Application of Akaike Information Criterion (AIC) to Assess Chlorophyll-a Concentration and Seasonal Changes in Lake Shikotsu based on Landsat/TM

Abstract

Akaike Information Criterion (AIC) is used to select the most appropriate algorithm when using multiple parameters. In this study, AIC was applied to assess the surface chlorophylla concentration in Lake Shikotsu, a typical oligotrophic lake in Hokkaido, Japan, using LANDSAT/TM data. Multiple regression analyses were conducted for all 63 patterns of combinations of bands 1–5 and 7 obtained from the Landsat/TM. Water sample data were set as the objective variable of chlorophyll-a concentrations. The AIC of three channels (band 1, 3, 7) was smallest and there was a significant correlation between the chlorophyll-a concentration measured in water samples and that determined based on bands of 1, 3 and 7 of the Landsat/TM (R=0.74, P<0.01). The algorithm selected using AIC tended to underestimate when the chlorophyll-a concentration was over 0.6 µg L^-1 and overestimate when the concentration was under 0.2 µg L^-1. The algorithm selected by AIC revealed seasonal variations in chlorophyll-a concentration distribution in Lake Shikotsu, as well as lower levels throughout the lake in winter. The chlorophyll-a concentration increased in April and peaked in May and June, then decreased from July, reaching levels similar to winter lows in October.