MACRO-PATTERNS AND PREDICTIVE MODELS OF ZOOBENTHOS ON MID-LOWER YANGTZE SHALLOW LAKES
-
-
Abstract
The Yangtze floodplain is one of the most important wetlands in the world. For decades, many lakes in this region have suffered from multiple stressors, such as over-exploitation fishery resources and cultural eutrophication. To solve these problems, it is necessary to establish a quantitative lake ecosystem management platform at a regional scale. Recognizing the lack of regional scale zoobenthos models with high predictive power, we carried out 105 lake-time investigations on 46 small-to medium-sized lakes along the mid-lower Yangtze River in this study. The results showed that the density and biomass of zoobenthos of these lakes were (847±248) ind./m2 (mean±SE) and (29.41±3.97) g/m2, respectively. The density and biomass of oligochaetes, gastropods and chironomids were (403±225) ind./m2 and (1.12±0.39) g/m2, (82±20) ind./m2 and (26.38±3.99) g/m2, and (356±62) ind./m2 and (1.86±0.58) g/m2, respectively. Further analyses showed that water depth, Secchi depth, water temperature, total phosphorus, phytoplankton chlorophyll a, and submersed macrophyte biomass were the important factors affecting the standing crops of zoobenthos. A series of models were therefore established. The explanation of variations in zoobenthos density among lakes was generally better than that of biomass. The explantion percentage of simple regression models were between 18%—33% for density and 7%—18% for biomass. In the multiple regression models, the explanation percentage was 46%—49% for density and 16%—55% for biomass. Considering the large sampling size of this study, the explanation of these research models was significantly higher than that of previous studies. When comparing the percentage predictive errors (PPE), the simple models and multiple models showed similar results: PPE was 76%—171% for density and 115%—1034% for biomass in simple models, and 88%—114% for density and 141%—1015% for biomass in multiple models. Therefore, these simple regression model with relatively few variables and good predictive power are suggested in practical application.
-
-