Modeling the Economic and Ecological Impact of Climate Change on Southern Forests 1. Robert Abt 2 Brian Murray 3 Steve McNulty 4

Transcription

1 Modeling the Economic and Ecological Impact of Climate Change on Southern Forests 1 by Robert Abt 2 Brian Murray 3 Steve McNulty 4 Abstract This paper describes the linkage of economic and ecological models of the southern forest ecosystems and markets to examine the potential impact of climate change. The core modeling system links the Sub- Regional Timber Supply (SRTS) model with PnET-II a process model of forest growth. Early results indicate significant variation in potential impact across the South based on variations in climate scenarios, ecological systems and management responses. Additional work by agronomists, agricultural economists, and hydrologists will link the forest sector model to the agriculture sector and hydrological modeling to provide a more comprehensive assessment of regional climate change impacts. INTRODUCTION The potential impact of global climate change on human welfare has become a core policy question over the last decade. The original research was focused on obtaining better estimates of climate change. Subsequent research has lead to a general consensus in the scientific community that human-induced changes are occurring (Intergovernmental Panel on Climate Change, 1996). Coupled models of transition climate provide the foundation for a new generation of impact analyses based on higher spatial and temporal resolution. Forested ecosystems cover over onehalf of the land area of the Southern U.S. and are an integral part of the culture and economy of the region. Ninety percent of the forest resource is owned by private landowners with a variety of objectives (Alig et al., 1990). Forest-based industry owns almost one-fourth of the timberland in the area. These industries are a significant component of the southern economy where the value of timber products exceeds agricultural crops in most states. The resource is also a critical and growing component of the U.S. timber economy since approximately 40 percent of the timberland and almost half of the removals (softwood and hardwood) are located in the region. Over the last 50 years, hardwood and softwood inventories have increased in the South. The latest cycle of surveys in the South, however, indicate that softwood removals slightly exceed growth for the region as a whole. Hardwood removals are currently 30 percent below growth, but if current removal trends continue, hardwood inventory could peak in the next decade. Though the ecosystems are predominately hardwood or mixed pine types, the softwood industry has traditionally been more economically important. Hardwood markets, however, have exhibited the greatest increase over the last decade. Sub-Regional Timber Supply (SRTS) model projections indicate that intensive management responses may increase growth enough for the softwood inventory to stabilize within the next decade. There are, however, limited options for intensive management on hardwoods. Forest industry owns primarily softwood timberland and is not likely to expand its land holdings for hardwood management. These factors imply that the next two decades may be significantly different from long-term historical trends. This competitive economic context implies that even the marginal climate changes expected in the next 50 years, might have 1 This work is funded by the Southeastern Center of the National Institute of Global Environmental Change 2 North Carolina State University, Raleigh, NC 3 Research Triangle Institute, RTP, NC 4 USDA Forest Service, Raleigh, NC

