Abstract: Based on data derived from Landsat multitemporal remote sensing images, the feature index and object-oriented classification were combined to extract the spatial information of mariculture area in Sansha Bay of Ningde, Northeast Fujian. Landscape pattern index and random forest algorithm were used to analyze the spatial and temporal expansion and morphological characteristics of mariculture areas, and to quantify the effects of natural, social, and economic factors on the spatial and temporal variations of mariculture areas. The results showed that object-oriented classification had a good outcome, with the total accuracy and Kappa coefficient being more than 81% and 0.72, respectively. During 2009 to 2018, patch area, patch density, and landscape shape index showed a trend of first increase and then decrease. The spatial aggregation of mariculture area was increasing, displaying the aggregation characteristics of the continuous block distribution from west to northeast. The spatiotemporal dynamics of mariculture area were mainly affected by the gross fishery production, the mean annual income of employees (agricultural, forestry, animal husbandry, sideline and fishery), the number of fishing motor boats, and the mean annual wind speed. The gross fishery production and the mean annual income of employees were more important in shaping landscape pattern, with a determination coefficient (R²) of 0.77. Our results can serve as a theoretical basis for culture zone planning and ecological environment protection in Fujian coastal areas.

Key words: remote sensing, spatiotemporal dynamics, driving factor, mariculture area, Sansha Bay.