Multitemporal soil pattern analysis with multispectral remote sensing data at the field-scale

Mono-versus multitemporal approach using RapidEye time-series was tested.Due to NDVI and phenology data bare soil images were selected from time-series.Principal component analysis of bare soil images allows detecting soil pattern.Multitemporal approach permits to evaluate the stability of detected soil pattern.A static soil pattern-based functional soil map for organic matter was generated. This research proposes a new model for the generation of basic soil information maps for precision agriculture based on multitemporal remote sensing data analysis and GIS spatial data modelling. It demonstrates (i) the potential of multitemporal soil pattern analysis (ii) to generate functional soil maps at field scale based on soil reflectance patterns and related soil properties and (iii) how to improve these soil maps based on the identification of static homogenous soil patterns by excluding temporal influences from the developed prediction model. Principal components and per-pixel analyses are used for the separation of static soil pattern from temporal reflectance pattern, influenced by (vital and senescent) vegetation and land management practices. The potential of the proposed algorithm is investigated using multitemporal multispectral RapidEye satellite imagery at a demonstration field "Borrentin" field in Northeast Germany.