Local and regional variation in hemlock seedling establishment in forests of the upper Great Lakes region, USA

Successful tree regeneration requires both successful seedling establishmen t and subsequent survival and growth sufficient to ensure recruitment. We e xamined patterns of initial seedling establishment in randomly selected sta nds of eastern hemlock (Tsuga canadensis) distributed across northern Wisco nsin and eastern Upper Michigan. In 1990, we counted all hemlock seedlings (defined as 4-29 cm tall) and sampled substrate conditions in two 7x7 m qua drats of 142 hemlock and hemlock-hardwood stands. In 1996, we resampled 109 of these stands more intensively using six 7x7 m quadrats. In 1997, we als o intensively surveyed which substrate-supported hemlock seedlings in three 50 m(2) replicate areas from each of five stands. In 1990, we found no hem lock seedlings at 64 of the 142 stands sampled, and 46% of the stands had m ore than 100 hemlock seedlings ha(-1) (mean density=480+/-90 ha(-1)). In co ntrast, only 14 (13%) of the 109 stands resampled in 1996 had no seedlings and 66% had over 100 seedlings ha(-1) (mean density=840+/-200 ha(-1)). Thes e results suggest that seedling establishment is patchy, causing the estima ted frequency of successful establishment to depend strongly on the area sa mpled. Seedling abundance within sites was correlated between years, sugges ting that site history may in part account for the abundance of seedlings. Analysis of variance to mean ratios reveals that seedlings are indeed stron gly clumped at the 7x7 m scale or smaller. Such clumping probably results f rom the tendency for hemlock seedlings to occur most frequently on particul ar substrates (i.e. decaying wood). Hemlock seedlings occur less frequently in stands with high basal area of sugar maple but do not decline in stands with greater hemlock basal area, suggesting that maple litter may restrict local opportunities for seedling establishment. Seedling densities do not vary among habitat-types, and no other stand or regional variable significa ntly affected hemlock seedling density. These results suggest that ground s urveys are needed to assess hemlock seedling abundance, and that such surve ys should be based on quadrats of at least 100 m(2). Successful hemlock reg eneration should capitalize on local patches of existing hemlock seedlings where interference by hardwoods is minimal. Even high initial seedling esta blishment, however, may not guarantee regeneration in areas where high deer densities preclude recruitment into larger size classes. (C) 1998 Elsevier Science B.V.