[pp.55-62]
Shingo YAMASHITA, Masatoshi DENDA and Nobukazu NAKAGOSHI (Hiroshima University / Public Works Research Institute)
Abstract:
To explore the useful spatial predictors in the estimating the species diversity of juvenile fish in floodplain pools, we used a regression tree technique to analyze the relationship between species richness of cyprinid juveniles and microhabitat characteristics in a major floodplain pool located in the middle reaches of the Chikuma River in central Japan. We surveyed the pool from 5 to 11 July 2000 and again from 25 to 30 July 2001, when the pool was separate from the main stream. During each survey period, we performed underwater observations and measurements of microhabitat characteristics using differential global positioning system instruments. We calculated the distribution of cyprinid juveniles and the microhabitat characteristics per 2-m grid cell using geographical information system software. Eight parameters described the microhabitat characteristics of the floodplain pool: depth, current velocity, ratio of cobble area to grid cell area, ratio of cover area to grid cell area, distance from cover, distance from pool edge, slope, and aspect. Nine species of cyprinid juveniles were found in the floodplain pool. In both the 2000 and 2001 daytime surveys, the regression tree models selected less distance from cover as the best predictor of the species richness of cyprinid juveniles. Our results suggest that the length of cover perimeter of pool is more important than other parameters as index of sustenance of juvenile fish diversity in estimating the function of floodplain pools as nursery area during isolated season.
Key Word:
data mining, floodplain pool, juvenile fish, microhabitat, regression tree, species richness