SEARCH

SEARCH BY CITATION

Keywords:

  • climate change;
  • degree-days;
  • Europe;
  • grapevines;
  • temperature;
  • trends

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References

The global increase of temperature, together with more frequent severe winters and summer heat waves may lead to a change in energy consumption and agricultural production. Cooling, heating, and growing degree-days (CDD, HDD, and GDD), respectively, are used to quantify the energy needed to condition or heat buildings, and to study the growing season. Using a new dataset made of 4023 daily TNTMTX series for the period 2001–2011 and 3897 monthly TM homogenized series for the period 1951–2011, we computed CDD, HDD, GDD, and Winkler Index (WI) for Europe. We developed a model that correlates degree-days calculated with daily TNTMTX data with degree-days obtained by monthly TM data, in the overlapping period 2001–2011. A set of parameters for each station was then applied to the corresponding 1951–2011 monthly records. We interpolated the parameters and the reconstructed degree-day series onto a European 0.25° × 0.25° grid: with these gridded parameters, one can estimate the degree-days for any European location if only monthly TM is available. We present maps of HDD, CDD, GDD, and WI for the period 1951–2010. To validate them, we run a comparison in the Carpathian area using an independent dataset (from the CARPATCLIM project). The regional records show high correlations, especially for HDD (r > 0.99) and WI (r > 0.98). Subsequently, we performed a linear trend analysis on European and regional basis. HDD showed a significant decrease almost everywhere in Europe, whereas CDD, GDD, and WI showed a significant increase in particular in the last 30 years in the Mediterranean region. Moreover, WI indicated that new areas in France and central Europe became suitable for grape cultivation in the last decades.


1 Introduction

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References

One of the main issues related to climate change is the temperature rise which involves Europe and in particular the Mediterranean area (Parry et al., 2007). A set of temperature-based variables, i.e. degree-day indicators, is usually computed to account for climate effects upon energy consumption (Christenson et al., 2006), agricultural production (Grigorieva et al., 2010), and crop phenology (Moriondo and Bindi, 2007).

A degree-day is a fictitious quantity defined as the deviation, in degree celsius (°C), from a reference temperature value (UK Met-Office, 2013). Degree-days represent the number of degrees by which the temperature has gone above or below a threshold. We distinguish between cooling degree-days (CDD), heating degree-days (HDD), and growing degree-days (GDD). CDD and HDD reflect the energy needed to cool and heat a building, respectively. CDD account for temperatures above a base value (from April to September, in the Northern Hemisphere) and HDD for temperatures below (from October to March). If autumn and winter are cold, this results in high HDD values, and if spring and summer are hot, this results in high CDD values.

GDD is an agroclimatic indicator related to the growth cycle of plants, fruits, and crops, on which the base temperature and the calculation period are based. For most of the crops, the growing season length (GSL) can be alternative to GDD, and when applied to grapevines (Vitis vinifera), GDD are often substituted by the Winkler Index (WI; Winkler et al., 1974; Santos et al., 2012) or the Huglin Index (HI; Huglin, 1978). GDD have also been applied to the cycles of pests, worms, parasites (Pruess, 1983), and the pollen sources (Jato et al., 2000).

This article has three principal aims: the description of a new model that allows the calculation of degree-days if only monthly mean temperature (TM) data are available, the construction of climatologies of CDD, HDD, GDD, and WI across Europe for the last six decades, and the investigation of their evolution and trends.

The article is organized as follows: Section 'Data' deals with the new datasets. Section 'Theory and methods' deals with the equations to compute the degree-day variables, the new model that estimates degree-days from monthly TM, and a validation in the Carpathian region using an independent dataset provided by the CARPATCLIM Project (Szalai and Vogt, 2011). Section 'Results and discussion' deals with the European degree-day climatologies, the trends at European and regional scale, and an application of WI to grape cultivation. Section 'Conclusions' deals with a summary of the key outputs and future developments.

2 Data

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References

2.1 Collecting European daily temperature series from three datasets

The agrometeorological Monitoring Agricultural ResourceS (MARS) database, hosted by the European Commission's Joint Research Centre (JRC), includes more than 5000 weather stations distributed across Europe and bordering Asian and African regions. Daily values of many variables, including temperature, are collected via the Global Telecommunicating System (GTS) and complemented by monthly updates from national and local providers.

The daily values are automatically quality checked, but they are not homogenized: gaps and outliers are present. To improve the data availability and quality, we collected temperature data from two additional sources: the Global Historical Climatology Network dataset of the National Climatic Data Center of the US Department of Commerce (NCDC-GHCNv3; Menne et al., 2012) and the European Climate and Assessment Dataset of the Royal Meteorological Institute of the Netherlands (KNMI-ECA&D; Klein Tank et al., 2002).

2.2 The daily complete temperature dataset for 2001–2011

After a preliminary analysis, we found that the availability of daily temperature data is greatest in 2001–2011, so we initially focused on this period. We collected minimum and maximum temperature (TNTX) data: 4487 stations from JRC-MARS, 3488 from KNMI-ECA&D, 4169 from NCDC-GHCN and its subset, the Global Summary of the Day (NCDC-GSOD).

After a quality check based on geographic and other metadata, we removed the overlapping stations and blended those whose relative distance is less than 7.5 km, provided that the elevation gap is smaller than 50 m. In the blending procedure, we followed a priority order based on the overall data quality: KNMI-ECA&D, NCDC-GHCN, and finally JRC-MARS.

