SEARCH

SEARCH BY CITATION

Keywords:

  • Biodiversity;
  • landscape;
  • local versus regional;
  • macroecology;
  • nutrient limitation;
  • soil;
  • species richness;
  • streams;
  • variance partitioning;
  • USA

ABSTRACT

Aim  In this continental study, species richness at local (LSR) and regional (RSR) scales was correlated and examined as a function of stream (local) and watershed (regional) environment in an effort to elucidate what factors control diatom biodiversity across scales.

Location  Conterminous United States.

Methods  Data on diatom richness, stream conditions and watershed properties were generated by the US Geological Survey. In the present investigation, RSR was estimated as the total diatom richness in a hydrologic study unit and, together with stream and watershed characteristics, was included in stepwise multiple regressions of LSR. The unique and shared contributions of RSR, stream and watershed environment to the explained variance in LSR were determined by variance partitioning. RSR was regressed against stream and basin features averaged per study unit.

Results  LSR responded most strongly to variability in stream manganese concentration and RSR. Other predictors included stream discharge and iron concentration, soil organic matter content and fertilization, and proportions of open water, barren land and forest in the watershed. Variance partitioning revealed that RSR had the lowest independent contribution to explained variance in LSR. Multiple regressions identified average stream iron concentration as the most important predictor of RSR.

Main conclusions  Local micronutrient concentration was the major predictor of LSR, followed by RSR. Since average micronutrient supply in the region was the chief determinant of RSR, it is proposed that micronutrients had both a direct effect on LSR and an indirect effect through RSR. The same argument is extended to watershed features with an impact on stream trophic status, because of their substantial contributions to the explained variance in both LSR and RSR. Considering that the major proportion of LSR variance explained by RSR originated from the covariance of RSR with stream and watershed properties, it is concluded that the LSR–RSR relationship was mediated by the local and regional environment.