Automated keys of soil diagnostic horizons based on case-based reasoning

Soil diagnostic horizon (SDH) in Chinese soil taxonomy (CST) is a representative feature for one soil type, the determination of SDH type is critical for soil classification. Usually, it is done by experts of soil classification with key rules in CST, but the amount of expert is far insufficient for SDH determination in future soil survey. Hence this paper presents a case-based reasoning (CBR) method for automated keys of SDH to improve the efficiency of soil classification and solve the problem of shortage of soil classification expert. Firstly the case base, as a foundation for automated keys of SDH, was established throughout the splitting and reorganizing of key rules in Chinese Soil Taxonomy (3rd edition). Then the algorithms for computing the similarity between case and fact were also established based on the theory of CBR. Finally the automated keys of SDH were implemented with fact data of soil profile. The result shows as follows: (1) The quality of case base of SDH is excellent and can be used for automated keys of all kinds of SDH in China. Moreover, the structure of case base can be used as the template for recording soil profile information in future soil survey. (2) The structure of case base affects reasoning accuracy. The reasoning accuracy with case base of single table is very low because of disturbing of the irrelative data field, and this can be improved by splitting single table off to multiple tables (each of them represents the key rules of one SDH type). (3) The weight setting of case attributes also affects reasoning accuracy. The accuracy can be improved through reducing the weight of case attributes possessed by all types of SDH. (4) The automated keys of SDH with CBR method can significantly improve efficiency of SDH determination, and can still obtain reasoning result under the circumstance of partially data missing, but it still can not replace the expert´s judgement completely.