After the blending, we obtained 7127 TNTX stations with 80.6% of valid data in 2001–2011, so we performed a completing procedure on the available records for the period 2001–2011. Whenever a datum was missing, we completed it using data from up to five ‘most similar’ stations, but without building the reference series. Such a procedure was based on an interpolation model which uses seven Gaussian weights: horizontal distance, distance from the coast, climate barriers, slope, aspect, angular distribution, and top/valley discrimination. A station record has been used to complete another one only if they mutually respected all the thresholds established for the seven weights.

We eventually obtained 4637 TNTX stations, 4023 of them are 100% complete for the period 2001–2011. We computed the daily TM series by summing TN and TX and dividing by two.

2.3 The monthly homogenized temperature dataset for 1951–2011

A similar completing procedure could not be applied to long-term daily records, due to the low data availability before 1971, thus we transformed the 1951–2011 daily records into monthly records if at least 80% of the data were available for each considered month. In order to complete the 1951–2011 monthly records, we also considered the gridded E-OBS data (v8.0, resolution 0.25° × 0.25°; Haylock et al., 2008) as input when it was impossible to use only the station data. In the completing procedure, we used two weights: the 3D distance and a monthly weight derived from comparing the data between the series over the entire overlapping period. This weight substituted the weights based on geographic and orographic features because a direct comparison over the data must include all the climatological features.

We reconstructed 4012 monthly and complete TNTXTM records for the period 1951–2011. Only 3897 passed the homogeneity tests based on the Multiple Analysis of Series for Homogenization software (MASH; Szentimrey, 1999): their spatial distribution is shown in Figure 1. Overall, the density is approximately 1/2500 km2 and the mean inter-distance between stations is 51.9 km.

image

Figure 1. Spatial distribution of the TNTX stations used for the period 1951–2011.

Download figure to PowerPoint

The final combination of the Gaussian weights performed very effectively; in Table 1, we report the statistics of the completing procedures, calculated with a jackknife cross-validation (Efron and Gong, 1983). For the 2001–2011 daily values, the mean absolute errors (MAE) were approximately 1.5 °C, but they reduced to 0.4 °C if the derived monthly records were compared. The homogenized 1951–2011 monthly records showed low MAE (≈0.5 °C) and low root mean squared errors (RMSE ≈ 0.8 °C).

Table 1. Cross-validation errors of the completing procedure: DD is for daily, MM for monthly, and HOM for homogenized monthly data. ME is for Mean Error, MAE for Mean Absolute Error, and RMSE for Root Mean Squared Error.
Period01–1101–1151–11
DataDDMMHOM
Error (°C)MEMAERMSEMEMAERMSEMEMAERMSE
TN0.01.41.90.00.40.70.00.50.8
TX0.01.62.00.00.40.70.00.50.8

3 Theory and methods

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References

3.1 Degree-days: equations and base temperatures

HDD and CDD can be computed considering daily TM, daily TN and TX (Matzarakis and Balafoutis, 2004; Hassan et al., 2007) or more advanced models (Allen, 1976). Because some impacts depend on the number of hours a temperature deviates from the baseline and some on the days a temperature is above or below the baseline, we must acknowledge that different approaches may lead to very different outputs. To obtain HDD and CDD, we followed the approach adopted by the UK MET-Office: it is based on the fraction of a day that exceeds a baseline (Equations (1)(2)(3)(4)), whereas conventional measures are based on whether TM is above or below the baseline.

  • display math(1)
  • display math(2)
  • display math(3)
  • display math(4)

We computed HDD from the 1st of October to the 31st of March and CDD from the 1st of April to the 30th of September. To obtain the annual values, we summed the daily values. The base temperature (Tb) is 15.5 °C for HDD and 22 °C for CDD. The relationship between degree-days and temperature is not unique (Thom, 1954) and different Tb can be chosen, according to the scope and the region considered. For example, Buyukalaca et al. (2001) set different Tb for CDD (18–28 °C) and HDD (14–22 °C) for Turkey; Valor et al. (2001) chose 10 °C for HDD and 25 °C for CDD for Spain; Papakostas et al. (2010) set various Tb (10–20 °C for HDD and 20–27.5 °C for CDD) for Greece; Wibig (2003) reported that in Canada a unique Tb (18 °C) is frequently used, as in the United States (65°F).

The choice of Tb depends on local climate and applications, should they refer to electricity production, energy budgets, or agroclimatic impacts. Dealing with an extended area like Europe, it is difficult to choose universally ‘correct’ Tb. We accepted the Tb suggested by the UK MET-Office as they fit a physically reasonable model based on TN and TX. In fact, setting Tb = 15.5 °C in Equation (1) implies notable values only if TM ≤ 15.5 °C which corresponds to TX = 20°C at European mid-latitudes considering an average daily winter temperature range of 9 °C (computed using our station data), thus it is assumed that the heating is turned on when the daytime external temperature does not go above 19–21 °C. In contrast, setting Tb = 22°C in Equation (3) implies notable values only if TM ≥ 22 °C which corresponds approximately to TX = 29 °C at European mid-latitudes considering an average daily summer temperature range of 14 °C (computed using our station data), thus it is assumed that the air conditioning is turned on when the daytime external temperature goes above 28–30 °C.

Although sophisticated – but rarely applied – approaches exist (Yang et al., 1995; McMaster and Wilhelm, 1997), we computed GDD and WI using only TM, as it is commonly done (Project team ECA&D, 2013, where a different Tb is chosen).

