A drought severity climatology for the Carpathian region has been produced
using the self-calibrating Palmer Drought Severity Index (Sc-PDSI) for the
period 1961–2010. Using the Sc-PDSI and the assumptions of the Palmer
drought model (PDM) the precipitation required for drought termination (when
Sc-PDSI reaches
Drought is one of the most far-reaching natural and socioeconomic disasters
(WMO/UNCCD/FAO/UNW-DPC, 2013). Traditionally, the acknowledgement and
attempts to manage droughts were mostly orientated towards crisis
management, while little attention has been given to proactive drought risk
management. More recently, European as well as international policies and
initiatives have highlighted the need for a more proactive, risk-based
management of droughts. Examples are the requirement for the set-up of River
Basin Management Plans, including Drought Management Plans under the
European Water Framework Directive (WFD), the High-Level Meeting on National
Drought Policies (HMNDP,
An essential element in risk management is the reduction of drought impacts (i.e., mitigation) based on an assessment of the cost of damages associated with droughts as compared to the costs for efficient early warning and preparedness, including the adaptation to climate change. Drought as a natural hazard has been the subject of a great number of studies, focusing on the definition of drought and the development of drought indicators (e.g., Palmer, 1965; McKee et al., 1993; Wells et al., 2004; Vicente-Serrano et al., 2010) as well as on drought assessment and monitoring (e.g., Briffa et al., 1994; Guttman, 1998; Lloyd-Hughes and Saunders 2002; Dai et al., 2004; van der Schrier et al. 2006; Dai, 2011). However, little attention was given to the analysis of probabilities that a given drought (and its impacts) could be ameliorated or terminated through adequate rainfalls. The number of studies addressing the drought recovery topic is few (Karl et al., 1986, 1987) and articles focused on drought as a natural hazard (Wilhite, 2000; Wilhite and Buchanan, 2005) as well as reports on drought management and monitoring (e.g., WMO, 2006; IPCC, 2007; ISDR, 2007), address the subject only in a general manner.
This paper provides a quantitative measure of the probability that any
drought could be ameliorated or terminated over some defined period of time
using the Palmer drought model (PDM) (Karl et al., 1987) with its
assumptions and limitations. The study was partially implemented in the
framework of the CARPATCLIM project (
The Sc-PDSI is a drought indicator based on the principles of balance between moisture supply and demand. A series of articles have pointed out the assumptions, strengths and weaknesses of the Palmer drought model along with details on calculation procedures (Alley, 1984; Karl, 1987, 1986a, b; Wells et al., 2004; van der Schrier et al., 2006 ). The PDSI or modified versions of PDSI have been used to quantify drought as a recurrent extreme climate event both at continental (Europe, North America) and global level (Dai et al., 2004; Dai, 1998, 2011; Wells et al., 2004; van der Schrier et al., 2006, 2007). By changing the standardization used by Palmer (1965), which was based on data from the US, Wells et al. (2004) proposed the Sc-PDSI and it was recognized as an improvement of the original PDSI (Dai, 2011). PDSI was developed with the intention of measuring the departure of soil moisture from the normal conditions, using a hydrological accounting system. Different from PDSI other drought indicators are based on past statistics of certain climate variables which often include only precipitation (Dai, 2011) and assumes that droughts are directly controlled by the temporal variability of the precipitation. Recent studies have confirmed the importance of the effect of other variables, such as temperature, on drought conditions. These studies (Williams et al., 2011; Martínez-Villalta et al., 2008; McGuire et al., 2010; Linares and Camarero, 2011) have shown that temperature rise affects the severity of the droughts and mainly the drought stress induced by heat waves on net primary production and tree mortality. For examples, the heat waves in Europe in 2003 and 2010 played extreme roles in influencing drought severity, which increased evapotranspiration and aggravated the drought severity (Rebetez et al., 2006). As a result – enhanced by precipitation shortage (Adams et al., 2009) – major decreasing in net primary production (Ciais et al., 2005) and high forest mortality occurred. This shows how drought stress – through increased evapotranspiration – is induced, to a large degree, by the availability of soil moisture. Therefore, the use of drought indices which is based on soil water balance model, such as PDSI or modified versions as Sc-PDSI, is required in order to calculate current soil moisture conditions. Moreover, the statistical based drought indicators are normalized measures with respect to location and period, which makes the frequency of their severity classes climatologically consistent for any site (Heinrich and Gobiet, 2012), not being able to identify regions that are more “drought-prone” than others (Hayes et al., 1999). Therefore, PDSI has been used as it allows for comparison of drought frequency within different severity classes on different locations and it is suitable to account the drought under global warming conditions. Various aspects of the hydrological model, on which the PDSI is based on, are directly used in the calculation procedure of the precipitation required to recover from drought, which not only confers homogeneity but also offers means of validation of the obtained results.
Carpathian region – geographical units.
