A Comparison Of Integrated Habitat Index For Albacore Tuna Between Quantile Regression Method And Support Vector Machine Method

Albacore tuna(Thunnus alalunga) is one of the high catch species in the world. China has paid more attention to this high value tuna resources. The sustainable utilization of albacore tuna resources in South Pacific Ocean has become an important task of the Western and Central Pacific Fisheries Commission(WCPFC).It will be beneficial to the sustainable utilization of albacore tuna resources to better understand the integrated influence of the environmental variables on its distribution. The catch rates and environmental variables(temperature, chlorophyll-a, horizontal current, vertical current) were collected at 56 sampling stations in waters near Cook Islands from Sep, 2013 through Dec, 2013. The prediction models of albacore tuna catch per unit fishing effort(CPUE) in each depth strata and the entire water column were built by quantile regression model(QRM) and support vector machine(SVM), respectively. The predicted CPUEs were calculated by inputting the environmental variables of the model verification stations to the prediction models of QRM or SVM, respectively. Wilcoxon test was used to test if there were significant differences between the predicted CPUE and the nominal CPUE of each stratum and entire water column in order to determine the prediction accuracy of both methods. The Spearman correlation coefficients between predicted IHI and nominal CPUE of each stratum and entire water column were calculated by inputting the environmental variables of the model building stations to the prediction models of QRM or SVM, respectively. The prediction power of QRM and SVM in each stratum and entire water column were analyzed.The results are as follow:1)There was no significant differences between the predicted CPUE and the IV nominal CPUE of each stratum and entire water column for the prediction models of QRM or SVM. Both models were suit for predicting the spatial distribution of the albacore tuna.2)The trend of IHI predicted by SVM was more consistent with the trend of the nominal CPUE. The predicted power of IHIQRM models in the depth strata of 40-79.9m and 160-199.9m were excellent by analyzing the Spearman correlation coefficients between predicted IHI and nominal CPUE of 42 model building stations and the Spearman correlation coefficients were 0.73 and 0.83, respectively. The predicted power of IHISVM models in the depth strata of 40-79.9m and 200-239.9m were excellent and the Spearman correlation coefficients were 0.81 and 0.82, respectively. The predicted power of IHISVM models(The Spearman correlation coefficients was 0.62) in the entire water column was higher than that of the IHIQRM models(The Spearman correlation coefficients was 0.54). The Spearman correlation coefficients of the IHISVM in the other depth strata and the entire water column were higher than that of the IHIQRM, except the depth stratum of 160-199.9m.3)In the entire water column, the critical environmental variables in both models that influence on the distribution of albacore tuna were similar. The critical environmental variables in the QRM was the temperature and chlorophyll-a, and that of the SVM was the temperature. The higher IHI areas predicted by both models were similar. In the QRM, both areas with high IHIQRM were 9°30′S~12°30′S, 158°00′W~162°00′W and 10°30′S~14°30′S, 164°00W~167°00′W. In SVM, three areas with high IHISVM were 9°00′S~12°20′S, 159°00′W~164°00′W, 13°30′S~14°30′S, 159°00′W~161°00′W and 10°30′S~12°30′S, 167°00′W~168°00′W.4)In each depth strata, the environmental variables that influence the distribution of albacore tuna were different. In QRM, the distribution of albacore tuna was mainly influenced by temperature, chlorophyll-a, and their interaction. In SVM, that was mainly influenced by chlorophyll-a, temperature and vertical current.5)Based on the measured data at sea and the results of analysis, the preferred swimming depth stratum of albacore tuna was 120.0 ~ 199.9m. We suggested that more fishing gear should be deployed to the depth stratum of 120.0 ~ 199.9m for reducing the bycatch and increasing the catch rate of albacore tuna as far as possible. 6)We suggested that the SVM could be used to predict the spatial distribution of albacore tuna in the waters near Cook Islands. The predicted power of SVM was higher than that of QRM.