  • display math(5)
  • display math(6)
  • display math(7)
  • display math(8)

We computed GDD from the 1st of March to the 31st of October, and WI from the 1st of April, because the grape growth cycle usually starts in April (Jones and Davis, 2000). Regarding GDD, we set Tb to 5.5 °C because it is the lowest Tb for the most important crops in Europe: 5.5 °C applies to wheat, barley, rye, oats, and lettuce, 8 °C to potatoes, 10 °C to American maize, rice, corn, and tomato (McMaster and Wilhelm, 1997; Miller et al., 2001). Regarding WI, Tb was set to 10 °C as suggested by Winkler et al. (1974).

While HDD, CDD, and WI are usually calculated using a baseline temperature (Tb), GDD can also be computed using an upper temperature threshold (Tu), which is very important for the development of certain crops (e.g. Berry and Bjorkman, 1980; Russelle et al., 1984; McMaster and Smika, 1988). A model based on two thresholds (Tb and Tu) would have probably produced better results for GDD than our model based only on Tb, but we preferred using the same methodology for all degree-day variables described in this article.

3.2 A simple model that allows degree-days to be obtained using only monthly data

Computing long-term degree-day records can be difficult; in fact, if a single daily temperature datum is missing, the annual degree-days cannot be calculated. If a daily temperature dataset is not available or it has missing data, degree-days can be estimated using other approaches, e.g. annual temperature profiles (McMaster and Wilhelm, 1997), but in general such methods are not accurate (Roltsch et al., 1999). We describe, step by step, a new methodology that allows the degree-day variables to be calculated using only monthly TM data. In the description, we refer to HDD, but it applies to CDD, GDD, and WI.

  • 1.
    Using the daily TNTMTX dataset, we computed the monthly and annual HDD for 2001–2011.
  • 2.
    From the same daily TNTMTX dataset, we obtained the corresponding monthly TM series.
  • 3.
    We inserted the monthly HDD (as obtained in step 1) and the corresponding monthly TM for the period 2001–2011 (obtained in step 2) in:
  • display math(9)

We performed a multiple linear regression (MLR) and obtained a set of parameters (α, β, γ) for each station with the ordinary least squares (OLS) method.

  • 4.
    We inserted the monthly TM (1951–2011) and the corresponding set of parameters (obtained in step 3) in Equation (9) in order to obtain monthly HDD for 1951–2011. We did this for the 3835 stations that are complete for both daily TNTMTX (2001–2011) and monthly TM (1951–2011).
  • 5.
    We transformed the monthly HDD (obtained in step 4) into annual HDD for 1951–2011.
  • 6.
    We compared the annual HDD obtained with our method (step 5) with the same annual HDD directly obtained from the daily data (step 1) in the overlapping period 2001–2011.

The complete procedure has been repeated using different regression models in Equation (9). We tested linear, parabolic, and hyperbolic functions with combinations of TN, TX, and TM as inputs. The validation described in step 6 has been performed to choose the regression model which best estimates the degree-day variables. The statistical errors, averaged over all the 3835 stations, are shown in Table 2 and prove the validity of the procedure. In general, the errors are very small and the MAE are lowest with a parabolic spline. Consequently, we have chosen the parabolic model based on TM to compute the degree-day series and the climatologies.

Table 2. Statistical errors (%) of the model used to estimate degree-days with monthly TM.
ModelLinearParabolicHyperbolic
InputTMTNTMTXTM
Error (%)MEMAERMSEMEMAERMSEMEMAERMSEMEMAERMSEMEMAERMSE
HDD0.2 1.0 1.2−0.12.02.50.00.60.80.12.22.7−0.1 3.9 5.9
CDD4.611.614.0−0.411.814.6−0.45.77.0−0.45.26.40.512.826.7
GDD0.9 1.4 1.7   0.10.81.0   0.2 3.2 4.3
WI1.0 2.0 2.5   0.10.91.1   0.4 4.5 6.8

Subsequently, we interpolated the annual HDD, CDD, GDD, and WI on the 0.25° × 0.25° European grid using a Kriging with External Drift scheme (KED; Hengl, 2009), with the smoothed elevation as external drift. Simultaneously, we interpolated the station parameters of the parabolic splines onto the same European grid. The grids can be requested from the authors.

By applying the procedure described above, one can obtain degree-days related to any period in which monthly TM data are available, provided that approximately 10 years of daily data are available to derive the parameters (α, β, γ). Otherwise, one can use the gridded parameters to obtain the degree-days for any European location if the corresponding monthly TM is available.

3.3 A validation of our model in the Carpathian Region

Before applying our methods to Europe, we tested them in the Carpathian area. We compared HDD, CDD, GDD, and WI as obtained with our reconstruction model and with an independent dataset, provided by the CARPATCLIM project (Szalai and Vogt, 2011).

The CARPATCLIM country members worked together with the Joint Research Centre to create a digital climate atlas of the Carpathian Region for the period 1961–2010, based on a daily gridded database of 16 meteorological variables (www.carpatclim-eu.org). The spatial resolution is 0.1° × 0.1° and the area is delimited by 17°–27°E and 44°–50°N, excluding Bosnia-Herzegovina. The temperature grids have been compiled starting from a station network of 258 stations and all the daily records have been quality checked, completed, and homogenized with the MASH (Szentimrey, 1999). The station data have then been interpolated onto the 0.1° × 0.1° grid with the Meteorological Interpolation based on Surface Homogenized data software (MISH; Szentimrey and Bihari, 2007), based on a regression-Kriging scheme with spatial distance, elevation, and the AURELHY principal components as auxiliary variables (Benichou and Le Breton, 1987).

