Occasional, rare measurements of stocks have limited use in constraining the estimates of other components of the C cycle. Long time series are particularly crucial for improving the analysis of pools with long time constants, such as SOM, woody biomass, and woody debris. Long-running forest stem surveys, and tree ring data, offer a rich resource that could be assimilated to provide an important constraint on C cycling of slow pools. For extending estimates of NEE across regions, DA can play a further important role, by assimilating remote-sensing data into the analysis of C cycles.We show, via sensitivity analysis, how assimilating an estimate of photosynthesiswhich might be provided indirectly by remotely sensed data -improves the analysis of NEE. I I I I I I I I I I f I 1 I f I I 1 estimates of ecosystem C stocks and fluxes, with reduced uncertainty compared with the original observations, or the model alone. The argument of this paper is that combining measurements and modelling through DA generates more precise estimates of C dynamics, and simultaneously highlights areas where model improvement is required. MethodsThe premise of DA is that neither models nor observations can perfectly describe a system, but an analysis that combines model and data will provide a better estimate of system dynamics than model or observations alone. DA is a process for the optimal combination of information about a system, which evolved from the engineering approaches to filtering and control theory applied in missile guidance and interception (Maybeck, 1979). DA has been applied in meteorology for forecasting (Lorenc, 1986), and ex- DA is the process of finding the model representation that is most consistent with observations (Lorenc, 1995). DA recognizes that there are never sufficient observations to represent the state of a system at any one time. For a detailed, complete picture, further information is required, such as knowledge of the behaviour and probable structure of the system. In DA, knowledge of system evolution in time is usually embodied in a model. In sequential assimilation, the approach we demonstrate here, the model organizes and propagates forward information from previous observations (Lorenc, 1995). When new information becomes available, the prediction, or forecast, of the model can be compared with these observations and corrected. A poor model will drift and will be frequently and heavily corrected; an effective model will require little ~initializa-tion by observations. However, it is not simply a question of fitting the model to the new data, as the assimilation process must also conserve the information provided by the model itself and by previous observations.The DA technique that we use here is the Kalman filter (KF) (Kalman, 19601, which has been widely used (Grewal, 1993), and, given various assumptions, has been shown to be an optimal, variance-minimizing analysis (Maybeck, 1979). The basic KF requires three assumptions: that a linear model can describe the system, and that ...

Abstract. The biogeochemical behavior of carbon in the forested watersheds of the Hubbard Brook Experimental Forest (HBEF) was analyzed in long-term studies. The largest pools of C in the reference watershed (W6) reside in mineral soil organic matter (43% of total ecosystem C) and living biomass (40.5%), with the remainder in surface detritus (14.5%). Repeated sampling indicated that none of these pools was changing significantly in the late-1990s, although high spatial variability precluded the detection of small changes in the soil organic matter pools, which are large; hence, net ecosystem productivity (NEP) in this 2nd growth forest was near zero (± about 20 g C/m 2 -yr) and probably similar in magnitude to fluvial export of organic C. Aboveground net primary productivity (ANPP) of the forest declined by 24% between the late-1950s (462 g C/m 2 -yr) and the late-1990s (354 g C/m 2 -yr), illustrating age-related decline in forest NPP, effects of multiple stresses and unusual tree mortality, or both. Application of the simulation model PnET-II predicted 14% higher ANPP than was observed for 1996-1997, probably reflecting some unknown stresses. Fine litterfall flux (171 g C/m 2 -yr) has not changed much since the late-1960s. Because of high annual variation, C flux in woody litterfall (including tree mortality) was not tightly constrained but averaged about 90 g C/m 2 -yr. Carbon flux to soil organic matter in root turnover (128 g C/m 2 -yr) was only about half as large as aboveground detritus. Balancing the soil C budget requires that large amounts of C (80 g C/m 2 -yr) were transported from roots to rhizosphere carbon flux. Total soil respiration (TSR) ranged from 540 to 800 g C/m 2 -yr across eight stands and decreased with increasing elevation within the northern hardwood forest near W6. The watershedwide TSR was estimated as 660 g C/m 2 -yr. Empirical measurements indicated that 58% of TSR occurred in the surface organic horizons and that root respiration comprised about 40% of TSR, most of the rest being microbial. Carbon flux directly associated with other heterotrophs in the HBEF was minor; for example, we estimated respiration of soil microarthropods, rodents, birds and moose at about 3, 5, 1 and 0.8 g C/m 2 -yr, respectively, or in total less than 2% of NPP. Hence, the effects of other heterotrophs on C flux were primarily indirect, with the exception of occasional 2 -yr) were small, larger quantities of C were transported within the ecosystem and a more substantial fraction of dissolved C was transported from the soil as inorganic C and evaded from the stream as CO 2 (4.0 g C/m 2 -yr). Carbon pools and fluxes change rapidly in response to catastrophic disturbances such as forest harvest or major windthrow events. These changes are dominated by living vegetation and dead wood pools, including roots. If biomass removal does not accompany large-scale disturbance, the ecosystem is a large net source of C to the atmosphere (500-1200 g C/m 2 -yr) for about a decade following disturbance and becomes a net si...

