判別分析による赤潮予知

概要

論文の詳細を見る
This study was made to explore the possibility of predicting the red tide by the discriminant analysis based on quality data of surface water in the northern part of Hiroshima Bay. All the data (N=526) obtained from 1976 to 1982 were divided into two groups, non-red tide group and red tide group, according to a critical value 250 μg/l for the PON concentration. Then, the discriminant analysis was performed using six variables (WT, SAL, TDP, DIN, DON, PON). The results obtained are as follows. 1) Mahalanobisdistance was 8.85 and the probability of misclassification was evaluated to be 6.8%. The rate of misclassification was actually 4.2% for 526 data. 2) When a station is classified into the non-red tide group and its misclassification probability becomes more than 25%, or less than 5%, there is a great tendency for red tide to occur around the station. In the latter case, salinity ranged from 5.1 to 15.1% or TDP was 45-141 μg/l, DIN was 439-618 μg/l respectively, and the red tide actually occurred at a rate of 52%. 3) When a station belonging to the red tide group has the misclassification probability more than 17%, there is a great tendency for red tide to disappear. 4) This method has a great advantage in judging a red tide occurrence, because only one misclassification probability is used instead of six variables.
公益社団法人日本水産学会の論文