To obtain the annual HDD, CDD, GDD, and WI from the CARPATCLIM data, we used Equations (1)(2)(3)(4)(5)(6)(7)(8) directly on the TNTMTX grids. For each variable, we extracted a secular record that represented the average annual degree-days over the entire Carpathian region for the period 1961–2010 (bold lines in Figure 2, where we show the period 1981–2010). The same records were extracted from the degree-day grids resulting from our new dataset and modelling procedure (dashed lines).

image

Figure 2. Records of HDD, CDD, GDD, and WI for the Carpathian region in the period 1981–2010.

Download figure to PowerPoint

Owing to the different spatial resolutions, we used 5895 grid points to compute the degree-day records from the CARPATCLIM data and 925 grid points from our dataset and model. The same digital elevation model has been exploited: the Shuttle Radar Topographic Mission at 90 m (SRTM-90 m-v4.1; Jarvis et al., 2008). In our grids, the average elevation of the Carpathian region is 370 m, whereas it is 383 m in the CARPATCLIM grids. We assumed that a 13-m gap is not relevant to our scopes, as it corresponds to 0.05 °C for a temperature lapse rate of 4 °C km−1 (summer) and to 0.09 °C for a lapse rate of 7 °C km−1 (winter).

Our model, based on monthly TM only, is able to reconstruct the degree-day variables with remarkable effectiveness (Figure 2). The Pearson's correlation coefficient (r; Rodgers and Nicewander, 1988), computed over the entire period 1961–2010, is very high for HDD (r = 0.998), high for WI (0.981) and GDD (0.961), and a bit lower for CDD (0.942).

In general, many institutions (e.g. the Bureau of Meteorology of the Australian Government, 2005) use simpler algorithms to compute the degree-days. We compared our model with two other models applied to the monthly data of the CARPATCLIM dataset. The model named as ‘other-1’ (Figure 2) is based on the simple deviation of TM from the selected baselines, whereas ‘other-2’ is based on Equations (1)(2)(3)(4)(5)(6)(7)(8) but uses also monthly TN and TX. Dealing with HDD, the differences are negligible, but our model does not underestimate GDD and WI, and it is the only one that is able to reconstruct CDD. Consequently, if only monthly temperature data are available, our methodology provides better results with respect to the most common techniques.

4 Results and discussion

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References

4.1 1951–2010 European degree-day climatologies

In Figure 3, we show the mean annual values of the degree-day variables in the period 1951–2010. We excluded 2011 to keep the analysis on exactly six decades. HDD showed the expected latitudinal gradient and a west-to-east smooth increase: in Europe, the lowest values were in southern Spain (500–800), and the highest in Finland (4500). CDD showed a south-to-north decrease: from the Mediterranean region (up to 600 in Europe) to Scandinavia (<5 in the Arctic Circle). GDD and WI showed a clear latitudinal pattern and the highest European values in the Mediterranean and the Black Sea regions (GDD close to 4000 and WI to 3000).

image

Figure 3. Climatologies of HDD, CDD, GDD, and WI for the period 1951–2010.

Download figure to PowerPoint

To analyse the last decade, we show the anomaly grids computed as the mean annual values in the period 2001–2010 minus the mean annual values in the period 1971–2000 (Figure 4). Compared to the period 1971–2000, the period 2001–2010 was characterized by lower HDD and higher CDD, GDD, and WI across almost the entire Europe. In particular, HDD showed the largest decrease in central Sweden, Finland and Russia and a small increase in Latvia and Turkey. Regarding CDD, the increase was more evident in the Mediterranean region and Turkey. Regarding GDD and WI, the western Mediterranean experienced the most evident increase.

image

Figure 4. Anomaly grids of HDD, CDD, GDD, and WI computed as 2001–2010 minus 1971–2000.

Download figure to PowerPoint

In Section 'A simple model that allows degree-days to be obtained using only monthly data', we affirmed that the MAE of our modelling technique is low on station basis for HDD, GDD, and WI, and low but improvable for CDD. We computed the gridded discrepancies between degree-day variables calculated by means of TNTX daily values and by our model applied to monthly TM in the test period 2001–2010. In Figure 5, the positive values mean that our model overestimated the degree-days computed with daily data. Regarding HDD, the differences exceeded 1% only in Croatia, Bulgaria, Slovakia, and Turkey; regarding CDD, the differences were smaller in absolute values, but up to 11% in Spain, the Republic of Macedonia, Serbia, and Turkey. If we skip the grid points with CDD < 5 and the extra-European territories, where the density of input data is low, the overall MAE for CDD reduced from 5.7 to 3.1%.

image

Figure 5. Gridded comparison (for 2001–2011) between HDD (left) and CDD (right) calculated with the daily dataset and with the model based on monthly TM.

Download figure to PowerPoint

It seems that our model, based on TM, slightly overestimates HDD and CDD in coastal regions, in particular on the Mediterranean Sea (excluding Italy) and on the Black Sea. The model failed to account for the mitigating sea effect, which characterizes the first 15–20 km from the coastlines and involves in particular TN in winter and TX in summer. In contrast, our model slightly underestimated HDD and CDD in the inland territories (especially Bosnia-Herzegovina, Bulgaria, and Slovakia), due to a continental climate effect characterized by cold winters and hot summers. Also the Po Valley is biased, probably because the urbanization, combined with the stagnation of air masses and the absence of winds, causes local summer heat waves (very high TX). The next versions of the model will include a special parameterization for the coastal areas.

