The study measured landscape level diversity of the understory plants of mature, upland forests in north-central Wisconsin USA. Habitat types were used to segregate the landscape along a moisture-nutrient gradient. Forty sites that had closed canopies, had been undisturbed for at least 20 years, and were at least 8 ha in size were used. The percent cover of groundlayer species was ocularly estimated in 12-18 randomly located, one meter square plots in June and August, 1995. Shrub cover was estimated by the line intercept method. Alpha, beta and gamma diversity were determined for early and late summer periods separately. Gamma diversity was quantified using a new method, affinity analysis, which generates a list of modal and outlier sites and calculates mosaic diversity, a measure of landscape complexity. Generally, communities in the middle of the moisture-nutrient gradient were modal, whereas those at the mesic end of the gradient were outlier. Mosaic diversity values were very similar for early summer and late summer (2.88 +/-0.04, 2.95 +/-0.03, respectively), but was much higher for both periods combined (3.95 +/-0.07). Whittaker's Index (beta diversity) revealed varying rates of species turnover along presumed moisture and nutrient gradients, whereas species densities and richness were relatively constant among habitat types. A one-way analysis of variance of Shannon-Weaver values found no significant differences among habitat types (p greater than or equal to0.05). Regional diversity mainly resulted from high beta values which appears to be primarily a function of the moisture gradient. The other factors influencing compositional differences among sites are variation in site history, especially disturbance, with niche partitioning and differences in seed dispersal capacity having a minor influence. The affinity analysis method indicated that sampling once per season is inadequate, and that many types of sites are modal. This method for estimating gamma (landscape) diversity shows considerable promise, but information on the processes that produce outlier sites is needed to fully understand and use the results of this method.
