Desai AGU05 B44B 05

Influence of Land Cover Heterogeneity, Land-Use Change and Management on the Regional Carbon Cyclein the Upper Midwest USA: 

Influence of Land Cover Heterogeneity, Land-Use Change and Management on the Regional Carbon Cycle in the Upper Midwest USA Ankur R Desai, Kenneth J Davis: The Pennsylvania State University Paul R Moorcroft: Harvard University Paul V Bolstad: University of Minnesota American Geophysical Union Fall 2005 Meeting B44B-05

Complex Regions: 1+1≠2: 

Complex Regions: 1+1≠2 Observational data scaling and ecosystem modeling of land-atmosphere carbon dioxide flux relying solely on dominant cover types is difficult in regions with complex land cover arising from topography, land management and other biotic & abiotic interactions (e.g., light environment, soil type)

Disturbance and Land Cover: 

Disturbance and Land Cover Past and current land use and forest harvest leaves its imprint on modern day land cover on the order of 100 yrs or longer in forested regions This imprint has the potential to alter land-atmosphere carbon exchange magnitudes and patterns The upper-Midwest was heavily clear-cut in the late 19th / early 20th century

The Upper Midwest: 

The Upper Midwest Today, the region is a complex, actively managed, heavily forested region with extensive wetland cover The densely-instrumented landscape is ideal for testing the roles of disturbance, management and scaling on the regional carbon cycle

A Very Tall Tower: 

A Very Tall Tower Park Falls, WI - WLEF Regionally representative CO2 fluxes at 30-396 m Tower shows small source (positive NEE) of CO2 to atmosphere, in contrast to stand-scale towers in region Larger ecosystem respiration (ER), similar gross production (GPP) to stand-scale towers Are we undersampling certain stand ages (young/old) or ecosystems (e.g., wetlands)?

Stand Scale Observations: 

Stand Scale Observations 12+ stand scale eddy covariance towers in region No significant difference in meteorology among sites Stand age/cover lead to significant differences in flux Coherent interannual variability in NEE and GPP

Multi-tower Aggregation Scaling: 

Multi-tower Aggregation Scaling Stand-scale towers were scaled to regional flux, based on: LandSat 30m land cover type Forest Inventory Analysis stand age Two equation parameter optimization Sites are assumed to observe same climate – mostly true

Regional Flux Comparisons: 

Regional Flux Comparisons Multi-tower aggregated fluxes (NEP, ER, GPP) for summer 2003 (blue) has smaller ER and larger GPP than tall tower (red) (Desai et al, in press) However when tall tower fluxes were decomposed & downscaled using footprint models (Wang et al., submitted) and regionally aggregated by land cover density (green) – upscaling and downscaling agree better

Biogeochemical models: 

Biogeochemical models We can further investigate regional flux with models Biome-scale biogeochemical models treat each “cell” as a single plant functional type (or fractions of a few) and 1-2 canopy layers with grid-average values of biomass/fluxes Age can only be modeled by following a cell with time as it builds and loses biomass

Dynamic Ecosystem Models: 

Dynamic Ecosystem Models On the other side of the spectrum are “gap” models that simulate the growth and fate of every plant with explicit interaction among them Computationally expensive Difficult to parameterize Can be complicated to scale Instead, we apply a height-and-age structured gap model to the region that uses concepts of statistical mechanics and ensemble averaging to simulate the dynamics of the mean-moment ensemble of gaps Moorcroft et al, 2001, Ecological Monographs Grid cell consists of multiple patches of different ages Patches also segregated by disturbance type Patches contains multiple cohorts of size and plant type Patch age affects light availability

The Ecosystem Demography Model: 

The Ecosystem Demography Model Farquhar leaf-level photosynthesis with soil water/N limitations and simple canopy light extinction Mean-moment differential equations for cohort density, active plant/root tissue size, non-active plant biomass, and patch CWD, fast, structural, slow and passive soil C and water pools Boundary conditions controlled by reproduction, mortality, disturbance and phenology Source: Hurtt et al., 2002 Source: Moorcroft et al., 2001

Model Data Assimilation: 

Model Data Assimilation Region divided by into soil/topographic sub-sites: mesic upland (N hardwoods/hemlock), xeric upland (N pines/ash-oak), lowland (shrub and forested wetlands) Constrained parameters USFS FIA: mortality, reproduction, harvest Chamber fluxes: component respiration rates, VcMax Biometric: site allometry, specific leaf area, C:N Input variables Meteorology: tower and NCDC air temperature, soil temperature, PAR, CO2, humidity, precipitation Land use/cover: Public land survey for presettlement vegetation (Schulte, 2002), Hurtt et al. land use change Time steps Hourly biogeochemistry, adaptive (days-month) growth and allocation, monthly ecosystem dynamics

Model Setup: 

Model Setup Model sub-sites (mesic, xeric, wet, water/ag/barren) summed by % landcover in 65-km radius around tall-tower Water, agricultural and barren lands are assumed to have 0 NEE, ER, GEP Four model scenarios: Full run (red) No anthropogenic disturbance (blue) Pre-industrial CO2 with anthropogenic disturb. (light red) No CO2 increase or anthropogenic disturbance (cyan) Each scenario and sub-site run from 1800-2004 Forest tent caterpillar infestation in 2001 included Results compared to tall tower (LEF/black), footprint decomposed and aggregated (LEF*) and stand-scaled tower based upscaled fluxes Caveat: results are very preliminary at this point

Results: Land Cover in 2004: 

Results: Land Cover in 2004

Results: Mean Fluxes 1997-2004: 

Results: Mean Fluxes 1997-2004

Results: Comparison to Tall Tower: 

Results: Comparison to Tall Tower

Results: Comparison to Tower Scaling: 

Results: Comparison to Tower Scaling Jun-Aug 2003 LEF = tall tower flux LEF* = downscaled regionally-integrated flux Towers = multi-tower upscaling Model = ED Full Run

Results: Impact of Stand Age: 

Results: Impact of Stand Age

Implications: 

Implications Model results fall within range observed by tall tower and by footprint-based downscaling and stand scale upscaling More observation sites needed in wetlands and young forests – currently underway Carbon fertilization significantly enhances net uptake today in response to logging 100 yrs ago May be artifact from lack of CO2 downregulation Modeled ecosystem respiration is lower than observed at tall tower, but larger than stand scale aggregated towers Young sites (esp. wetlands) have large ER:GPP ratio, leading to positive NEE – role of disturbance residue Carbon sink strength declines in mature and old sites except for xeric sites which continue to strengthen Mesic sites: age of max. ER precedes age of max. GPP Sampling of dominant cover types (mature mixed forest) cannot solely explain regional carbon flux

Future Work: 

Future Work Uncertainty analysis of model output Continued investigation of Climate-carbon flux coupling as a function of stand type Observation network density needed for scaling Minimum resolution required for spatial data Scaling to larger and smaller regions Role of local vs. global parameterizations Carbon fertilization effects Forest product lifecycle, net biome productivity Incorporation of wetland dynamics and biogeochemistry Comparison to atmospheric tracer based regional carbon flux observations: see next two talks – B44B-06 (Davis) / B44B-07 (Uliasz) Incorporation of new biometric and flux data: see talk after that – B44B-08 (Bolstad) Running model into the future with IPCC scenarios