家鱼池塘底泥耗氧率与理化因子的相关性分析

采用原位底泥耗氧测定法,研究了10口家鱼鱼池底泥耗氧率与底部水体理化因子(溶氧、温度、pH值、氧化还原电位)和底泥有机质含量及深度的相关关系。结果显示:池塘平均底泥耗氧率(SOD)为0.91 g/(m2.d),变动范围为0.76～1.09 g/(m2.d)。双变量相关性分析表明,底泥耗氧率与池塘底部水体理化指标的相关性均达到极显著水平(P<0.01),与溶氧相关性最高(Pearson相关系数为0.779),其次是温度、pH值和氧化还原电位,相关系数分别为0.587、0.557和-0.421;底泥耗氧率与底泥深度相关性达到显著水平(P<0.05)。偏相关分析结果表明,底泥耗氧率与溶氧和温度呈极显著相关(P<0.01),与其它因素均未达到显著水平。影响底泥耗氧率最重要的环境因子是溶氧,其次是温度。利用BP神经网络分析影响SOD的理化因子,以溶氧、温度和底泥深度为BP神经网络模型的输入变量建立BP神经网络模型对SOD进行预测分析,BP神经网络模型训练和测试相关系数分别为0.911和0.879,平均相对误差分别为11.6%和10.4%,预测值与真实值偏差较小,拟合度较高,可有效预测池塘底泥耗氧率。

The pond’s sediment oxygen demand and the relationship with physical and chemical factors

The relationship of sediment oxygen demand(SOD) and physical and chemical factors(dissolved oxygen,temperature,pH,oxidation-reduction potential) of bottom water and sediment characteristics(sediment organic matter,sediment depths) in 10 fish ponds were studied by measuring sediment oxygen demand in situ.The results showed that:The average of SOD rate was 0.91 g/(m2·d) in all ponds and the range of changes in different ponds were from 0.76 to1.09 g/(m2·d).Bivariate correlation analysis showed that the relationship between the SOD rate and the physical and chemical factors of the ponds bottom water veried significantly(P<0.01),and the Pearson correlation coefficient with DO was the highest(R=0.779),followed by temperature(R=0.587),pH(R=0.557) and ORP(R=-0.421).The correlation between SOD rate and sediment depth was significant(P<0.05).The partial correlation analysis results showed that the correlations of SOD rate and dissolved oxygen and temperature were highly significant(P<0.01),and other factors did not reach significant levels.The most important environmental factor affectting the SOD rate was dissolved oxygen,followed by temperature.Using BP neural network to analyze the physical and chemical factors influencing the SOD rate,dissolved oxygen,temperature,and sediment depth were regarded as the input variables to establish the BP neural network model and predict the SOD rate.The BP neural network model’ s training and testing correlation coefficient were 0.911 and 0.879,and the average relative error were 11.6% and 10.4%,respectively.There were small differences in the predictive value and the true value,and the model had high accuracy.It can effectively predict the SOD rate.