Rossby Centre is developing a set of high-resolution calibrated climate indices that will help to climate-proof the Swedish and Finnish forestry sector. The forest sector operates under lead times of 50-100 years, or more, when it comes to the rotation period of forest trees commonly used in Scandinavian forestry. To be competitive, seedlings planted today both have to have maximum survival rate under present climatic conditions and show optimal growth in the future climate.
The Forest Research Institute of Sweden, Skogforsk, is together with partner institutions currently developing new production functions and survival functions for Scots pine (Pinus sylvestris L.) for Sweden and Finland. These statistical functions are used to estimate biomass production – growth – and survival rate of pine seedlings of different origin (provenance and seed orchard) based on location and climatic conditions. Fundamental for the new approach in developing these functions is availability of high-resolution gridded observational datasets of the required meteorological variables. These reference datasets serves two purposes. Firstly, they are used to derive the relevant climate indices used to develop the functions for production and survival under present day climate. Secondly, they are used to calibrate the climate scenario ensemble so that there is no noteworthy inconsistency (“jump”) when moving from the reference data period into the scenario data period.
We use an updated and improved version of the software tool implementing the DBS bias-correction method (Yang et al., 2010) partly developed with support from the EU-Genesis project. In essence, the DBS method implements a parametric quantile-quantile mapping in which the temperature is conditioned on precipitation occurrence.
For Sweden the reference dataset has a spatial resolution of 4×4 km, and for Finland the resolution is 10×10 km, and the focus is on climate indices derived from daily mean temperature. The following indices were selected for use in the statistical analyses: monthly mean temperature for January, February and July (used to calculate a continentality index); start, end and length of the vegetation period (Figure 1a); in addition to growing degree-days (>+5°C) during the vegetation period. In this note we use the length of the vegetation period to illustrate the results.
For the future we used an ensemble of regional scenarios produced by our regional model RCA3 forced by six different global climate models (GCMs) representing the SRES A1B emission scenario. This ensemble is consistent with one of the ensemble datasets that SMHI presents in the web-based climate scenario tool (Swedish only). The regional climate model and its performance have been extensively described in an open-access special issue of the scientific journal Tellus A (Jones et al., 2011). A comparatively small bias in daily mean temperature may, however, have a strong effect on climate indices that involve thresholds. This is a well-known fact which is clearly illustrated in Figure 1b. It is, however, worth pointing out that because the reference data has higher resolution than the scenario data (~50×50 km) a preliminary step was to apply a correction for the altitudinal difference between the model grid and the reference data grid. As this correction makes use of the simplistic assumption of a vertical temperature gradient of 0.0065 K/m according to the standard atmosphere, some of the bias may in fact be introduced at this step rather than being a genuine model bias. In the mountainous region, where the altitude corrections are largest, this effect is likely a major component of the substantial biases seen in the region. Nonetheless, after calibrating the daily mean temperature using the reference datasets the bias is substantially reduced (Figure 1c) and stays within a margin of a few percent.
Interestingly enough, while calibration is clearly necessary when looking at the performance of the ERA-40/ERA-interim downscaled data (Figure 1b-c), the calibration makes less difference when considering the ensemble mean for a future period (Figure 1d-e). Here model output daily mean temperature of each ensemble member was independently calibrated to the reference datasets. An explanation of this is that a bias is largely constant and thus cancels out when taking the time difference. However, a careful look at differences between the two maps shows that the calibration step substantially reduces the remnants of the large-scale gridcell pattern carried over from the RCM grid.Rossby Centre is carrying out this work within the framework of the Swedish research programme Mistra-Swecia which is hosted by SMHI. A focus area for the programme is research on how climate change influences forestry, the forest sector as well as land use in the natural and managed environment.
The Rossby Centre is carrying out this work within the framework of the Swedish research programme Mistra-Swecia which is hosted by SMHI. A focus area for the programme is research on how climate change influences forestry, the forest sector as well as land use in the natural and managed environment.
References
- Jones, C.G., Samuelsson, P., Kjellström, E. 2011: Regional climate modelling at the Rossby Centre. Tellus A, 63(1), 1-3.
- Yang, W., Andreásson, J., Graham, L.P., Olsson, J., Rosberg, J. and Wetterhall, F. 2010: Distribution-based scaling to improve usability of regional climate model projections for hydrological climate change impacts studies. Hydrology Research, 41 (3–4), 211-228.