Effect of low count sums on quantitative environmental reconstructions: an example using subfossil chironomids   总被引：7，自引：0，他引：7  

Oliver Heiri  André F. Lotter 《Journal of Paleolimnology》2001,26(3):343-350

The concentrations of chironomid remains in lake sediments are very variable and, therefore, chironomid stratigraphies often include samples with a low number of counts. Thus, the effect of low count sums on reconstructed temperatures is an important issue when applying chironomidtemperature inference models. Using an existing data set, we simulated low count sums by randomly picking subsets of head capsules from surfacesediment samples with a high number of specimens. Subsequently, a chironomidtemperature inference model was used to assess how the inferred temperatures are affected by low counts. The simulations indicate that the variability of inferred temperatures increases progressively with decreasing count sums. At counts below 50 specimens, a further reduction in count sum can cause a disproportionate increase in the variation of inferred temperatures, whereas at higher count sums the inferences are more stable. Furthermore, low count samples may consistently infer too low or too high temperatures and, therefore, produce a systematic error in a reconstruction. Smoothing reconstructed temperatures downcore is proposed as a possible way to compensate for the high variability due to low count sums. By combining adjacent samples in a stratigraphy, to produce samples of a more reliable size, it is possible to assess if low counts cause a systematic error in inferred temperatures.  相似文献   

Relationships between environmental variables and zooplankton subfossils in the surface sediments of 36 shallow coastal brackish lakes with special emphasis on the role of fish   总被引：2，自引：0，他引：2  

Susanne Lildal?AmsinckEmail author  Erik?Jeppesen  Frank?Landkildehus 《Journal of Paleolimnology》2005,33(1):39-51

Subfossil zooplankton assemblages (Cladocera 22 taxa, Rotifera 1 taxon) were identified from the surface sediments of 36 shallow (median depth = 0.7 m) Danish coastal brackish lakes differing in epilimnic salinity (SAL, range 0.2–17.4), summer-mean total phosphorus (TP, 27–327 g l^–1) and total nitrogen (TN, 0.850–2.629 mg l^–1), as well as in submerged macrophyte coverage and planktivorous fish density (PL-CPUE). Cladoceran species richness declined significantly with increasing SAL, TP and TN, while no significant correlation was found to either PL-CPUE, macrophyte coverage or lake surface area. Bonferroni-adjusted forward selection within canonical correspondence analysis (CCA) showed that 22.1% of the variation in zooplankton data was explained by PL-CPUE, SAL and TP uniquely; each variable explaining an almost equally significant amount of variation in the zooplankton data. Predictive models to infer PL-CPUE, SAL and TP were developed using variance weighted-averaging (WA) procedures. Almost similar values of boot-strapped coefficient of determination (r²_{boot-strapped} 0.22–0.38) were produced by the WA inference models of PL-CPUE, SAL and TP, while the inference models of TP produced the lowest boot-strapped root-mean-squared-error of prediction (RMSEP_{boot-strapped} 0.29–0.36 log(TP + 1), g l^–1). Yet, zooplankton TP and SAL optima (WA) were strongly correlated (r² = 0.46), while PL-CPUE optima (WA) were independent of both TP and SAL optima, indicating that only the PL-CPUE inference models are suitable for making reconstructions.  相似文献