I. Prior Results and Background to the Harvard Forest (HF) LTER
The Harvard Forest LTER has developed into an integrated research program addressing fundamental and applied ecological questions in temperate forest ecosystems that are dynamic as a consequence of disturbance and environmental change (Fig. 1; Foster & Aber 2000). LTER I emphasized site-based investigations in which static and annual estimates of forest pattern and ecosystem process were interpreted in relation to current mean conditions. Long-term experiments contrasted natural disturbance (hurricanes) vs. anthropogenic stresses that are indirect consequences of current land use (N deposition and soil warming). Major findings included: (1) historical land use conditions modern ecosystem structure, process, and dynamics; (2) ecosystem function and trajectories exhibit large inter-annual variation; and (3) interpretations of ecosystem pattern and process require investigations across broad spatial and temporal scales (Fig. 2). Consequently, in LTER II we added retrospective (historical/paleoecological) studies; incorporated land-use history and inter-annual variation into experiment and measurement design; and broadened our investigations to sub-regional (Central Massachusetts) scales. A decade of research has produced major accomplishments: > 300 publications including 22 theses, 12 books, two video presentations; a synthesis volume linking 1000 years of forest dynamics and natural and human disturbance to ecosystem structure, function, and composition (Foster & Aber 2000); a Summer Research Program for 35 students annually; unique methodological and experimental approaches to ecological, atmospheric, and historical research; and regional, national and international cross-site studies.
We are poised in LTER III to (1) expand many studies to regional (New England-wide or larger) scales; (2) evaluate inter-decadal dynamics of key processes and responses; (3) pursue new mechanistic studies of ecosystem, population, and atmospheric processes to interpret long-term trajectories of major experiments; (4) incorporate critical disturbance and stress phenomena (major forest pathogens, forest logging and conversion, ozone, Little Ice Age climate change, exotic species) with ongoing studies of land-use, hurricanes, fire, N deposition, and drought; and (5) apply results to local, regional, and global issues in climate change, water and air quality, forest management, land protection, and conservation. Results below, reported primarily from LTER II, set the stage for these new studies.
Specific objectives of Harvard Forest LTER II included: (1) contrasting pre-European and historical forest dynamics and disturbance; (2) using these histories to interpret vegetation patterns at sub-regional scales; (3) evaluating controls (land-use, climate, vegetation, soils) on C and N dynamics; and (4) synthesizing results through integration and model development.
Retrospective Studies using paleoecological and historical approaches enabled us to analyze dynamics over millennia to understand fundamental ecological processes, to identify key factors shaping modern forest conditions, and to provide management and conservation insights (Foster et al. 1996; Foster & Motzkin 1998). In LTER II, we sought to interpret the impact of New England's history of deforestation, agriculture, farm abandonment, and reforestation on modern forest patterns with a focus on Harvard Forest and the C. Massachusetts sub-region. Important findings include: (1) at the time of European settlement, tree distributions were correlated with regional climate, whereas modern vegetation is more homogeneous as a consequence of historical land use and is not reverting to pre-settlement patterns (Foster et al. 1998b); (2) pre-European vegetation was dynamic as a consequence of climate change (Little Ice Age), fire, and Indian activity (Fig. 3; Fuller et al. 1998); (3) historical decreases in long-lived, shade-tolerant trees (e.g., beech, hemlock) and increases in sprouting/successional species (e.g., birch) were driven by land-use and lagged response to the Little Ice Age (Foster 2000); (4) pre-European forest communities were controlled by site conditions, climate change, and fire; modern assemblages are novel and reflect specific land-use histories (McLachlan et al. 2000, Foster et al. 2000).
Hurricane Modeling and Reconstruction enables us to interpret the landscape and regional impacts of important storm events based on historical data (Boose et al.1994, 2000). The technique provides meteorological data, regional maps of actual damage, and estimates of wind speed, direction, and damage for each storm and compilations of spatial and temporal gradients for all storms. Landscape damage variation is examined with a simple exposure model. This technique can be applied to any region and allows incorporation of hurricane disturbance regimes into research and management (Foster et al. 1999). Results for New England since 1620 include: (1) hurricanes are key factors regionally, decreasing in frequency and intensity from SE to NW due to weakening over land and cold ocean water (Fig. 4); (2) landscape-level gradients of exposure exist due to topography, whereas stand-level damage is dependent on land-use and stand history; and (3) hurricane impacts varied widely on annual and decadal scales.
Community, Population, and Ecosystem Studies are used to evaluate the role of history vs. site conditions in controlling modern species distributions and assemblages. Separating the influence of these factors is complicated by: (1) inter-correlated environmental gradients, (2) paucity of historical data, and (3) confounding of disturbance history with environment. To address these problems, we used three approaches. Intensive studies of vegetation on homogeneous sand plains allowed assessment of disturbance effects in the absence of site variation. Applicability of the results was evaluated on all similar sites in our C. Massachusetts sub-region. Studies were then extended to complex uplands where variation in environment and disturbance coincide. Results include: (1) modern sand plain communities are more strongly related to past land-use than fire or current site factors (Fig. 5; Motzkin et al. 1996; 1999a). (2) on heterogeneous uplands, species respond individualistically to site factors and disturbance, but distributions are most strongly related to moisture and land-use history (Motzkin et al. 1999b); (3) constraints on dispersal/establishment prohibit some species from re-colonizing sites for centuries (Motzkin et al. 1996; Donohue et al. 2000); and (4) prior land use exerts long-term effects on forest C and N dynamics that varies with site and vegetation (Compton et al. 1998; Compton & Boone 2000).
