Abstract: Objective To analyze and compare the spatio-temporal structure and risk factors of county-level seroprevalence of Schistosoma japonicum infection in lake and mountainous regions. Methods Bayesian spatio-temporal models were used to analyze the county-level data from serological tests， which was part of the annual reports on S. japonicum infection in China from 2002 to 2005; also used were normalized difference vegetation index （NDVI）， land surface temperature （LST）， land use type from Moderate Resolution Imaging Spectroradiometer （MODIS）， and the index of economic level. Results The seroprevalence was positively associated with the mean of LST from July to August， the proportion of water body and that of grassland in lake region （regression coefficient: 0.650， 0.662 and 0.832， respectively）， while in mountainous region， the seroprevalence was positively associated with the mean of LST from January to February and the proportion of grassland （regression coefficient: 2.631 and 0.400， respectively）， and negatively associated with the mean of LST from July to August （regression coefficient: -0.288）. The spatial correlation coefficients ranged from 0.868 to 0.945 for lake region while they were not significant for most years in mountainous region. Conclusion The impact of environmental factors on seroprevalence of S. japonicum infection varies in different regions. Seroprevalence presents a str-ong spatial correlation in lake region with certain yearly variability， but such spatial correlation is weak in mountainous region.

Key words: Bayesian statistics, Schistosoma japonicum, Spatio-temporal distribution, Seroprevalence