4.2 European and regional degree-day trends

To investigate the evolution of degree-days from 1951 to 2010, we performed a linear trend analysis. The statistical significance of each trend has been tested with a Student's t-test (Gosset, 1908) and the values are reported in degree-days year−1.

Figure 6 shows the gridded linear trends of HDD, CDD, GDD, and WI. HDD decreased almost everywhere in Europe excluding Iceland, northern Scandinavia, the Balkans, and Turkey, where HDD increased. For CDD, GDD, and WI, the Mediterranean region experienced the largest positive trends in the Provence (France) and Sardinia (Italy); GDD and WI showed significant positive trends in the majority of Europe, though the Baltic Sea region was characterized by negative trends, but the normal values were so small there (GDD ≈ 800, WI ≈ 450) that the crop cultivation was rarely possible.

image

Figure 6. Maps of degree-day trends in 1951–2010 at 95% confidence level. Grey: not significant trends; white: trends close to null values; black: no valid data.

Download figure to PowerPoint

According to geographical and political features, we divided Europe into 14 regions. For each region and variable (Table 3), we calculated a 60-year record computed as the mean value amongst all the grid points located within the corresponding borders.

Table 3. Linear trends of HDD, CDD, GDD, and WI in degree-days year−1 for the period 1951–2010.
No.RegionELEVHDD year−1CDD year−1GDD year−1WI year−1
  1. In brackets: the statistical significance. ELEV is the average elevation of the region.

1Iceland518.5  1.9 (99)0.7 (99)
2UK + Ireland148.8−2.7 (99)0.1 (99)3.8 (99)2.7 (99)
3Iberian Peninsula639.1−2.8 (99)1.7 (99)5.7 (99)4.6 (99)
4France + Belgium + Netherlands245.6−3.9 (99)1.0 (99)6.3 (99)5.3 (99)
5Alps1370.3−3.9 (99)0.9 (99)5.2 (99)4.1 (99)
6Germany + Denmark263.8−4.1 (99)0.5 (99)4.8 (99)3.7 (99)
7Scandinavia331.3−4.1 (95)0.1 (99)2.5 (99)1.6 (99)
8Italy390.8−2.8 (99)1.9 (99)6.2 (99)5.3 (99)
9Balkans536.7−2.4 (95)1.1 (99)4.4 (99)3.5 (99)
10Greece + Cyprus431.5    
11Carpathians368.5−3.6 (98)0.6 (99)3.5 (99)2.7 (99)
12Baltic Republics136.1−5.0 (99)0.3 (99)3.8 (99)2.7 (99)
13Black Sea393.1 1.1 (99)3.8 (99)2.8 (99)
14West Russia + Ex-USSR151.9−5.9 (99)0.1 (99)2.8 (99)1.8 (98)

Almost every region showed a negative trend for HDD and a positive one for CDD, GDD, and WI, in accordance with the temperature rise in Europe, which is the main driving factor for degree-day trends. The negative trend in HDD was largest in Russia and northern central Europe. Regarding CDD, the largest positive trends occurred in Italy and the Iberian Peninsula; regarding GDD and WI, they occurred in the western Mediterranean region, but also in the Alps, and this may open the way for new cultivations of crops at higher elevation. However, if GDD are too high due to hot and maybe dry summers, the crop growth can be affected negatively.

In general, the decrease of HDD was faster in the last three decades than in the entire period analysed, as well as the increase of CDD, GDD, and WI has been characterized by larger values in the period 1981–2010 than 1951–2010. However, a few regions have been characterized by trends with opposite signs in the periods 1951–1980 and 1981–2010 (Table 4). For example, regarding HDD, Iceland, Greece, and Cyprus showed a positive trend in 1951–1980 and a negative one in 1981–2010. In southeastern Europe, CDD decreased in 1951–1980 and increased in 1981–2010. Similarly, GDD and WI decreased in 1951–1980 and increased in 1981–2010 in the Iberian Peninsula, Italy, the Balkans, the Carpathians, Greece, and Cyprus.

Table 4. Linear trends of HDD, CDD, GDD, and WI in degree-days year−1 for the periods 1951–1980 and 1981–2010. Only the trends significant at 95% level are shown.
  HDD year−1HDD year−1CDD year−1CDD year−1GDD year−1GDD year−1WI year−1WI year−1
No.Region51–8081–1051–8081–1051–8081–1051–8081–10
1Iceland6.3−11.3  −2.48.4 2.6
2UK + Ireland −6.7   8.4 5.6
3Iberian Peninsula   2.7−6.19.7−5.68.7
4France + Belgium + Netherlands −6.0 1.4 10.5 8.8
5Alps −5.8 1.3 8.5−2.87.0
6Germany + Denmark   0.8 6.9 5.9
7Scandinavia −11.8 0.3 7.0 4.7
8Italy −7.8−2.03.3−6.311.8−7.110.4
9Balkans −7.3−2.32.8−5.49.1−7.08.1
10Greece + Cyprus2.7−4.2−5.55.3−8.010.3−10.29.1
11Carpathians  −1.71.9−6.28.9−7.27.9
12Baltic Republics   1.0 7.9 6.8
13Black Sea −7.2−1.64.3 12.1 10.6
14West Russia + Ex-USSR   1.2 7.3 5.8

We defined a 15th region, the European Mediterranean coast: from Gibraltar to Antalya (Turkey), delimited by a belt of 50 km from the coastline, excluding the points with an elevation higher than 600 m. For this area, we computed the anomaly series of the degree-day variables using the mean annual value of 1951–2010 as the reference value (Figure 7). HDD showed a significant decrease only in the last 30 years and the anomalies of CDD, GDD, and WI turned from negative to positive values in the 1980s. Regarding the period 1981–2010, the trends were all significant at 99% level: −5.2 HDD year−1, 4.1 CDD year−1, 10.7 GDD year−1, and 9.7 WI year−1.

image

Figure 7. Anomaly series (vs normal 1951–2010 values) of HDD, CDD, GDD, and WI for the European Mediterranean coastal region.

