多变量分位数回归构建印度洋大眼金枪鱼栖息地指数

Multivariate Quantile Regression on Habitat Suitability Index of Thunnus obesus in the Indian Ocean

以0～300m水层加权平均水温、50～150m水层的温差和氧差及其交互变量为影响因子,运用分位数回归法,寻找出环境变量与大眼金枪鱼(Thunnus obesus)延绳钓钓获率的最佳上界分位数回归方程,计算出栖息地指数(HSI),并应用地理信息系统(GIS)软件绘制各月HSI空间分布图。研究表明:大眼金枪鱼延绳钓钓获率(HR)依加权平均水温(x)、温差(y)、氧差(z)与的最佳上界分位数回归方程为HR0.70=-15.596+2.124x-0.003x3+0.033xyz-0.036y2z+0.107yz2-0.337z3;HSI空间分布为:16°S—10°N印度洋海域HSI高于0.7,HSI>0.8的海域随季节发生显著变化,马达加斯加外海至100°E、16°S—26°S海域常年存在一片HSI<0.4的区域,26°S—40°S海域的HSI介于0.4～0.5,40°S以南海域HSI<0.4,东非外海季节性地出现一片HSI<0.6的海域。利用多个环境变量的栖息地指数模型来预测分析大洋金枪鱼资源分布效果较好。

Some important variables are selected to carry out quantile regression including weighted average water temperature in the 0 - 300m water layer, temperature difference and dissolved oxygen difference between 50m and 150m water layer and their interactions. The optimum upper boundary equation is regressed for these variables against bigeye tuna Thunnus obesus longline hooking rate to compute habitat suitability index （HSI）. The computation results are visualized by geographical information system. It shows that the optimum upper boundary equation of weighted average water temperature（x）, temperature difference（v）, dissolved oxygen difference （z） against bigeye tuna longline hooking rate（HR） is as follows： HR0.70=-15.596＋2.124x-0.003x3＋0.033xyz-0.036y^2z ＋0.107yz^2- 0.337z^3. The GIS maps illustrate that HSI is greater than 0.7 within 16°S--10°N in the Indian Ocean. Area significantly varies seasonally where HSI is greater than 0.8. HSI ranges from 0.4 - 0.5 between 26°S - 40°S. There is an area of HSI less than 0.4 that exists within Madagascar to 100°E and 16°S -26°S throughout the year. Area of HSI less than 0.6 seasonally appears away off East Africa. It is found that the better results of resources distribution could be predicted by use of multi-environmental variables in the estimation of HSI.