Island ecosystems are characterized by self-containment and vulnerability due to their limited and isolated areas. They are also important for biodiversity maintenance, freshwater conservation, and other basic ecological functions. Moreover, island ecosystems play important roles in global carbon cycle, which makes the estimation of net primary productivity (NPP) more significant for ecological construction and the study of carbon source and sink on islands. The rapid development of RS and GIS technologies provides a practical method for NPP estimation. In this study, five southern islands of Miaodao Archipelago were examined and an NPP estimation model was established based on the Carnegie-Ames-Stanford Approach (CASA) and regional characteristics. To characterize each season, 4 temporal remote sensing images were utilized. The normalized difference vegetation index (NDVI) and land cover types were extracted, and these constituted the data used for the NPP estimation model, along with meteorological and topographical data. The NPP values for different seasons, islands, and land cover types were calculated using GIS methods. The linear correlations between NPP values and various parameters including NDVI, altitude, slope, aspect, and land cover types were obtained to determine the importance of each parameter. Finally, the relationships between 10 soil environment-chemical properties and NPP at 36 sampling sites were analyzed. The total NPP for the five islands was 11043.52 t C/a. The mean density of NPP (MDN) was 340.19 g C m^-2 a^-1, which was higher than that of western regions at a similar latitude (e.g., Gansu Province and Qinghai Province) and lower than that of coastal zones in the eastern region (e.g., Guangdong Province, Jiangsu Province, and Panjin City in Liaoning Province). NPP in the summer accounted for approximately 80% of the total yearly NPP, followed by spring (11%), autumn (7%), and winter (1.3%), indicating that NPP in the northern area differed greatly among seasons. The total NPP was highest for Nanchangshan Island, followed by Beichangshan Island, Daheishan Island, Miao Island, and Xiaoheishan Island, and the total NPP was positively correlated with island area. MDN was highest for Daheishan Island, followed by Beichangshan Island, Miao Island, Nanchangshan Island, and Xiaoheishan Island, and was negatively related to the proportion of construction land. The land cover types included broad-leaf forest, coniferous forest, farmland, grassland, construction land, and bare land, in descending order based on MDN. The high NPP for forest land indicated that the plantations on the five southern islands were of great ecological significance. NDVI and land cover types were the most important parameters with respect to the spatial distribution of NPP, and topographical parameters indirectly affected NPP via their effects on NDVI and land cover type. NPP had a significant negative relationship with soil pH (P < 0.01), available phosphorus (P < 0.01), total phosphorus (P < 0.01), and total potassium (P < 0.05), and a significant positive relationship with total carbon (P < 0.01), total organic carbon (P < 0.01) and total nitrogen (P < 0.05). No significant relationships with water content, available potassium, or salinity were observed.