Case study: application of a process model to climate change prediction

 

The following case study presents the application of a process model - GroMIT - to predictions of future growth of oak in southern England. The model has been parameterised for the stand in question - the Straits Enclosure, Alice Holt, Hampshire - and has been validated against CO2 and water flux data measured using the technique of eddy correlation at the Straits flux station. Acclimation of physiological processes to rising CO2 levels has been accounted for by adjusting relevant parameters (photosynthesis, respiration, stomatal conductance) for the effects of elevated CO2 that were observed in impact studies at the Headley open top chamber site.

Model Overview

GROMIT (Growth Model of Individual Trees) is a long term model of forest stand growth working at two temporal scales, obtaining annual carbon budgets from a mechanistic hourly flux model (ForestFlux), and scaling these up to simulate long-term growth through allocation patterns based upon the pipe model of Shinozaki. Water balance, and thus limitation of productivity is inherent in the flux model, acting through a simple three pool water supply module limiting conductance. However, nutrient cycling is not included, and thus limitation of the CO2 fertilisation effect by nutrients is not addressed. The output of the model is directed towards forecasting timber production and not ecosystem function. The impact of climate change and atmospheric composition changes on the carbon and water cycles are intermediates in model computations, but the primary objective of these simulations is the accurate prediction of changes in timber production, resulting in the ability to alter forest management tables for forecasting productivity. These predictions are for the short to medium term when placed in the context of ecosystem function, and address commercial rotations over the coming century.

Technical manuals:

GroMIT ^(PDF-347K) 

ForestFlux ^(PDF-535K)  

Climate change scenarios

Baseline climate data are based upon daily meteorological records from the Meteorological Office weather station at Alice Holt. An integral weather generator within ForestFlux, based upon fitting the daily records to a sine function was used to derive hourly data. The weather generator was parameterised and validated using (hourly) data from an automatic weather station for 1996 and 1997 (calibration and validation, respectively). Future scenarios are based upon results from the HADCM2 climate experiment, which have been further down-scaled under the UK Climate Impacts Programme (UKCIP98) to the level of a 10 km grid over the whole of the UK . Monthly means of climatic variables are given for three, thirty year time slices (2010-2039, 2040-2069, 2070-2099), which have been superimposed on the baseline climate data. Note that the UKCIP98 climate change scenarios are less extreme than the more recent ukcip02 scenarios.

Carbon dioxide concentration

The CO2 concentration profile follows the UKCIP98 Medium High scenario. This assumes equivalent CO2 concentration changes are those used in the HADCM2 GGa scenario (1% increase per year). Under this scenario, CO2 is assumed to be responsible for 75% of the total greenhouse gas forcing resulting in mean CO concentrations for the thirty year periods centred on the 2020s (i.e. 2010-2039), 2050s and 2080s of 447 ppm, 554 ppm and 697 ppm, respectively. Past CO2 concentrations are estimated from measurements made in 1935 (307 ppm: Friedli et al., 1989), 1956 (313 ppm: Friedli et al., 1989) and 1978 (342 ppm: Keeling et al., 1979). Linear increases of 1.59 ppm/yr and 1.53 ppm/yr are assumed for the 1980s and 1990s respectively (IPCC, 1995).

Growth simulations

The simulations are carried out for the Straits Enclosure experimental stand which was planted with a mixture of Quercus robur and Q. petraea during the 1930s. Simulations of both height and diameter growth overestimate early growth in the baseline scenario, and diameter growth is underestimated towards the end of the rotation. This is likely to be a result of the representation of the stand as a series of uniform trees; thinning is assumed to be carried out according to management tables (Edwards and Christie, 1981), resulting in uniform spacing in the simulation, rather than the removal of weaker trees as would occur in reality. The stand is a mensuration permanent sample plot, and at the time of the last measurement in 1995 (age 60), mean height, DBH and volume were 18.8 m, 20.6 cm and 0.282 m3 respectively. These values are closer to the model predictions, certainly for height and diameter, (18.3 m, 20 cm, 0.37 m3) than the forest management tables for general yield class six (20.2 m, 24.2 cm, 0.41 m3). The low production of the stand is likely to be a result of sporadic management intervention, with current tree numbers and production similar to those expected for GYC4. These observations therefore highlight the importance of descriptions of management intervention for modelling productivity in commercial forestry.

The 2010 simulation of rising atmospheric CO2 concentration and climate change suggests a relatively large increase in production, with the site index rising from GYC4-6 to GYC6-8. This is likely to be a result of both the lengthening growing season and the CO2 fertilisation effect (CFE). The magnitude of this increase is surprising, given the predicted reduction in summer rainfall (up to 25%), and increases in evaporative demand. The simulations also predict a modest increase in leaf area (mean leaf area index rises from 4.4 to 5.2), which would increase both interception and transpiration losses, making the effects of the predicted droughts more severe. This is compensated for by the reduction in stomatal conductance in response to elevated CO2.

Water Balance

Canopy water use expressed as total transpiration (i.e. ignoring canopy interception and evaporation although these are assessed in the simulations) averaged over the entire rotation lengths was 191 mm/yr and 220 mm/yr for the baseline and 2010 oak simulations, respectively; as a consequence, the moisture content of the top 50 cm of soil during July and August was lower in the 2010 simulation than the baseline rotation average (23% vs. 16%), although these predictions for soil moisture content also include losses resulting from increased canopy interception. This indicates the magnitude of the predicted increase in water use, and also highlights the increased soil moisture deficits that are predicted. However, since mortality is assumed to remain unchanged, these simulations do not address the problem of tree death resulting from long periods of drought. The fact that soil moisture content showed a large reduction does indicate a significant perturbation in the soil moisture regime. The importance of modelling water use under conditions of climate change is further emphasised by the fact that a simulation in which there was no stomatal closure to rising CO2 did not show a reduction in yield as compared to the 2010 rotation. This suggests that in these simulations, reduced carbon assimilation as a result of water limitation is compensated for by higher rates of assimilation (due to a lack of response to CO2) when water is not limiting, thus potentially exacerbating summer droughts. Although evidence from a large number of studies indicates significant reductions in stomatal conductance at elevated CO2, there is also some evidence that for some species, elevated CO2 reduces stomatal sensitivity to soil moisture deficits and VPD. This last simulation in which there is no stomatal response to CO2 may therefore not be unrealistic for some species.

Conclusions

The model simulations of the growth of oak in southern England shown here suggest that during the 21st century, there is likely to be an increase in productivity, raising site indices. This may already have occurred in the latter half of the 20th century, although it is difficult to assign changes in productivity to one factor, since there have been concomitant changes in management practice, silviculture and nitrogen deposition. Climate change scenarios suggest that in the south of England, water resources may become limiting, and consequently it is essential that accurate predictions of water use by trees are available, and that we are able to model the effects of these predicted water deficits on tree growth and their ability to survive. The model simulations presented here address potential assimilation and growth based solely on climatic and atmospheric changes, but do not predict the effects of stochastic events such as drought. In addition, the many indirect consequences of environmental and climatic change have not been addressed such as changing insect and disease epidemiology, nor has nutrient availability, which may become limiting as a result of increased growth rates, management practice or reductions in atmospheric deposition.

The modelling work described here was the first step to an enhanced modelling capability that has developed ForestEtP and ForestGrowth within the Core Model Programme of Biometrics Division. These models will be applied to climate change predictions in the near future. ForestFlux has also been applied to, and integration with, data collected as part of the Level II programme in a pilot study on 'upgrading the Level II protocol for physiological modelling of cause-effect relationships' ^(PDF-2137K) including an assessment of the likely effects of climate change.