2 noticeable effects. While management-based adaptation may be a realistic assumption for pine types, the passive management associated with hardwoods implies that natural adaptation may be more characteristic for hardwoods. To capture the ecological and management variation across the South, a spatially detailed modeling system is required. Forest-Agriculture Modeling System The initial model linkage was focused on linking changes in forest growth due to temperature, precipitation, and CO 2 fertilization. Each of the models is able to model biological and economic information at the sub-regional level. This is important because the biological diversity of the south implies that even with relatively homogenous climate changes across the South, the ecological impact may be spatially diverse. The economic response will depend partially on the biological impact but also on the current level of harvest and management, as well as differences in regional ownership objectives and patterns. The modeling system developed here is focused on modeling the spatial diversity of climate change impacts in the south. The overall system is shown in Figure 1, the focus on this paper is the linkage between the biological model, PnET, and the Sub-Regional Timber Supply Model (SRTS). A more detailed description of these two models follows. Additional research to make land use and management endogenous to the system are briefly described below. Forest Process Model PnET-II is a forest process model developed to predict forest productivity and hydrology across a range of climates and site conditions (Aber and Federer, 1992; McNulty et al., 1998). PnET-II calculates the maximum amount of leaf-area that can be supported on a site based on the soil, the climate, and tree species specific vegetation attributes (Aber et al. 1995). The model does not account for differences in sites due to insect, disease, or specific management activities (i.e., burning or thinning). Predicted net primary productivity (NPP) is a principle model output and is calculated as total gross photosynthesis minus growth and maintenance respiration for leaf, wood and root compartments. Gross photosynthesis is first calculated without water stress effects as a function of temperature, foliar N concentration, and vapor pressure deficit. Potential transpiration is calculated from potential gross photosynthesis and water-useefficiency. Actual transpiration is a function of potential transpiration and available soil water. The latter quantity is related to the soil water holding capacity, a soil moisture release parameter, and incident soil water. After the water balance is updated, actual gross photosynthesis is calculated from water stress and potential gross photosynthesis. Wood, root and leaf respiration is a function of the current and previous month s average minimum and maximum air temperature. Forest Economic Model Timber market and inventory modules are the two major components of the Sub- Regional Timber Supply (SRTS) model. For the analysis presented here, FIA survey units and industry and other private ownerships in the South were used to define 102 (51 units x 2 owner types) sub-regions in the model. Public lands and harvest were excluded from the model because market forces do not drive their harvest and management decisions. Market Model Structure Usually market equilibrium is modeled to determine the price and quantity that result from exogenous shifts in supply and demand. SRTS was developed to link to inventory models that use timber harvest as the control variable. Thus the SRTS default mode is to take aggregate regional harvest levels and solve for the implicit demand, price, and sub-regional harvest shifts. Market parameters are first used to solve for equilibrium price changes, where the market is defined by all of the included sub-regions. Second, the price and supply shift information from the individual regions is used to calculate harvest change by sub-region. At the aggregate region level, SRTS models year t harvest quantities as determined by the supply function: Q S t = Q S (P t, I t, v t ) And the demand function: Q D t = Q D (P t,z t ). where in the reduced form, current harvests, Q t, are a function of timber prices, Pt, and beginning of period inventory, I t, and other supply and

3 demand shifters (v t, Z t ). A constant elasticity or log-linear functional form is assumed. This structure is consistent with empirical analysis of timber supply (Adams and Haynes, 1996, Newman, 1987). While these studies estimate elasticities at a broad regional level, there is little information on price or inventory elasticities at the sub-regional level. Other factors affecting supply levels (v t ) might include input prices, technological factors such as land quality or management, and landowner characteristics. Some of these issues can be addressed by changing ownership or management type parameters in the model as described below. In harvest exogenous mode, SRTS determines the price and demand curve position in each year of a given harvest level and the supply shift due to modeled inventory changes. The solution sequence proceeds as follows. The region is assumed to start in equilibrium. Since the equilibrium quantity, Q t, and starting inventory, I t, are known, the reduced form equation can be used to solve for P t and the implicit demand shift, Z t. An initial estimate of harvest by sub-region is found by using the same supply specification with the estimated regional price change and sub-regional inventory change to estimate harvest change by sub-region. Because the Cobb-Douglas functional form is not additive, each sub-region s harvest is adjusted proportionately to match regional harvest. The model can be run with the assumption that the sub-regional supply specifications hold and the aggregate price is found by using a binary search algorithm that determines the market -clearing price by summing the supply response across sub-regions and owners. In either top-down or bottom-up mode, demand shifts or equilibrium price trends can be exogenous, and the model will solve for the remaining equilibrium parameters. The runs described below maintained the aggregate market relationship or top-down assumption. These assumptions imply a competitive market with regions and ownership s facing the same price trend. SRTS is not a traditional spatial equilibrium model where a single point with associated transportation costs represents demand. Instead, demand is assumed to be mobile either through shifts in procurement regions (e.g., chip mills) or new capacity (e.g., OSB mills) and is assumed to respond to intraregional differences in stumpage prices. In this formulation, all regions and owners included in model run are assumed to follow the same stumpage price trend, although levels may differ. Harvests will be shifted among owners and subregions based on comparative supply advantages. Inventory Model Structure The internal inventory module in SRTS is based on USDA Forest Service Forest Inventory and Analysis (FIA) timberland area, timber inventory, timber growth rates, and timber removals data. The data are classified into 10-year age class groups by broad species group (softwoods and hardwoods) and forest management type (planted pine, natural pine, oak-pine, upland hardwood, and lowland hardwood). FIA data by species group, forest management type, and 10-year age class are summarized for each relevant region in the analysis. Land area trends by forest management type are exogenous to the model. The SRTS model uses tree and plot level data as a basis for the age and growth analyses described below. Growth SRTS uses 10-year age classes and species/survey unit/owner/ management type cells to account for inventory change. To avoid wide variations or empty cells, the following growth per acre (gpa), regression equation was estimated by species-group (hardwood, softwood), physiographic region (delta, coastal plain, piedmont, mountain), and management type (plantation, natural pine, mixed pine, upland hardwood, lowland hardwood): gpa = f (state, owner, age, owner*age interaction). A cubic age relationship was estimated. This approach allows the shape of the growthage function to be modeled based on data from an entire physiographic/type combination, but allowed the level of growth to vary between states, and the level and shape of the growth curve to vary between owners. In the FIA database, some plots are not assigned ages. For these plots a regression relationship between plot characteristics and age was used to assign ages to the plots.

