Time Series Analysis and Forecasting of Tile Fish Fishery in Guei-Shan Island Waters, Taiwan

Abstract

日本馬頭魚爲本省高價魚種之一，近年來其漁獲量有減少的趨勢，然而有關其漁業生物學的研究至今仍付之闕如。本研究使用之漁獲量資料係根據大溪魚市場每月拍賣統計資料，由1977年1月至1992年12月共計192個月，漁獲努力量則由船籍資料及氣象資料推估而得。此漁獲統計資料以時序列分析法解析其時間結構並導出漁況預測方程式，結果發現漁獲量具有一個月的時間延遲而單位努力漁獲量則有4個月的延遲。一階自我相關預測方程式C(下标 t)=444.682+0.4137 C(下标 t-1)可預估未來一個月的漁獲量，此方程式可解釋漁獲量17%的變異。四階自我相關預測方程式U(下标 t)=0.574 U(下标 t-1)+0.242 U(下标 t-2)+0.113 U(下标 t-3)-0.194 U(下标 t-4)可預估未來一個月的單位努力漁獲量，此方程式可解釋單位努力漁獲量40%的變異。合理的假設和容易使用使時間序列分析法成爲漁業管理中一種經濟且有效率的工具。The tile fish, Branchiostegus japonicus, is one of the highest price commercial species in Taiwan; however, its fisheries biology is unknown. The catch data of Guei-Shan Island waters, January 1977 to December 1992, were from the monthly catch record in Tahsi fish market. The effort data were estimated from the monthly weather report and registered fishing boat record. Time series analysis was applied to analyze the time structure and develop equation for month-ahead forecasting of tile fish fishery. A 1-month time lag was found for the catch series and a 4-month time lag was found for the CPUE series. The one month-ahead forecasting equation for a first order autoregressive model ARIMA (1, 0, 0) and a fourth order autoregressivc model ARIMA (4, 0, 0) can be estimated from C(subscript t)=444.682+0.4137 C(subscript t-1) and U(subscript t)=0.574 U(subscript t-1)+0.242 U(subscript t-2)+0.113 U(subscript t-3)-0.194 U(subscript t-4), respectively. The easy-to-use and no unrealistic assumption make the time series analysis an economic and effective tool for fishery management.