Abstract: According to the theory of system science, this paper established an evaluation system of forest health including 13 indicators from system vitality, composition and resilience and forest environment aspects. To solve the problem in which the health evaluation values are unknown, that is, no decision attribute, this paper applied attribute importance degree of rough set theory to analyze the indexes’ influence on forest health, with the forest health evaluation index weight being obtained based on knowledge granularity and attribute importance. We took eight secondary forest communities and two plantations in Daweishan Nature Reserve of Hunan Province as study cases. The results showed that the weights of 13 indexes were basically consistent with the expert scoring results. The evaluation results indicated that only one plot was at health level, seven at subhealth level and two at unhealthy level. Compared with the analytic hierarchy process method (AHP) and the principal component analysis method (PCA), this method did not need any prior information but took the information related to the forest health system as the basis to mine correlation and importance degree among factors directly reflecting forest health. It did not depend on experts’ experience, and the greater the amount of information, the more objective the attribute weight was. The results suggested that the evaluation method was real and effective and the evaluation result was more objective, providing a theoretical basis for forest health assessment and management.

Key words: late rice, heading and flowering stage, irrigation, yield, low temperature, physiological characteristics.