4 Harvest Harvest in SRTS is handled in three steps. The allocation of regional harvest to a sub-region/owner is based on supply shifts and is part of the market equilibrium calculation described below. Within a sub-region/owner, harvest is allocated across management-types and age-classes based on assigned parameters. Allocation of harvest across the five management types can be related to historical removal proportions, current inventory or growth, or any weighted combination of the above. For example, to allocate removals based on the average of starting removal and current, year t, inventory proportions, a 0.5 weight would be assigned to each. Within a management type, the model can allocate harvest across age classes based on starting harvest proportions, current inventory proportions, or oldest age class first. Weighted average combinations of these procedures can also be specified. Empirical examination of harvest allocations in the FIA data indicate for all management types other than pine plantations, harvest allocations across age classes are highly correlated with the distribution of inventory across age classes. Area Timberland area trends are exogenous to SRTS. The default specification is to apply one set of management type trends to each region/owner combination. For example, a one percent annual increase in pine plantation acreage would be applied to the current plantation acreage in each region. Acres added to a management type begin at age zero. Acres leaving a management type are removed proportionately across all age classes. Growing stock on these acres contributes to current harvest. PnET-SRTS Model Integration PnET model prediction of forest NPP were first derived from historic climate data to develop a historical grid at a 0.5 o x 0.5 o across the southern region. The model is then re-run with various climate scenarios to examine the impact of changing air temperature, precipitation, and atmospheric CO 2 on potential forest productivity for each grid cell. The PnET model only predicts potential productivity because actual stand stocking is not input to the model. The relative climate change impact on forest productivity was calculated as climate scenario productivity/ historic productivity. Ratio values greater than 1.0 indicate that forest productivity will increase for a specific cell under climate change, while values less than 1.0 indicate that climate change will have a negative impact on forest growth. The ratio for each grid cell and year was then combined with the USDA Forest Service Forest Inventory Assessment (FIA) data of stand growth. The FIA divides the southern US into 51 survey units. Individual FIA plot level historic forest volume and growth data is aggregated up to the survey unit scale for analysis. A GIS mask of the survey units is overlaid on the 0.5 o x0.5 o PnET grid of productivity ratios. A Weighted average of productivity in then calculated for each survey unit based on all of the predicted PnET grid cells. This procedure results in a productivity ratio mask at the FIA survey unit scale for each year and climate scenario. RESULTS The results presented below are based on the Phase I SRTS-PnET linkage. The climate change scenarios are the 2x CO 2 climate change scenarios from the following general circulation models shown in Table 1. Since the focus on this paper is the model linkage and spatial diversity of results, only the minimum change and scenarios are discussed. The minimum change scenario simply increases temperature in the model by 2 degrees C and increase precipitation by 20 percent. The meteorological office scenario, by contrast, had temperature increases of up to 10 degrees C. Figure 2 shows the differential impact of the scenarios on species groups. While the scenario implies lower growth, the minimum change scenario increases growth. Hardwoods fare better than softwoods in both scenarios. This average effect masks significant regional variation as shown in Figure 3. Figures 4 and 5 show how growth changes, current harvest levels, and ownership patterns affect shifts in harvest from 1990 to 2030 in these two scenarios. In the scenario, pine harvest shifts to the Mid-Atlantic region and the plantations of the Gulf Coast. In the scenario, however, the severe growth decline in the Gulf Coast region leads to a distinct shift in harvest northward.

