To detect suspended sediment by remote sensing technology, researchers have brought up various kinds of experiential or deductive models.The precision and/or accuracy of these models is, more often than not, unable to be assured due to the lack of atmospheric parameters or enough statistics.In this paper, we take Lake Chaohu as an example area, and process the TM data of Lake Chaohu as follows:(1) we get the similar spectrum character between the relative reflection rate and the actual reflection rate by the atmospheric correction which is called the internal average relative reflectance;(2) we get the map of water with different sediment levels by extracting water, creating a map with 0/1 encoding, masking and acquiring a map of SI=(TM2+TM3)/(TM2/TM3);(3)we get the map of suspended sediment relative distribution by applying the density slicing to the map of SI with the criteria for the classification.The results show that:(1)we can get an accurate extraction of water information from Lake Chaohu by the way of considering the relationship between the spectrum;(2)The relative suspended sediment revealed are in accordance with the instrumental data in situ.(3)According to the remote sensing image, the high-level suspended sediment areas had expanded, i.e 1.5 times larger during the past 13 years, indicating certain changes of lakes estuary, lake shoreline and the suspended sediment content.(4)The comprehensive survey shows that the main source of suspended sediment of Lake Chaohu is from the river transportation, and from leftovers caused by the erosion of lakeshore.