Coastal forests of Tanzania are diverse in plant species that make them included as part of the 34 world biodiversity hotspots. It was aimed at determining plant species diversity, richness, and evenness and to identify the parameter that best defines plant species diversity of the coastal forests. Transect method was used for data collection; analysis of variance and multiple regression were used to analyze the vegetation data. The plant species diversity ranged from 2.26 to 2.77 in Kazimzumbwi, 2.31 to 2.46 in Pande, and 1.76 to 2.48 in the Zaraninge Forest that was significantly lower than those from other forests. Regardless of high species diversity in Kazimzumbwi it was recorded the lowest plant species evenness (0.485 to 0.490) and the difference of values among forests was not significant. The diversity was strongly positive correlated with both evenness and richness whereas perfect positive correlation (r =1) was observed with evenness and strong positively correlation existed with species richness in Zaraninge (r = 0.88), Pande (r = 0.91) and Kazimzumbwi forest (r =0.79). This implies that richness and evenness portrays different ecological interpretation and cannot be used interchangeably to describe the biodiversity value of the coastal forest ecosystem. Regression models showed that evenness significantly influenced the plant species diversity, whereas richness had insignificant influence. It can be concluded that the regression model is suitable to predict the trend of change in plant species diversity and evenness is the best predictor and an adequate measure of the coastal forests’ conservation value than richness.