Using auxiliary information to adjust fuzzy membership functions for improved mapping of soil qualities

Agronomic soil management and decision-making frequently requires the joint classification of soil variables. Fuzzy set theory is often used to accomp lish this task. This paper addresses the issues of objectively defining fuz zy membership functions (FMF) and reducing classification uncertainty with hedge operators. As an example, soil in North-east Thailand was classified according to its inherent potential to support the recovery of a rice crop after a drought spell. The utility of auxiliary information not directly in cluded in the classification was explored. A tree cover density index was employed for an objective definition of the FMF to classify soil organic matter content and plant-available potassium. Mapping units were allocated to classes having low, medium or high availabi lity of these plant nutrients. It was shown that crisp, Boolean style class ifications severely misclassify land in all but one class. Adjusted FMF dec reased the uncertainty contained in thematic class maps. Single FMF values for soil organic matter and plant-available K were then jointly modelled an d the soil classified as having low, medium and high potential for rice pla nts to recover from drought impacts. The very and more or less hedge operat ors were applied to increase or decrease the joint FMF values using farmer' knowledge about soil fertility. Overall classification uncertainty using F MF was decreased by 14% if the standard FMF was adjusted and the generated membership values were hedged. It was shown that adjusting FMF influenced t he uncertainty components vagueness and ambiguity differently; the former i ncreased slightly but the latter was drastically reduced.