An analysis of extreme flood events during the past 400 years at Taihu Lake, China

Summary Considerable attention has been paid to the extreme floods affecting Taihu Lake and the lower reaches of the Yangtze River (China) since they have caused serious socio-economic problems past and present. To fully understand these low probability events, it is necessary to build longtime-series of flood occurrence using flood-level proxies from archaeology and sedimentology that extends the period of observation beyond that of instrumental data. Using historical stele flood markers relicts and lacustrine sediment records from Taihu Lake, this paper attempts to reconstruct the historical flood events, to compare the stele-measured flood levels with those recorded by the modern gauges, and to identify the extreme flood signals among the flood events over the past 400 years. Results indicate that the lowest lake level of the 15 extreme floods by stele-records in the period 1600–1954 AD was 4.07 m a.s.l., equivalent to the 80th percentile of hydrological-gauged lake levels during 1921–2004 AD; this comparison provides a quantitative analogue for the floods reconstructed from lake sediments. Flood signals from coarse silt–sand sediments and low-frequency magnetic susceptibility profiles captured 85% of flood years in the historical period, and five extreme flood years that were missed in the stele flood records were found. There were three flood years that were identified from the historical documents in 1766, 1875 and 1882 AD, respectively, when lake-flood levels were estimated at 4.0–4.1 m, 4.1–4.2 m and 4.13–4.23 m, respectively. Spectral analysis was used to identify return periods from the three time series of the stele flood records, grain flood index and magnetic flood index, respectively; these analyses indicated some synchronous patterns and showed common return periods of 90–102 years, 60–62 years and 42–44 years. To test if the historical extreme floods have statistical relationships with climate variability, a two-variables-conditional test following Poisson’s distribution was applied: the estimated probabilities of 0.17–0.20 between a proxy of the Pacific Decadal Osculation (PDO) and the flood events were obtained. These estimates indicate that the hypothesis that the two variables have significant correlations can be accepted, suggesting that the Taihu historical extreme floods were significantly linked with the changes in the Pacific monsoon and summer precipitation. This study demonstrate the possibility of reconstructing past extreme floods where gauge data are not available and has relevance to the prediction of future extreme floods which will have possible disastrous socio-economic impacts on the Taihu catchment.