Long-Term Experiments enable us to evaluate critical ecosystem processes and to contrast forest response to natural disturbance and anthropogenic stress. The Experimental Hurricane mimics the 1938 storm and was stimulated by recognition of the importance of hurricanes and the lack of integrated studies that address forest recovery and reorganization following wind disturbance (Foster 1988a,b; Cooper-Ellis et al. 1999). Results include: (1) most uprooted and damaged trees re-leafed, died in 2-6 years, and were replaced by sprouts, leading to only minor change in microenvironment or composition (Fig. 6); (2) due to continuity of LAI and the soil environment, changes in nutrient cycling and soil C effluxes were minor (Bowden et al. 1993); (3) seedling regeneration was controlled by resource congruence (water, light, nutrients); species success varied across five microsite types; seed dispersal of major tree species varied considerably: heavy-seeded oaks accumulated in uproot pits while light-seeded birch occurred on all microsites (Carlton & Bazzaz 1998a,b); and (4) the study forced re-evaluation of the 1938 hurricane (Foster et al. 1998a). Stands damaged in 1938 were mostly pine on old fields, which were subsequently salvaged; thus, increases in river flows and changes in forest composition in 1938 were more a consequence of land-use than natural disturbance.
The Soil Warming Experiment evaluates the effect of a 5o C rise in soil temperature, similar to that predicted by many global models, on ecosystem processes. Results include: (1) over 7 years, CO2 flux-increases due to warming dropped from 30-40 % to ~9 % (Peterjohn et al. 1994, 1995). Annual decreases occurred in dry and cold years; (2) net N mineralization rates in the soil increased 106-140% in most years, but dropped to 31% during a drought year (Fig. 7). Nitrification was consistently low, less than 5% of annual net mineralization; and (3) trenched plots indicate that growing season microbial respiration was 69-74% of total respiration; root respiration was 26-31%. Heating increased microbial respiration (20-32%) more than root respiration (9-15%), with annual variation due to drought (Melillo et al. 1995).
The N Saturation Experiment tests hypotheses on the response of N-limited ecosystems to increases in N deposition (Aber et al. 1989). Plots in adjacent hardwood and red pine forests are subjected to 0 (control), 5 (low) and 15 (high) g N.m-2.yr-1 additions of NH4NO3 from May to October. Major N fluxes are measured (net mineralization and nitrification, aboveground uptake, litter fall, decomposition, leaching loss of DON (dissolved organic N), N2O emissions) along with aboveground NPP, foliar N, and soil CO2 flux. Results include: (1) Increases in nitrate leaching occurred as predicted with N additions, but the timing differed strongly between sites with differing histories (Fig. 8; Magill et al. 2000). The pine stand on fields fertilized in the 19th C has lower N demand and immediate nitrate loss with high N. The hardwood stand, originally low in N cycling due to prior pasturing, cutting and fire, retained N and showed nitrate loss only in year 9 under high N; (2) N retention occurred without increased soil CO2 efflux (Aber et al. 2000) or apparent microbial nitrate immobilization (Berntson & Aber 2000), suggesting the importance of abiotic or mycorrhizal N retention (Aber et al. 1998). Soil and decomposing litter are greater sinks for N inputs than tree biomass, but the proportions of N uptake by trees vs. soils increased with deposition (Nadelhoffer et al. 1999a); (3) elevated nitrate losses in the pine stand are associated with declines in biomass production (Magill et al. 2000) despite large increases in foliar N. Needle retention time and total LAI declined and net rates of photosynthesis per unit leaf area are unchanged. In the hardwood high N plot, there is an increase in foliar N and woody biomass production; (4) decomposition of litter is depressed with high N additions, supporting hypotheses on the suppression of enzyme decomposition systems with abundant mineral N. N accumulation in litter continued in the absence of mass loss, suggesting chemical interactions between organic substrates and N. Up to 30% of leaf mass loss occurred through leaching dissolved organic carbon; (5) 15N tracer studies were consistent with studies at 6 European sites showing that soils are the dominant sink for N saturation inputs to temperate forests and that N deposition does not dramatically increase C uptake (Nadelhoffer et al. 1999c,d); and (6) variation in plant performance is related to availability of nitrate or ammonia. Tree species differentiate between these N forms, the spatial distribution of ammonia and nitrate affect seedling growth, and growth increases with nitrate deposition.