Download figure to PowerPoint

As our methodology does not account for Tu, climatologies and trends of GDD can be slightly biased in the regions involved by relevant summer heat waves. In fact, extremely high temperatures should not be considered as GDD as they are expected to cause a reduction of primary productivity, as it happened during the European heat wave of 2003 (Ciais et al., 2005).

Theoretically, an increase in CDD leads to an increase in energy consumption to cool the internal environments, whereas a decrease in HDD leads to a saving, because the heating systems need to be turned on fewer days. To evaluate if these effects were balanced, we defined the energy degree-days (EDD) as the sum of HDD and CDD. In Figure 8, we show the linear trends of EDD: regarding 1951–2010, a negative trend was found for central Europe, Russia, UK, Ireland, and Scandinavia; regarding 1981–2010, the negative trend was even larger in Iceland, Scandinavia, UK, Ireland, Northern France, Italy, and former Yugoslavia.

image

Figure 8. Maps of EDD trends in 1951–2010 (left) and 1981–2010 (right) at 95% confidence level. Grey: not significant trends; white: trends close to null values; black: no valid data.

Download figure to PowerPoint

HDD and CDD are roughly comparable, because the energy related to CDD and HDD varies if the heating (cooling) systems are based on different technologies and also because buildings are usually heated (cooled) to different temperatures according to the use, the local laws, and personal comfort. However, from a climatological point of view, a rational use of the air conditioning and the heating systems may lead to a save in energy consumption due to climate change (Christenson et al., 2006). In the last three decades, the decrease in HDD was greater than the increase in CDD in Italy and former Yugoslavia. Also, in northern Europe HDD decreased and CDD showed values so small that there was no need to cool the environments in summer.

4.3 The Winkler Index applied to grapevines

Grapes are one of the most important crops in Europe: according to the official dataset of the Statistic Division of the Food and Agriculture Organization of the United Nations (FAOSTAT; website: http://faostat3.fao.org/faostat-gateway/go/to/home/E), in 2010 Europe produced 46.8% of the world supply of grapes, i.e. more than 31.4 billion tons, and the countries on the Mediterranean Sea produced 42.4%. In particular, Italy (2nd), Spain (4th), France (5th), Turkey (6th), Egypt (11th), and Greece (15th) were amongst the greatest producers (Figure 9).

image

Figure 9. Production of grapes in 2010 in % of the global production. Data obtained from the official dataset of FAOSTAT, available at: http://faostat3.fao.org/faostat-gateway/go/to/home/E.

Download figure to PowerPoint

The WI was first proposed and applied to grapevines by Amerine and Winkler (1944), but we used the improved version presented by Winkler et al. (1974), see Equations (7) and (8), and we adapted the classification used in California (whose climate is roughly similar to the Mediterranean climate) to Europe as follows: Region I (1000 < WI ≤ 1500), II (1500 < WI ≤ 2000), III (2000 < WI ≤ 2500), IV (2500 < WI ≤ 3000), and V (WI > 3000).

In Figure 10, we show the 1961–1990 (left) and the 2001–2010 (right) climatologies of WI. In Europe, the most important producers were those countries whose land fell into more than two categories (Spain, France, Italy, Greece, and Turkey) because many different types of grapevines could be cultivated there (Johnson and Robinson, 2007). Moreover, following WI, new areas became suitable for grapevines from 1961–1990 to 2001–2010 in Belgium, the Netherlands, Germany, Slovakia, Poland, and Belarus.

image

Figure 10. Climatologies of WI: 1961–1990 (left) and 2001–2010 (right).

Download figure to PowerPoint

According to this preliminary analysis, WI proved to be a valid indicator to detect the countries suitable for grape growth, in fact the greatest producers in Europe in 2010 (Figure 9) were those countries where WI indicated ‘favourable’ conditions (Figure 10). To improve these results, we plan to use soil information, construct higher resolution grids to capture the local features (south or north-facing hill slopes), and investigate the correlation between WI records and the grape production at country level.

5 Conclusions

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References

We constructed a new European dataset composed of 4023 quality checked 2001–2011 daily TNTMTX records and 3897 homogenized 1951–2011 monthly TM records and we set a methodology based on the estimation of degree-days using monthly temperature data. We found a set of parameters for each station and interpolated them with a KED spatial scheme onto a regular 0.25° × 0.25° European grid. Such parameters, valid for every period, allow the user to calculate HDD, CDD, GDD, and WI if only monthly TM are available, thus a daily TNTMTX dataset is no longer needed. In general, monthly TM can be more easily collected, homogenized, and completed than daily temperature values.

To test such a methodology, we performed a double validation procedure. Firstly, on the station basis and for 2001–2011, we compared the degree-days obtained using measured daily data versus the modelled degree-days: our model estimated HDD, GDD, and WI with MAE < 1% and CDD with MAE ≈ 5.7%. Secondly, we compared HDD, CDD, GDD, and WI obtained with our model versus those calculated with an independent dataset (CARPATCLIM) in the Carpathian region: a high correlation was found, in particular for HDD (r > 0.99) and WI (r > 0.98). Though the validations proved the effectiveness of the presented datasets and model compared with the most commonly used techniques, our methodology still suffers a bias in coastal regions.

