3. Material and methods
3.2 Observations
Four land cover maps were adopted for REMO simulations in this study (Table 1). In Paper I,
20
version 2.0 of Global Land Cover Characteristics Database (GLCCD) was replaced with the 2006 version of the Corine Land Cover (CLC) for the majority of the Fennoscandinavian domain in REMO, excluding Russia and Belarus where the CLC does not cover. In Paper II, the 1st Finnish National Forest Inventory (FNFI1) and the 10th Finnish National Forest Inventory (FNFI10) were implemented in REMO in order to study the biogeophysical effects of peatland forestation on climate conditions in Finland. The implementation of FNFI maps were based on the work that was done in Paper I.
Table 1: Summary of land cover maps used in this study.
Map Time tag Spatial No. of land
cover types Reference
Coverage Resolution Paper I
GLCCD version 2.0 (Global Land
Cover Characteristics Database) 1992-1993 Global 1 km 74 U.S. Geological Survey
(2001) CLC 2006 (CORINE Land Cover) 2006+/-1 Most areas of
Europe 100 m 44 European Environment
Agency (2007) Paper II
FNFI1 (1st Finnish National Forest
Inventory) 1921-1924 Finland 3 km 10 Ilvessalo (1927);
Tomppo et al. (2010) FNFI10 (10th Finnish National Forest
Inventory) 2004-2010 Finland 3 km 10 Korhonen et al. (2013)
The standard land cover map employed in REMO is GLCCD. Based on AVHRR satellite data from April 1992 to March 1993 mainly, the U.S. Geological Survey (1997) constructed GLCCD (version 1.0) to display the global distribution of major ecosystem types (Olson, 1994a, b) at a 1 km horizontal resolution (Loveland et al., 2000). Later, land cover classes that cover 10% of the global land area in GLCCD (version 1.0) were revised and GLCCD (version 2.0) was produced (U.S. Geological Survey, 2001). Similarly, CLC 2006 is a land cover map based on satellite images, which were measured with the HRVIR and LISS-III instruments in 2006 ± 1 year (European Environment Agency, 2007). The horizontal resolution of CLC 2006 is 100 m, and 44 land cover classes were included. The FNFI1 describes the land cover of Finland in the 1920s before peatland forestation (Ilvessalo, 1927; Tomppo et al., 2010), while the FNFI10 represents the present-day land cover in the 2000s (Korhonen et al., 2013). In Paper II, we substituted the CLC with FNFI10 to describe the present-day land cover, in order to avoid uncertainties in comparing land cover maps with different land cover classification methods and different spatial resolutions. The differences between FNFI1 and
21
FNFI10 show the largest historical changes on land cover in Finland due to peatland forestation, which has started at 1950s. According to the FNFI series, the area of undrained mires was 8.83 million ha at 1951-1953 (FNFI3), 4.32 million ha at 1986-1994 (FNFI8), 4.14 million ha at 1996-2003 (FNFI9) and 4.00 million ha at 2004-2010 (FNFI10) (Päivänen and Hånell, 2012). Both the FNFI land cover maps are at a 3 km resolution and include 10 land cover classes that follow the CLC nomenclature. Both FNFI1 and FNFI10 land cover maps are post-products that were especially prepared for this study from the respective FNFI field measurement data (see detailed description of the procedures in Appendix A in Paper II).
The FNFI land cover maps are at a 3 km resolution and include 10 land cover classes that follow the CLC nomenclature, where the land cover type “peat bogs” is defined as naturally treeless peatland and mires where the stocking level is low or the mean height of trees is below 5 m at maturity.
In order to utilise the existing land surface parameters for the default land cover types, translations of the land cover types in the newly introduced land cover maps to the Olson land cover types in GLCCD (version 2.0) have been conducted through comparing the definitions and matching the surface characteristics of land cover types. It is obvious to find appropriate analogues for some land cover types; for instance, matching the coniferous forest, mixed forest and broad-leaved forest in FNFI maps with conifer boreal forest, cool mixed forest and cool broadleaf forest in GLCCD, respectively. However, for some land cover types, such as transitional woodland/shrub in FNFI maps, it is not straightforward to find correspondence land cover types in GLCCD, and GLCCD land cover types with suitable land surface parameters were adopted. All the translations are listed in Table 1 in Paper I and Table B1 in Appendix B in Paper II.
3.2.2 Meteorological observations and ecosystem flux data
A number of meteorological observations were used in this work (Table 2). The E-OBS gridded observational data (Haylock et al., 2008) were adopted in Paper I and Paper II for model evaluation. The E-OBS dataset covers the area between 25 °N - 75 °N and 40 °W - 75 °E. It is a daily high-resolution gridded dataset that aims to provide the best estimate of
22
gridbox values rather than point values to enable direct comparison with the results from regional climate models. The dataset has five elements that include daily mean temperature, daily minimum temperature, daily maximum temperature, daily precipitation sum and daily averaged sea level pressure. It has been found that the uncertainty in E-OBS data is largely dependent on the season and number of observations. Paper I assessed the simulated mean monthly/seasonal maximum and minimum 2-m air temperatures, diurnal temperature range and precipitation with E-OBS data (version 7.0). In Paper II, the temperature trends over 40 years (1959-1998) for March and April were calculated based on monthly mean daily maximum and monthly mean daily minimum surface temperatures over Finland from E-OBS data (version 10.0), so as to compare with the simulated effects on surface temperature in spring from peatland forestation.
Moreover, the gridded meteorological data compiled by the Finnish Meteorological Institute (FMI gridded observational data; Aalto et al. (2013)) from site observations in Finland were used as the baseline climate for the bias correction of JSBACH forcing data, and as inputs for the calculation of observation-based drought indicators (Paper III). The data contain daily mean temperature, daily minimum temperature, daily maximum temperature, precipitation, relative humidity and incoming shortwave radiation on a 0.2° longitude × 0.1° latitude grid over Finland.
Meteorological data at the three sites were used as meteorological forcing for site-level simulations by JSBACH, and in Paper III the site measured soil moisture were compared with the simulated soil moisture. Two of those three sites were also studied in Paper IV with GPP and ET fluxes derived from EC measurements.
In addition, ERA-Interim reanalysis data (Simmons et al., 2007) was used to drive REMO in Paper I and Paper II, and the 10-m wind speed of ECWMF ERA-Interim reanalysis data was used in Paper III to calculate the reference evapotranspiration (ET0) from the Penman-Monteith equation (Allen et al., 1994).
23
Table 2: Summary of meteorological observations and ecosystem flux data used in this study.
Data Spatial coverage or
The forest drought damage percentages in Finland of forest health observation data from 2005 to 2008 in Muukkonen et al. (2015) were adopted in Paper IV. The forest health observation data are products of the pan-European monitoring programme ICP Forests (the International Co-operative Programme on the Assessment and Monitoring of Air Pollution Effects on Forests). Forest drought damage symptoms have been identified since 2005 in Finland through visual inspections, following internationally standardized methods (Eichorn et al.,
24
2010) and national field guidelines (e.g. Lindgren et al., 2005). The inspections have been carried out at forest stands during July and August annually by 10-12 trained observers in Finland. A drought damage site was recognized when a single sample tree in a study site showed drought symptoms. Therefore, uncertainties in the data can rise from subjective interpretations and inappropriate time point of the visual inspections. In the summer of 2006, 24.4% of the 603 forest health observation sites over entire Finland showed drought symptoms, in comparison to 2-4% drought damaged sites in a normal year. Most of the drought damaged sites located in southern Finland, totalling to 30% of the observation sites in southern Finland.
3.3 Studied indicators