Comparison of ecosystem response to natural disturbance versus anthropogenic stress showed that although hurricane impacts appear catastrophic, many key ecosystem processes are relatively unchanged and stands recover structure and function rapidly, in keeping with the cyclic history of disturbance and recovery. By contrast, N addition and soil warming have no visible impact yet measurements of ecosystem function suggest serious imbalances with long-term implications for ecosystem function (Foster et al. 1997).
The Environmental Measurement Station (EMS; Wofsy et al. 1993) is a unique eddy covariance system that we developed in order to evaluate carbon exchange in relationship to environmental variables. Major results include: (1) over 75,000 hourly rates of Net Ecosystem Exchange of CO2 (NEE) provide 9 years of annual NEE estimates (Goulden et al. 1996a, b); this spurred creation of flux measurement networks in the US (AmeriFlux) and abroad to examine pattern and process at ecosystem to continental scales (Fig. 9); (2) C uptake (mean ~2.1 t C ha-1 yr-1; Goulden et al. 1996a,b; Frolking et al. 1996) is a consequence of recovery from prior land use and the 1938 hurricane; (3) annual variation in NEE is controlled by growing season length, cloudiness, precipitation, and winter soil temperature (Goulden et al. 1996a).
Measurements of atmospheric exchanges of reactive trace-gases at the EMS include dry deposition for nitrogen oxides, deposition of ozone, and emission of hydrocarbons (isoprene and terpenes) that influence ozone and nitrogen oxides in the regional and global atmosphere. These observations help define how biogenic hydrocarbons mediate N deposition by the formation of hydroxyalkylnitrates (Fig. 10) and elucidate the mechanisms, including local forest processes, that facilitate the escape of nitrogen oxides from regional sources to continental and global scales.
The DIRT (Detritus Input and Removal Treatments) Experiment investigates the mechanisms of carbon dynamics in forest ecosystems and seeks to characterize the role of plant inputs in determining soil properties and organic matter dynamics. Treatments over 10 years include: doubling of leaf litter inputs, no leaf litter inputs, no root inputs, no leaf or root inputs, and soil "impoverishment" by replacing O and A horizons with B horizon soil. Results include: (1) soil respiration measurements estimated fine root production (Bowden et al. 1993b); roots and associated rhizosphere soils are much more responsive to temperature variations than are bulk soils (Fig. 11; Boone et al. 1998); and (2) soil C and N contents, humus turnover, net N mineralization, nitrification, dissolved organic C and N production, and microbial communities are strongly influenced by litter and root inputs (Nadelhoffer et al. in review).
Collaborative Modeling with the Hubbard Brook (HB) LTER and U.S. Forest Service enables spatial extrapolations of ecosystem function based on HF and HB LTER research. We developed spatial data sets for climate (Ollinger et al. 1993,1995), soils and vegetation, and a simple model of forest C, water, and N dynamics (Aber et al. 1995, 1996, 1997, Aber & Driscoll 1997) validated against NPP and water yield at the site and regional (Ollinger et al. 1998) scales, and nitrate leaching losses at HB (Aber & Driscoll 1997). The model includes ozone effects on photosynthesis and forest production (Ollinger et al. 1997) and responses to climate change and N deposition in terms of NPP, water yield, and nitrate leaching in streams. Results include: (1) ambient ozone is reducing forest NPP by ~10%; (2) climate change is predicted to increase regional forest production by about 30% and decrease water yield by 15% due to change in precipitation inputs; (3) annual climate variability exerts a large effect on watershed nitrate losses; and (4) responsiveness to climate variability may make it very difficult to detect changes in nitrate leaching due to N deposition (Fig. 12). At current levels of deposition, more than 100 years of stream chemistry data would be required to demonstrate predicted increases.
Cross-site Research is a growing part of HF LTER research; it provides insight into the range of variation in fundamental processes, strengthens our ability to develop generalizations, and supports our efforts at spatial extrapolation. Controls on C sequestration in forests are examined through comparison of annual and interannual rates of C flux at Howland ME (USFS, DOE), Morgan, IN (DOE), Thompson, Manitoba (NASA), other AmeriFlux sites, and the Brazilian Amazon. Our results indicate that conifer forests (Maine, Manitoba) have a longer season of uptake but lower efficiencies for photosynthesis than Harvard Forest (Goulden et al.1997), whereas the deciduous Indiana forest acts much as Harvard Forest does. The 90-year old stand at Howland takes up C at similar rates, but the old forest in Manitoba releases C due to warming and ablation of permafrost.
Interactions among land use, climate variation, and natural disturbance in controlling forest landscape patterns and dynamics are being compared in New England, Ireland, Puerto Rico (LUQ LTER), and the S Yucatan (Mexico), with support from LTER, NSF-International, NASA, and University College, Dublin (Boose et al. 2000; Foster et al. 1999, 2000). In all areas, historical and modern land use are more important drivers of vegetation structure, composition, dynamics, and function than natural disturbance such as hurricanes or fires or environmental variability. This research confirms similarities among very different landscapes, has re-oriented major research programs in each region, and has been a very successful means of exchanging faculty, scientists, and students.