Subsequently, we interpolated the degree-day variables onto a regular 0.25° × 0.25° European grid: the last decade showed a generalized decrease of HDD and an increase of CDD, GDD, and WI. A linear trend analysis showed that the temperature rise caused an increase in CDD, GDD, and WI almost everywhere in Europe (with the Mediterranean area being most affected) and a decrease in HDD, in particular in northern and eastern Europe. In general, the trends from 1981 to 2010 showed larger values compared to those from 1951 to 2010 and this can be seen as further evidence that the last decades have been subjected to a climate shift.

In central and northern Europe, the decrease in HDD has outweighed the increase in CDD in the period 1951–2010; therefore, a rational use of heating and cooling systems may lead to a decrease in energy consumption. The same applies to Iceland, UK, Ireland, northern France, northern Italy, Scandinavia, and the Balkans, regarding the period 1981–2010.

With regard to WI, we proved the effectiveness of WI as an indicator for grape production in Europe where, in the last decades, new areas turned into areas climatically suitable for grape cultivation.

Though the methods described in this article can be applied to any baseline, the results presented depend on the selected baselines. As a specific user might want degree-day climatologies based on different baselines, we plan to construct European degree-day grids based on a set of base and upper temperatures, in particular for GDD. In the future, we plan to apply the described methodology to smaller areas with denser datasets: some preliminary tests for Italy are promising.

Acknowledgements

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References

We sincerely thank the four anonymous referees that contributed to improving the paper with their precious suggestions and comments. We also acknowledge Dr Martha Dunbar, a colleague of the authors at the Joint Research Centre and a native English speaker, for the language editing.

