Hemi‐marsh concept prevails? Kaminski and Prince (1981) revisited

Abstract

AbstractThe hemi‐marsh concept has prevailed in avian habitat selection and wetland ecology and management since the 1970s. Hemi‐marsh is a stage of wetland succession when approximately equal proportions of emergent vegetation and open water occur. In 1981, researchers reported on a field experiment in Delta Marsh, Manitoba, Canada, where they artificially created interspersion levels of emergent vegetation and open water (30%:70%, 50:50 [hemi‐marsh], and 70:30) by mowing or rototilling to test for hypothesized differences in indicated breeding pair indices (IBPs/ha; pairs + lone males) of dabbling ducks and species diversity among treatment combinations. Greatest dabbler densities and species diversity were in 50:50 hemi‐marsh plots. This research has been widely cited; however, the experiment was pseudo‐replicated because of non‐independence of IBP counts within and among weeks and years of surveys. We reanalyzed data from the study to support or refute original results. More robust and generalized analyses revealed vegetation:water interspersion characteristics influenced IBP densities of blue‐winged teal (Spatula discors), mallard (Anas platyrhynchos), gadwall (Mareca strepera), northern shoveler (S. clypeata), and species diversity. We estimated greatest IBP densities and species diversity in 50:50 hemi‐marsh plots, agreeing with the previous study. Including nested random effects to account for temporal pseudo‐replication improved model fits by 2–30%. Despite accounting for temporal variation in dabbler density and species diversity, 61–93% of variation remained unexplained, indicating unresolved complexities of breeding dabbler habitat use. Nonetheless, our reanalysis generalized and strengthened the original finding that dabbling duck densities and species diversity were greatest in hemi‐marsh plots. Managers may strive to reproduce hemi‐marsh conditions for dabbling ducks and evaluate other waterbird responses. Our reanalysis also emphasized the need for researchers to account for or prevent temporal and spatial pseudo‐replication, examine model fit, and provide reproducible documentation of data and code, especially now that analytical tools and databases are freely available.