5 Figures 6 and 7 show the harvest trends associated with maintaining constant real prices. This gives first-order effects of climate change on harvest trends; subsequent changes in market prices due to changes in supply relative to demand would generate subsequent harvest responses. For softwoods the harvest trends vary significantly across scenarios. Large decreases in harvest are associated with the scenario, while the scenario implies increases in softwood harvest. For hardwoods the variation between scenarios is smaller and all are trending upward, though the scenario show relatively flat trends by SUMMARY AND CONCLUSIONS Results from the linkage of the PnET and SRTS models indicate that there may be wide variation in the impact of climate change across the South. The spatial diversity of forested ecosystems in the South, as well as sub-regional differences in ownership patterns and intensity of management insure that even relatively homogenous climate changes will have varied affects on the landscape. At this time, the system does not include links to other sectors or the global economy, which would be required for a more complete assessment of potential impacts. The model system described is based on biological and economic relationships linked to historical relationships. While these approaches have advantages in predicting near term responses, they are susceptible to structural changes in either the ecological or economic systems. Long-term structural changes may be better addressed by using optimization models of behavior. These models, however, may not model current behavior well and are likely to underestimate the welfare costs of transition by assuming optimal foresight and response of landowners. Our modeling system provides a detailed look of potential variation of impacts within the South based on empirical responses. Together with models that address other sectors and regions, this information will contribute to an assessment of the need for, and potential impact of, climate change policy in the South. in temperate and boreal forest ecosystems, Oecologia 92: Aber, J.D., Ollinger, S. V., Federer, C. A., Reich, P. B., Goulden, M. L., Kicklighter, D. W., Melillo, J. M., Lathrop, R.G. (1995). Predicting the effects of climate change on water yield and forest production in the Northeastern U.S. Climate Res. 5: Adams, Darius and Richard W. Haynes The 1993 Timber Assessment Market Model: Strucutre, Projections, and Policy Simulations. PNW-GTR-368. U.S. Department of Agriculture, Forest Service. Alig, R.J., K.J. Lee, and R.J. Moulton. Likelihood of timber management on nonindustrial private forests: evidence from research studies USDA Forest Service, Southeastern Forest Experiment Station, General Technical Report SE pp. Intergovernmental Panel on Climate Change (IPCC) The Science of Climate Change. Contribution of Working Group I to the Second Assessment Report of the IPCC. J.T. Houghton et al., eds. New York; Cambridge University Press. McNulty, S. G., Vose, J. M., Swank, W. T. (1998). Predictions and projections of pine productivity and hydrology in response to climate change across the southern United States. In: The Productivity and Sustainability of Southern Forest Ecosystems in a Changing Environment (S. Fox and R. Mickler, eds.). Springer-Verlag, New York, NY. Newman, D. H An Economic Analysis of the Southern Softwood Stumpage Market: Forest Science 33: LITERATURE CITED Aber, J.D., and Federer, C.A. (1992). A generalized, lumped-parameter model of photosynthesis, ET and net primary production

6 Table 1 Sources of Climatic Scenarios Abreviation Description GF General Fluid Dynamics Model GI Goddard Institute Model OS Oregon State Univeristy Model United Kingdom Meteorological Office Model Minimum Change Scenario Figure 1. Forest/Ag Modeling System GCM Climate Scenarios Temp/Precip Hydrology Durrans/U Ala PnET-Productivity McNulty/USFS Yields Crop Model Jim Jones/UF Yields SRTS Inventory Model Abt NCSU Murray RTI Inventory Harvest SRTS Stumpage Market Abt NCSU Murray RTI Ag Mgt Model/Auburn Land By MT Mgt Type Allocation Timberland Price LAND USE Parks, Hardie, Wear

7 Figure 2. Average Growth Rate Changes 100% 50% 0% -50% -100% Softwood Hardwood Figure 3. Spatial Diversity in Growth Rate Change Percent Change in Growth Rate: < to to to 60 > 60 Pines

8 Figure 4. Harvest Shifts , Minimum Change Scenario SWLEVEL Figure 5. Harvest Shifts , MO Scenario SWLEVEL

9 Figure 6. Softwood Harvest Trends Based on Constant Real Price Scenario Figure 7. Hardwood Harvest Trends Based On Constant Real Price Scenario Ambient GF GI OS Ambient GF GI OS