NIPPON SUISAN GAKKAISHI
Online ISSN : 1349-998X
Print ISSN : 0021-5392
ISSN-L : 0021-5392
Simulation of Red Tide by Means of Multiple Regression and Fourier Analysis
Akira Ouchi
Author information
JOURNAL FREE ACCESS

1986 Volume 52 Issue 2 Pages 203-207

Details
Abstract
I have already shown that red tides having the index PON can be expressed by five predictor variables (WT, SAL, TDP, DIN, DON) based on multiple linear regression (MRA) model in the northern part of Hiroshima Bay. If future values of five predictor variables at each station can be estimated in some way, it becomes quite possible that red tide with the PON index simulation can be done. Thus, the red tide (PON) simulation in August, 1979 and 1983 were performed using MRA model and Fourier analysis (FFT). The results obtained are as follows.
1) The FFT was used to calculate the future values of five predictor variables at six stations and then correlation coefficients between actual variables and estimated ones were determined to be 0.54 for DON and 0.69-0.93 for four predictor variables.
2) By the simulation I ascertained the possibility of red tide in August, 1979 in which Gymno-dinium nagasakiense actually occurred. The correlation coefficients between actual PON and simulated PON were 0.91 on August 7, 0.86 on August 20 and 0.20 on August 30, respectively. The differences between actual PON and simulated PON ranged from 16 to 230μg/l, and became greater when the red tide occurred.
3) In the case of August, 1983, differences between actual PON and simulated PON ranged from -19 to 13μg/l, indicating that these two PON have the same values. Therefore, this method seems to be very useful for predicting the red tide occurrence in August.
Content from these authors
© The Japanese Society of Fisheries Science
Previous article Next article
feedback
Top