In order to implement scientific assessment of heavy metal pollution of farmland soils in regions well developed in industry, a modified empowered fuzzy comprehensive evaluation model was established based on the concept and fuzzy nature of risk, and moreover, analysis was done of factors affecting the risk. The model consists of a target layer, a criterion layer and an index layer. The target layer is the target object for evaluation, that is, risk of a soil getting polluted by heavy metals. The target layer is composed of 3 criterion layer indexes (tendency, vulnerability and hazardness). The index layer of tendency covers accumulation rates of heavy metals (Hg, Cd, Cr, Pb, As, Cu, Zn and Ni); the index layer of vulnerability consists of pH, organic matter, soil texture; and the index layer of hazardness refers to potential ecological risk index. Weights of the criterion layer are obtained with the analytic hierarchy process, while comprehensive weights of the indices in the index layer are determined with the group decision making analytic hierarchy process (the subjective weight) and the entropy weight method (the objective weight). Risk level is determined with the fuzzy comprehensive evaluation results vector. Validity of the maximum membership degree law is tested in the process of determining risk level. The weighted average method is used to determine risk level when the validity is low. Priority order of the risk management of various indices can be determined according to the importance degree and risk degree of each index. A case study was carried out of the Pearl River Delta Region. Results show that the principle of maximum membership degree is inefficient in Guangzhou, Dongguan, Zhongshan and Jiangmen, with risk grade value being 2.8, 2.5, 2.4 and 2.4, respectively; and the principle of maximum membership degree is valid in Shenzhen, Zhuhai, Foshan and Huizhou, with maximum membership degree being 0.50 (third-level), 0.47 (second-level), 0.46 (third-level) and 0.63 (second-level), respectively, both suggesting that Guangzhou, Shenzhen, Foshan and Dongguan, as a result of high human disturbance, are high in risk, reaching up to Grade Ⅲ or "moderate risk", while Zhuhai, Huizhou, Zhongshan and Jiangmen, relatively low in human disturbance are in Grade Ⅱ or "slight risk". Analysis of affecting factor indicates that major risk affecting factors vary with the administrative region. All the findings are expected to serve as scientific basis for designing risk management strategies and defining a priority order of management.