多种数据源下栖息地模型及预测结果的比较

由于多来源的海洋环境数据常以不同时间、空间分辨率呈现,并具有不同的误差,因此,有必要分析数据源的差异是否会对研究结果产生显著影响,是否会影响基于不同数据源估计的模型对其他数据的适用性。为此,本研究利用多个网站提供的叶绿素浓度与海表水温数据,采用线性回归与随机检验方法,分析了不同数据源对栖息地模型构建及其预测效果的影响。研究结果表明,不同数据源的数据之间常存在系统性偏差,从而使得模型参数的估计具有显著性差异,该模型不适合于其他数据源的数据;多源环境数据间的离散性反映数据存在随机误差,环境数据的随机误差将使模型结果具有随机性,因此本研究建议定量分析模型结果的不确定性,以使模型结果得到科学应用。

Comparisons of the habitat suitability index models developed by multi-source data and forecasting

Marine environmental data are always multi-source and multi-versional. This is because the different methods used for data collection vary in their retrieval algorithms, procedures used and purposes, so that data processing has various spatial and temporal resolutions with different errors. Hence, it is important to know whether results derived from different versions of the same data are consistent and the models can be correctly used by testing other versions of the data. For this purpose, we collected sea surface temperature data from different web sizes using linear regression and randomization tests to evaluate the effects of different data versions on the parameter estimations and predictions of habitat suitability index models. The results showed that because of system errors in the data, the parameters estimated differed significantly and the models were unable to make correct forecasts by inputting other versions of the data. Dispersion between different data versions reflected the random errors inherent in the data and led to uncertainty in the results of the habitat suitability index models. Accordingly, we suggest that model outputs are quantified for uncertainty to ensure that scientific data can be reliably used in fishery resource management.