD) of LAI can be explained by the superimposition of RFDs of the main crops’ LAI. While upscaling (by averaging) does not change the spatial mean LAI, it reduces the standard deviation. At spatial resolutions less than 300 m, the RFDs of LAI change from asymmetric multi-peak to symmetric single-peak shapes.

Spatial structure in terms of theoretical semivariogram ranges shows two characteristic lengths of 100 to 300 m and 2 to 20 km corresponding to the dimensions of agricultural fields and soil properties. Information on the shorter lengths is conserved during upscaling down to resolutions of about 500 m.

Results show that for the test area, information required to comprehensively describe spatial heterogeneity (spatial variability and structure) of LAI is not conserved at spatial resolutions less than 300 m.