By multi-temporal analysis of multispectral RapidEye data sets visible spectrally defined spatial units are distinguished within an agricultural field in stable pattern (due to differences in soil properties influenced) (ground pattern) and temporary pattern (created by management, vegetation and weather). Identified patterns are analyzed with respect to their soil characteristics and assessed their spatial and temporal stability. The aim is to develop functional maps based on spatially and temporally stable soil patterns for a more economical and more sustainable land management.

In order to cover the entire geomorphological, pedological and topographical range of the project area, on several reference areas in total 731 in-situ surface soil samples are taken and were analyzed in the laboratory with regard to their physiochemical soil properties. 

The method of the mulitemporal soil pattern analysis for field-specific modeling of organic matter consists of the following steps:

(1)   selection of the most appropriate data (bare soil images) from the satellite data time series,

Method: Data Mining using NDVI thresholds and phenological data of cash crops

(2)   detection of soil reflectance pattern

Method: image mining using standardized principal component analysis (PCA Standardized)

(3)   analysis of the spatio-temporal stability of soil samples,

Method: pixel-based change detection

(4)   preparation of soil function maps based on statistical analyzes for creating the prediction model and the progressive elimination of temporary effects (eg management, vegetation).

Method: regression analysis, percentile analysis

Detailed explanations of the methodology and preliminary results provided in our publication "multitemporal soil pattern analysis with multispectral remote sensing data at the field-scale" are removed (Blasch et al. 2015): 

Blasch, G., Spengler, D., Hohmann, C., Neumann, C., Itzerott, S., Kaufmann, H. (2015): Multitemporal soil pattern analysis with multispectral remote sensing data at the field-scale. - Computers and Electronics in Agriculture, 113, p. 1-13.

The method has been transferred from field scale to landscape scale for the project area Demmin. For all reference test areas a joint prediction model was established for organic matter on the inclusion of all 731 soil samples and the common PCst1 . This predictive model was transferred to the detected ground pattern (PCst1) of all bare soil fields in the project area "Demmin". Therfore a synthetic data set of bare soil fields, selected from different images have been used. Figure 3 shows the finally created soil function map of Demmin.

Blasch, G., Spengler, D., Itzerott, S., Wessolek, G. (2015): Organic Matter Modeling at the Landscape Scale Based on Multitemporal Soil Pattern Analysis Using RapidEye Data. - Remote Sensing, 7, 9, p. 11125-11150.