LONG-TERM FOREST GROWTH AND DECADAL SOIL CHEMISTRY CHANGE ACROSS THE WHITE MOUNTAIN NATIONAL FOREST

Forest ecosystems are subject to a variety of stressors including human land use, air pollution and climate change. A challenge for detecting temporal change, however, is disentangling heterogeneity at multiple spatial scales. Therefore, we need a better understanding of the mechanisms influencing forest growth and soil formation, how to improve existing long-term sample designs, and quantify variability at multiple spatial scales. Identifying the areal extent of bedrock outcrops and shallow soils has important implications in understanding the spatial dynamics of surrounding vegetation, stream chemistry gradients, and soil properties. Manual methods of delineating bedrock outcrops and associated shallow soil are still commonly employed in the northeastern US and retain numerous limitations associated with the geometry of polygon units. Chapter 2 objectives were to assess the accuracy of visually interpreted high-resolution relief maps for locating bedrock outcrops and associated shallow soil as well as automate the delineation soil using predictive analytics. Visual interpretation of lidar-derived 1 m shaded relief maps at Hubbard Brook Experimental Forest resulted in a 92% accuracy in distinguishing the presence of bedrock outcrops and shallow soil. A generalized additive model had 88.1% overall accuracy using independent validation data and 90.1% overall accuracy predicting bedrock outcrops-shallow soil presence and absence in a second validation watershed 16 km northwest. Chapter 3 objectives were to predict the asymptotic range of stand relative density and biomass as well as explore the influence of topographic metrics as proxies for site quality. Predicting the asymptotic range of stand relative density and biomass has important implications on silvicultural practices and understanding forest carbon pools in late successional forests across northeastern US. In addition, quantifying the influence of site quality on carrying capacity in a mixed species forest is a long-standing challenge and has not been thoroughly tested with long-term longitudinal data. Logistic and Chapman-Richards growth functions were fitted to eight decades of Bartlett Experimental Forest, New Hampshire inventory measurements from 1931-32, 1939-40, 1991-92, 2001-03, and 2015-17 as nonlinear mixed effects models. The variability associated with the plot-level random effects suggested broad differences in structure among the plots could be accounted for with topographic metrics. The variance components of predicted stand relative density and biomass asymptotes in this study influenced by a topographic covariate highlighted the importance of incorporating landscape-soil-water dynamics when characterizing stand dynamics and growth. Finally, Chapter 4 objectives were to investigate overall change in soil chemistry of mid-elevation, northern hardwood Spodosols across the White Mountain National Forest, calculate the variance components at multiple spatial scales, and stratify sampling sites by dominant hydrologic pathways to determine if groundwater influenced soils were more responsive to acidification recovery processes compared to soils developed vertically via unsaturated flow. Forty permanent plots were sampled across the White Mountain National Forest (WMNF), USA in 2001-02 and resampled in 2014. Paired t-tests detected significant increases in carbon and base cations and a decrease in Al in the Oa horizon while base cations decreased and Al increased in some mineral horizons. Additionally, within-site variability was comparable to overall variability across the WMNF. When study sites were stratified into hydrologic groups, we found a strong signal of increasing carbon and base cation concentrations from 2001-02 to 2014 for the Oa horizon, suggesting that soils influenced by shallow groundwater contributions from upslope were more responsive to acidification recovery than soils influenced only by vertical percolation. The combined approach to hydrologic stratification and estimating variance components simultaneously at the landscape and within-plot scales is crucial for calculating sample size needed to detect temporal change