References

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Data
  5. 3 Theory and methods
  6. 4 Results and discussion
  7. 5 Conclusions
  8. Acknowledgements
  9. References
  • Allen JC. 1976. A modified sine wave method for calculating degree days. Environ. Entomol. 5(3): 388396.
  • Amerine MA, Winkler AJ. 1944. Composition and quality of musts and wines of California grapes. Hilgardia 15: 493675.
  • Benichou P, Le Breton O. 1987. Prise en compte de la topographie pour la cartographie des champs pluviometriques statistiques. La Meteorologie 19: 2334.
  • Berry J, Bjorkman O. 1980. Photosynthetic response and adaptation to temperature in higher plants. Annu. Rev. Plant Physiol. 31(1): 491543.
  • Bureau of Meteorology of the Australian Government. 2005. Average Annual & Monthly Heating and Cooling Degree Days. www.bom.gov.au/jsp/ncc/ climate_averages /degree-days.
  • Buyukalaca O, Bulut H, Yilmaz T. 2001. Analysis of variable-base heating and cooling degree-days for Turkey. Appl. Energy 69(4): 269283.
  • Christenson M, Manz H, Gyalistras D. 2006. Climate warming impact on degree-days and building energy demand in Switzerland. Energy Convers. Manage. 47(6): 671686.
  • Ciais P, Reichstein M, Viovy N, Granier A, Ogée J, Allard V, Aubinet M, Buchmann N, Bernhofer C, Carrara A, Chevallier F, de Noblet N, Friend AD, Friedlingstein P, Grünwald T, Heinesch B, Keronen P, Knohl A, Krinner G, Loustau D, Manca G, Matteucci G, Miglietta F, Ourcival JM, Papale D, Pilegaard K, Rambal S, Seufert G, Soussana F, Sanz MJ, Schulze ED, Vesala T, Valentini R. 2005. Europe-wide reduction in primary productivity caused by the heat and drought in 2003. Nature 437(7058): 529533.
  • Efron B, Gong G. 1983. A leisurely look at the Bootstrap, the Jackknife, and cross validation. Am. Stat. 37(1): 3648.
  • Gosset WS. 1908. The probable error of a mean. Biometrika 6(1): 125.
  • Grigorieva EA, Matzarakis A, de Freitas CR. 2010. Analysis of growing degree-days as climate impact indicator in a region with extreme annual air temperature amplitude. Clim. Res. 42: 143154.
  • Hassan QK, Bourque CP, Meng F, Richards W. 2007. Spatial mapping of growing degree days: an application of MODIS-based surface temperatures and enhanced vegetation index. J. Appl. Remote Sens. 1: 013511(24 April 2007).
  • Haylock MR, Hofstra N, Klein Tank AMG, Klok EJ, Jones PD, New M. 2008. A European daily high-resolution gridded dataset of surface temperature and precipitation. J. Geophys. Res. 113: D20119.
  • Hengl T. 2009. A Practical Guide to Geostatistical Mapping. Office for Official Publications of the European Communities: Luxembourg, Germany. ISBN: 978-90-9024981-0.
  • Huglin P. 1978. Nouveau mode d'évaluation des possibilities héliothermiques d'un milieu viticole. C. R. Acad. Agric. 64: 11171126.
  • Jarvis A, Reuter HI, Nelson A, Guevara E. 2008. Hole-filled SRTM for the globe Version 4. Available from the CGIAR-CSI SRTM 90m Database. http://srtm.csi.cgiar.org.
  • Jato MV, Frenguelli G, Rodriguez FJ, Aira MJ. 2000. Temperature requirements of Alnus pollen in Spain and Italy (1994–1998). Grana 39(5): 240245.
  • Johnson H, Robinson J. 2007. The World Atlas of Wine (6th edition). Mitchell Beazley: London. ISBN: 978-1-84533-301-0(PCLC number: 213401236).
  • Jones GV, Davis RE. 2000. Climate influences on grapevine phenology, grape composition, and wine production and quality for Bordeaux, France. Am. J. Enol. Vitic 51(3): 249261.
  • Klein Tank AMG, Wijngaard JB, Konnen P, Bohm R, Demaree G, Gocheva A, Mileta M, Pashiaridis S, Hejkrlik L, Kern-Hansen C, Heino R, Bessemoulin P, Muller-Westermeier G, Tzanakou M, Szalai S, Palsdottir T, Fitzgerald D, Rubin S, Capaldo M, Maugeri M, Leitass A, Bukantis A, Aberfeld R, van Engelen AFV, Forland E, Mietus M, Coelho F, Mares C, Razuvaev V, Nieplova E, Cegnar T, Antonio Lopez J, Dahlstrom B, Moberg A, Kirchhofer W, Ceylan A, Pachaliuk O, Alexander LV, Petrovic P. 2002. Daily dataset of 20th century surface air temperature and precipitation series for the European Climate Assessment. Int. J. Clim. 22(12): 14411453.
  • Matzarakis A, Balafoutis C. 2004. Heating degree-days over Greece as an index of energy consumption. Int. J. Clim. 24(14): 18171828.
  • McMaster GS, Smika DE. 1988. Estimation and evaluation of winter wheat phenology in the central Great Plains. Agric. For. Meteorol. 43(1): 118.
  • McMaster GS, Wilhelm W. 1997. Growing degree-days: one equation, two interpretations. Agric. For. Meteorol. 87(4): 291300.
  • Menne MJ, Durre I, Vose RS, Gleason BE, Houston TG. 2012. An overview of the Global Historical Climatology Network-Daily Database. J. Atmos. Oceanic Tech. 29(7): 897910.
  • Miller P, Lanier W, Brandt S, 2001. Using growing degree days to predict plant stages. Ag/Ext Comm Coord, Comm Services, Montana State University-Bozeman, Bozeman, MO. http://msuextension.org/publications/AgandNaturalResources/MT200103AG.
  • Moriondo M, Bindi M. 2007. Impact of climate change on the phenology of typical Mediterranean crops. Ital. J. Agron. 3: 512.
  • Papakostas K, Mavromatis T, Kyriakis N. 2010. Impact of the ambient temperature rise on the energy consumption for heating and cooling in residential buildings of Greece. Renew. Energy 35(7): 13761379.
  • Parry ML, Canziani OF, Palutikof JP, van der Linden PJ, Hanson CE. 2007. Climate change 2007: impacts, adaptation and vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press: Cambridge, UK, 982.
  • Project team ECA&D. 2013. Royal Netherlands Meteorological Institute (KNMI) - European Climate Assessment & Dataset (ECA&D) – Algorithm Theoretical Basis Document (ATBD) Version 10.7, Retrieved September 16, 2013. http://eca.knmi.nl/documents/atbd.pdf.
  • Pruess KP. 1983. Day-degree methods for pest management. Environ. Entomol. 12(3): 613619.
  • Rodgers JL, Nicewander WA. 1988. Thirteen ways to look at the correlation coefficient. Am. Stat. 42(1): 5966.
  • Roltsch WJ, Zalom FG, Strawn AJ, Strand JF, Pitcairn MJ. 1999. Evaluation of several degree-day estimation methods in California climates. Int. J. Biometeorol. 42: 169176.
  • Russelle MP, Wilhelm W, Olson RA, Power JF. 1984. Growth analysis based on degree days. Crop Sci. 24: 2832.
  • Santos JA, Malheiro AC, Pinto JG, Jones GV. 2012. Macroclimate and viticultural zoning in Europe: observed trends and atmospheric forcing. Clim. Res. 51: 89103.
  • Szalai S, Vogt J. 2011. CARPATCLIM – high resolution gridded database of the Carpathian Region and calculation of drought indices as a contribution to the European Drought Observatory. In WRCP Conference, 24–28 October 2011. Denver, CO, T185A.
  • Szentimrey T. 1999. Multiple analysis of series for homogenization (MASH). In Proceedings of the 2nd Seminar for Homogenization of Surface Climatological Data. Budapest, Hungary, WMO, WCDMP-No. 41, 27–46.
  • Szentimrey T, Bihari Z. 2007. Mathematical background of the spatial interpolation methods and the software MISH (Meteorological Interpolation based on Surface Homogenized Data Basis). In Proceedings from the Conference on Spatial Interpolation in Climatology and Meteorology. Budapest, Hungary, COST Action 719, COST Office 17–27.
  • Thom HCS. 1954. The rational relationship between heating degree days and temperature. Mon. Weather Rev. 82: 16.
  • UK Met-Office, 2013. http://www.metoffice.gov.uk/
  • Valor E, Meneu V, Caselles V. 2001. Daily air temperature and electricity load in Spain. J. Appl. Meteorol. 40: 14131421.
  • Wibig J. 2003. Heating degree days and cooling degree days variability in Lodz in the period 1931–2000. In Fifth International Conference on Urban Climate. Lodz, Poland.
  • Winkler AJ, Cook JA, Kliewer WM, Lider LA. 1974. General Viticulture: Second Revised Edition. University of California Press: Berkeley, CA. ISBN: 9780520025912, 0520025911.
  • Yang S, Logan J, Coffey DL. 1995. Mathematical formulae for calculating the base temperature for growing degree days. Agric. For. Meteorol. 74(1): 6174.