As forests age, their structure and productivity change, yet in some cases, annual rates of water loss remain unchanged. To identify mechanisms that might explain such observations, and to determine if widely different age classes of forests differ functionally, we examined young (Y, approximately 25 years), mature (M, approximately 90 years) and old (O, approximately 250 years) ponderosa pine (Pinus ponderosa Dougl. ex P. Laws.) stands growing in a drought-prone region of central Oregon. Although the stands differed in tree leaf area index (LAIT) (Y = 0.9, M = 2.8, O = 2.1), cumulative tree transpiration measured by sap flow did not differ substantially during the growing season (100-112 mm). Yet when water was readily available, transpiration per unit leaf area of the youngest trees was about three times that of M trees and five times that of O trees. These patterns resulted from a nearly sixfold difference in leaf specific conductance (KL) between the youngest and oldest trees. At the time of maximum transpiration in the Y stand in May-June, gross carbon uptake (gross ecosystem production, GEP) was similar for Y and O stands despite an almost twofold difference in stand leaf area index (LAIS). However, the higher rate of water use by Y trees was not sustainable in the drought-prone environment, and between spring and late summer, KL of Y trees declined fivefold compared with a nearly twofold decline for M trees and a < 30% reduction in O trees. Because the Y stand contained a significant shrub understory and more exposed soil, there was no appreciable difference in mean daily latent energy fluxes between the Y stand and the older stands as measured by the eddy-covariance technique. These patterns resulted in 60 to 85% higher seasonal GEP and 55 to 65% higher water-use efficiency at the M and O stands compared with the Y stand.

Spatial patterns of tree species in forested landscapes are regulated by a variety of environmental and disturbance factors. Biological factors such as disturbance and competition that operate within a local neighborhood (neighborhood factors) might also influence these patterns. We sought empirical evidence for the role of neighborhood factors in determining spatial patterns of abundance of dominant tree species in the 3160‐ha Hubbard Brook Experimental Forest (HBEF), a second growth, northern hardwood–conifer forest in New Hampshire, USA. We measured tree abundance patterns and a suite of environmental and disturbance factors expected to regulate these patterns in 0.05‐ha plots distributed throughout the Hubbard Brook Valley. Environmental and disturbance effects were modeled using linear regression with spatially correlated errors described by semivariograms. These models explained 26–62% of the variation in abundance among the seven tree species that comprise 90% of the total basal area in the HBEF. Semivariograms described residual spatial autocorrelation of the abundance of each tree species after accounting for environmental and disturbance effects. One species, Betula alleghaniensis, did not exhibit significant residual spatial autocorrelation in its spatial pattern of abundance. The other six species exhibited abundance patterns with highly significant residual spatial autocorrelation, suggesting that factors other than environment and disturbance are needed to adequately explain their spatial patterns of abundance. The range and normalized sill semivariogram values from the residual spatial autocorrelation of the different species were consistent with a role of seed dispersal distance and root sprouting in regulating patterns of abundance of several tree species in the HBEF. We argue that neighborhood factors significantly influence patterns of tree species in forested landscapes.

[1] We investigated the relative importance of climatic versus biotic controls on gross primary production (GPP) and water vapor fluxes in seasonally drought-affected ponderosa pine forests. The study was conducted in young (YS), mature (MS), and old stands (OS) over 4 years at the AmeriFlux Metolius sites. Model simulations showed that interannual variation of GPP did not follow the same trends as precipitation, and effects of climatic variation were smallest at the OS (<10%), largest at the MS (>50%), and intermediate at the YS (<20%). In the young, developing stand, interannual variation in leaf area has larger effects on fluxes than climate, although leaf area is a function of climate in that climate can interact with age-related shifts in carbon allocation and affect whole-tree hydraulic conductance. Older forests, with well-established root systems, appear to be better buffered from effects of seasonal drought and interannual climatic variation. Interannual variation of net ecosystem exchange (NEE) was also lowest at the OS, where NEE is controlled more by interannual variation of ecosystem respiration, 70% of which is from soil, than by the variation of GPP, whereas variation in GPP is the primary reason for interannual changes in NEE at the YS and MS. Across spatially heterogeneous landscapes with high frequency of younger stands resulting from natural and anthropogenic disturbances, interannual climatic variation and change in leaf area are likely to result in large interannual variation in GPP and NEE.