The selection of certain indicators is critical to undertake ecological risk assessments of long-term oil pollution and other environmental changes. The indicators should be easily and routinely monitored, be sensitive to pollution, respond to pollution in a predictable manner, and match the spatial and temporal scales of investigations. To compare the effectiveness of indicators for the long-term risk assessments, this study investigated the multiple ecological effects of chronic oil pollution on the plant community dominated by reed (Phragmites australis). The physiology, growth and reproduction of reed, together with the composition and productivity of the reed community, were measured around oil wells that have operated for approximately 10 years in the Yellow River Delta, eastern China. The predictive power of each indicator was evaluated using the coefficients of determination (R2) of linear regression models established for each indicator and soil Total Petroleum Hydrocarbons (TPH) concentration. The sensitivities of indicators were evaluated by comparing slopes of new established regression lines using standardized data. The top three indicators in terms of predictive power were leaf length, width and number, followed by the Shannon-Wiener index, Pielou evenness index and Simpson's diversity index. Community aboveground biomass, foliar projective coverage and species richness showed predictive power lower than those of the three diversity indexes, but higher than those of leaf net photosynthetic rate, reed height, aboveground biomass and vertical projective coverage of reed plants. Leaf transpiration, chlorophyll concentration and reed stem density showed no significant linear response to elevated soil TPH concentration. In terms of sensitivity, the top three biological indicators were Pielou evenness index, Simpson's diversity index and Shannon-Wiener index, followed by community vertical projective coverage, community aboveground biomass, and species richness. Leaf number, length and width were moderately sensitive, followed by reed coverage, aboveground biomass and height. The sensitivity of net photosynthetic rate was the lowest. The predictive power and sensitivities of indicators were compared in terms of their spatial and temporal scales. In conclusion, scale can be used to facilitate the selection of indicators, and the combination of different indicators may yield improved understanding of the various effects of elevated soil TPH concentration at the different biological levels.