基于分位数回归的库克群岛海域长鳍金枪鱼栖息环境综合指数

An integrated habitat index for albacore tuna(Thunnus alalunga) in the waters near the Cook Islands based on the quantile regression method

根据2012年9-11月在库克群岛（the Cook Islands）海域利用金枪鱼延绳钓调查所获得的共计43个站点的长鳍金枪鱼（Thunnus alalunga）渔获率数据,以及测得的温度、盐度、叶绿素浓度、水平海流及垂直海流数据等环境因子数据,采用分位数回归方法分析了各水层（40-280 m,每40 m为一层）及整个水体中各个环境因子与长鳍金枪鱼渔获率的关系,并利用43个站点内随机选择的验证站点对不同水层的研究结果进行了验证。研究结果表明：（1）长鳍金枪鱼在各水层及整个水体的单位捕捞努力量渔获量（CPUE）分布呈偏正态分布；（2）调查期间建模站点和验证站点内的预测CPUE与名义CPUE间均无显著性差异；（3）栖息地综合指数（IHI）模型的预测能力较好,且在水深40-80 m、160-200 m及整个水体范围内能有效预测长鳍金枪鱼的分布情况；（4）不同水层影响长鳍金枪鱼分布的因素不同,如在较浅水层（40-80 m）长鳍金枪鱼的渔获率与水色的的关联较大,在80-120 m 水层则主要受水温的影响、在混合水层所在的120-160 m水层则主要受海流的影响,在较深的水层（160-240 m）则主要受饵料分布及水温的影响；（5）长鳍金枪鱼偏好觅食的水层应为160-240 m 水层；（6）长鳍金枪鱼 IHI 指数分布较高的两个海域分别为13°S-15°S,162°W-167°W与11°S-12°S,161°W-167°W。建议在上述两个海域作业时,应使钓具沉降到160-240 m水层,从而在避免兼捕其他水层渔获的同时,提高长鳍金枪鱼的捕捞效率。

We developed an“Integrated Habitat Index （IHI）”model based on the quantile regression method using sur-vey data collected at 43 sites in waters near the Cook Islands from September, 2012 through November, 2012. The model variables included vertical profile data for temperature, salinity, chlorophyll-a, horizontal current, vertical current and catch per unit effort （CPUE） of albacore tuna（Thunnus alalunga）, and the interactions among these variables. Mod-els were developed for five 40 m water strata between 40 m and 240 m and the entire water column to predict the spatial distribution of albacore tuna. The environmental variables measured at modeling sites were used as inputs to the IHI models to predict the IHI value of the 5 strata and the entire water column. We tested for a significant difference be-tween the observed CPUE and predicted CPUE within the 5 water strata and the entire water column using a Wilcoxon test. The Spearman correlation coefficients were assumed to indicate the predictive power of the IHI model. The trend line of the arithmetic average about the predicted IHI for the 5 strata was compared with the CPUEs at the specific depth stratum. The environmental variables at validation sites were used to validate the model’s predictive power. These data were input into the CPUE models to predict the CPUE of the 5 water strata and the entire water column. We used a Wilcoxon test to compare between the predicted and observed CPUEs within the 5 water strata and the entire water column to validate the IHI results. The CPUE for albacore tuna in the 5 strata and the entire water column exhibited a skewed normal distribution with a longer left tail. There was no significant difference between the nominal CPUEs and predictive CPUEs of albacore in the 5 water strata and the entire water column at the modelling sites or the validation sites. The IHI models had good predictive power, and were able to accurately predict the distribution of albacore tuna in the 40-80 m and 160-200 m strata and in the entire water column. The key environmental parameters in the IHI models differed among the depth strata. In the shallow water （40-80 m）, there was a close relationship between the CPUE and water color. Conversely, water temperature, sea currents, and bait and water temperature were the most significant vari-ables in the 80-120 m, 120-160 m, and 160-240 m strata, respectively. Albacore tuna prefer to feed at around 160-240 m, and generally between 160 and 200 m. The CPUE was 20.31 ind/（1000 hooks） in this water strata. The IHIs for al-bacore tuna in the area between 13°S-15°S, 162°W-167°W and 11°S-12°S, 161°W-167°W were relatively high. When fishing in these areas, our results suggest that fishing gear should be deployed at a depth of 160-240 m to increase CPUE and reduce by-catch.