Daily values of minimum and maximum temperature, sunshine duration, humidity and wind speed for more than 1100 weather stations for period 1980-2011 were derived from the WebWerdis portal of the German Meteorological Service DWD (DWD, 2012a; DWD, 2012b; DWD, 2012c). In addition, the portal provided access to daily gridded precipitation at 1 km × 1 km resolution (Regnie data set) and to grids of monthly mean values of daily sunshine duration, daily minimum temperature, daily maximum temperature and daily mean temperature. These grids were developed by the DWD by interpolation of weather station data using a digital elevation model to support the interpolation (DWD, 2014). Daily values for temperature and sunshine duration Xgrid,d (°C) were
computed for each 1 km × 1 km grid cell and for each day of the period 1980-2011 by using a procedure described in Zhao et al. (2015) as
m ws m grid d ws d grid X X X X , , , , (1)
where Xws,d was the daily value measured at the nearest DWD weather station, Xgrid,m was the
monthly mean at the grid cell according to the 1 km × 1 km grid and Xws,m was the monthly mean
at the nearest weather station. Use of this procedure ensured that the monthly mean value was equal to the value computed by the DWD for each grid cell in the 1 km × 1 km grid, while the day- to-day variation was equal to the variation reported for the nearest weather station (Siebert and Ewert, 2012). Daily solar radiation was then calculated from daily sunshine duration by using the Ångström–Prescott approach (Almorox and Hontoria, 2004). Extra-terrestrial radiation was calculated according to Allen et al. (1998) while the Ångström coefficients a and b were computed by comparing, on sunny and overcast days, incoming shortwave radiation derived from satellite imagery to computed extraterrestrial radiation. Daily mean incoming shortwave radiation (W m-2) and daily mean fractional cloud cover (%) were derived from the Satellite Application Facility on Climate Monitoring (CMSAF, 2012a; CMSAF, 2012b) and analyzed for period 2005-2012. Daily
66
wind speed was calculated by averaging daily mean wind speed across the weather stations of the DWD network. In total, 378 weather stations measured wind speed in period 1980-2011 but only stations with a measuring height of maximal 20 m above ground and an altitude of not more than 900 m. were considered when calculating the mean across stations, so that the number of stations considered in this study was 236. Measured wind speed was corrected to a sensor height of 2 m according to Allen et al. (1998) as
) 42 . 5 8 . 67 ln( 87 . 4 2 z u u s (2)
where u2 was the wind speed in 2 m height (m s-1), us the wind speed at the sensor (m s-1) and z the
sensor height (m). The stations were selected because wind speed was measured there in an appropriate height and because the stations were located on or close to cropland. The calculation procedure resulted in wind speed that was similar for all grid cells in Germany but varied from day to day.
4.2.2.2. Soil properties at high resolution
Maximum rooting depth and volumetric water content at full saturation, field capacity and wilting point were derived from the Bodenübersichtskarte (BÜK) 1000 N data set developed by the Federal Institute for Geosciences and Natural Resources, BGR (BGR, 2013). The soil data set was developed in the period 2000 – 2007 by combining soil information with land use information derived from the Corine land cover classification. The BÜK 1000 N distinguishes 71 soil mapping units, each of them in 5 climatic zones and for the land uses “cropland”, “grassland and heterogeneous agricultural land” and “forest” (BGR, 2013). In addition, the BGR provided descriptions of representative soil profiles with up to 12 soil layers for each of the soil mapping units, in each climate zone and each land use type as MS-Access databases (Dr. Andreas- Alexander Maul, B4.2 Geodaten, Geologische Informationen, Stratigraphie, personal communication). To derive maximum rootable soil depth it was assumed that roots cannot grow into layers with the following properties/horizon designation suffixes:
- r (low oxygen content)
67
- Cn (unweathered rock)
- d (extremely dense, high bulk density)
- C-horizons derived from parent materials marls, sandstone, moraines loams, shale, lime stone, granite, and basalt,
- C-horizons with more than 50% -Vol gravel or stones.
It was also assumed that roots cannot grow into layers below such a non-rootable layer. Furthermore maximum rooting depth for winter wheat was constrained to 1.2 m, even when the soil properties would allow for a larger rooting depth.
4.2.2.3. Emergence date at high resolution
Emergence day of winter wheat was obtained from the phenology database of the German Meteorological Service DWD provided by the WebWerdis portal (DWD, 2013). The phenology database contained 91230 observations for period 1950-2009, of which 72507 were selected after application of an outliers filtering procedure. The filtering procedure computed for each of the 89 DWD eco-regions the mean and standard deviation of the observations and filtered out all records deviating from the computed mean more than ±2 times the standard deviation (Siebert and Ewert, 2012). From these 72507 records we selected 39680 records at 3018 locations for the period 1980- 2009. To reduce the effect of observations in years with uncommon phenology we selected the records for 2243 locations with at least 5 years of observations and computed for each of them the mean emergence day. The point observations were then interpolated to a grid at 1 km × 1 km resolution by using the Inverse Distance Weighted (IDW) method with a power of 2 and considering 12 neighbor points (Mabit and Bernard, 2007). Consequently, the calculated emergence day was similar for all years but varied across grid cells.
4.2.2.4. Aggregation of the input and output data to lower resolutions
Weather data, soil properties and emergence day was aggregated from 1 km × 1 km resolution to 10 km × 10 km, 25 km × 25 km, 50 km × 50 km and 100 km × 100 km resolution. First a cropland mask was applied to avoid that weather and soil properties in mountainous regions, which are mainly covered with forest or grassland, impact the crop model input data. At 1 km × 1 km resolution, grid cells were masked out and not considered in subsequent calculations when no cropland or mosaic of cropland and other land cover was contained according to the Corine land
68
cover 2006 (EEA, 2010). Daily weather data were then aggregated to lower resolution by computing the mean of the values of the 1 km × 1 km grid cells contained in grid cells of larger extent. Soil data were aggregated by selecting the properties of the dominant soil (the soil type with the largest extent on cropland in each specific grid cell). Emergence day was computed for each grid cell as the mean of the emergence observations made in the respective grid cell. In particular at the higher resolutions it happened frequently that grid cells did not contain any phenology observation point at all. Then the mean of the emergence day of the enclosed 1 km × 1 km cropland grid cells was computed. The application of the cropland mask and the aggregation methods described before resulted in 231,601 grid cells for Germany at 1 km × 1 km resolution, 3440 grids at 10 km × 10 km resolution, 609 grids at 25 km × 25 km resolution, 171 grids at 50 km × 50 km resolution, and 51 grids at 100 km resolution by 100 km resolution. The aggregation of output data was performed by averaging high resolution model outputs such as heat and drought stress reduction factors or crop yield at coarse resolution.
4.2.2.5. Crop yield data
To evaluate the crop model, winter wheat yield at district level was derived for period 1999-2011 from the Regional database of the German Federal Statistical Office (DESTATIS, 2013). There are 404 districts in Germany but several of them (e.g. larger cities) do not contain winter wheat growing areas, reducing the number of districts with reported winter wheat yields to 345.