基于地理加权回归的渤海沙氏下鱵鱼仔稚鱼栖息地指数

Habitat suitability index of larval Japanese Halfbeak( Hyporhamphus sajori) in Bohai Sea based on geographically weighted regression

为研究渤海鱼类资源早期补充过程,本文将地理加权回归法(GWR)引入栖息地指数(HSI)模型,选取海表温度、海表盐度、水深和叶绿素a浓度4个环境因子建立基于GWR的渤海沙氏下鱵鱼仔稚鱼的HSI_(GWR)模型.模拟发现:在2015年8月渤海的HSI_(GWR)模型中,海表温度和叶绿素a浓度为全局变量,两者的回归系数分别为-0.027和0.006,对HSI影响较小.海表盐度和水深为局地变量,两者回归系数绝对值的平均值分别为0.075和0.129,对HSI的影响较大.其中,海表盐度在渤海中部与HSI呈负相关,负相关系数最大,为-0.3,在三湾呈微弱正相关,相关系数最大值为0.1;水深在整个渤海均与HSI呈负相关,且在三湾的负相关程度明显大于渤海中部,三湾的负相关系数最大,为-0.16.该HSI_(GWR)模型的泊松相关系数为0.705,拟合效果较好,可为今后的鱼类栖息地环境研究提供一种新的方法.

To investigate the early supplementary processes of fishre sources in the Bohai Sea, the geographically weighted regression （GWR） was introduced to the habitat suitability index （HSI） model. The Bohai Sea larval Japanese Halfbeak HSIGWR model was established with four environmental variables, including sea surface temperature （SST）, sea surface salinity （SSS）, water depth （DEP）, and chlorophyll a concentration （Chl a）. Results of the simulation showed that the four variables had different performances in August 2015. SST and Chl a were global variables, and had little impacts on HSI, with the regression coefficients of -0.027 and 0.006, respectively. SSS and DEP were local variables, and had larger impacts on HSI, while the average values of absolute values of their regression coefficients were 0.075 and 0.129, respectively. In the central Bohai Sea, SSS showed a negative correlation with HSI, and the most negative correlation coefficient was -0.3. In contrast, SSS was correlated positively but weakly with HSI in the three bays of Bohai Sea, and the largest correlation coefficient was 0.1. In particular, DEP and HSI were negatively correlated in the entire Bohai Sea, while they were more negatively correlated in the three bays of Bohai than in the central Bohai Sea, and the most negative correlation coefficient was -0.16 in the three bays. The Poisson regression coefficient of the HSIGWR model was 0.705, consistent with field measurements. Therefore, it could provide a new method for the research on fish habitats in the future.