Based on these considerations and using the assumptions of the Palmer
drought model (PDM), the precipitation needed to end or ameliorate a drought
(in a 1-, 3-, or 6-month period) for different levels of severity (moderate when
Sc-PDSI
The region covered by this study, depicted in Fig. 1, is centered on the
Carpathian Mountains and the surrounding lowlands (17–27
Gridded monthly precipitation (from the CARPATCLIM project at 0.1 Gridded monthly mean surface air temperature (from the CARPATCLIM project
0.1
Temperature and precipitation gridded data have been interpolated within the
CARPATCLIM project from quality-checked, completed, homogenized and
harmonized daily station data. Please see Spinoni et al. (2013) for a more
detailed description.
The available water holding capacity (AWC) of the soil is computed from the soil
texture classes and soil profile depths in the European Soil Database (
Available water holding capacity (AWC) of the soil (mm) in the Carpathian region.
For the computation of the precipitation needed to recover from drought gridded data sets of CAFEC (climatically appropriate for existing conditions) precipitation (precipitation needed to maintain a normal soil moisture level), climate characteristic coefficient and the moisture anomaly index from the Palmer drought model for 1961–2010 period were used. A complete description of these variables is presented in Sect. 2.2.2
Sc-PDSI is based on the Palmer Drought Severity Index (PDSI), first introduced by Palmer (1965), and modified by Wells et al. (2004) in order to allow a more accurate comparison of the index at different locations. Sc-PDSI measures the cumulative departure of moisture supply and demand from the normal conditions and is computed on monthly time scale. The supply in this model is the precipitation, the water demand is the potential evapotranspiration and the outputs are the actual evapotranspiration and runoff. Often discussed in other studies (e.g., Alley 1984; Karl, 1986a; Guttman et al., 1992; Weber and Nkemdirim, 1998; Wells et al., 2004; Dai et al., 2004; Vicente-Serrano et al., 2010) the strengths, weakness and differences of these two drought indicators will not be examined in this study. The major difference comes from the reduced frequency of extreme events of Sc-PDSI when compared with PDSI. This aspect is the result of the overall effect of the calibration based on the actual climatic characteristics of a given location that makes Sc-PDSI more consistent over diverse climatological regions.
The Sc-PDSI calculation procedure starts with the calculation of the monthly hydrological parameters of a rather complex soil water balance system: evapotranspiration, recharge, runoff, water loss from the soil and their potential values. The hydrological system is confined by the assumptions that the soil is split in two layers (with the upper soil layer holding 25.4 mm of water) and the saturation level (of both soil layers) is conditioned by the top layer, both on supply and demand.
The potential evapotranspiration (PET) is estimated following Thornthwaite (1948), while the other potential parameters are defined as follows: the potential recharge (PR) is the amount of moisture required to bring the soil moisture up to filed capacity (AWC less the total amount of moisture stored in both soil layers), the potential loss (PL) is the moisture that could be lost from the soil if precipitation is zero for the month and the potential runoff (PRO) is defined as total AWC less potential recharge (PR). The climatically appropriate for existing conditions (CAFEC) precipitation (or the precipitation needed to maintain a normal soil moisture level) is obtained by summing the monthly mean potential values which are previously scaled by their ratio with the monthly mean actual values.
The difference between monthly precipitation and CAFEC-precipitation, weighted by a local climate characteristic coefficient (an empirical derived normalization factor) results in Palmer moisture anomaly index (Palmer's Z-Index). A description of the modifications made to obtain Sc-PDSI is presented in Appendix A.
The Sc-PDSI
However, before being able to compute
This represents the transformation of the moisture anomaly
index (
From the PDSI severity classes (Palmer, 1965), adopted also for the Sc-PDSI
(Table 1), it can be stated that a drought event ends when the Sc-PDSI
increases above
Considering the same severity classes, it can be assumed that a drought is
ameliorated when the Sc-PDSI reaches a value of
The weighted factors
The next step is assigning values to the CAFEC precipitation (
Cumulative frequency, severity classes, and SC-PDSI values in the Carpathian region.
In the third step, the precipitation needed to end or ameliorate the drought is
computed as in Eq. (1), using both the moisture anomaly index needed to end
(
The climatological probability of receiving the amount of precipitation needed to end and ameliorate the drought was calculated using the gamma distribution. Gamma distribution has been frequently used in literature to represent precipitation (Thom, 1966; Wilks, 1990, 1995; Oeztuerk, 1981) due to the advantage that it excludes negative values, being bounded on the left at zero (Thom, 1966; Wilks, 1995). Analysis of rainfall data strongly depends on its distribution pattern (Sharma et al., 2010). This is especially important as gamma distribution is positively skewed and represents an advantage as it mimics the actual rainfall distributions for many geographical areas (Ananthakrishnan et al., 1989). Also it provides a flexible representation of a variety of rainfall regimes while utilizing only two parameters, the shape and the scale (Wilks, 1990).
Average number of months per year with moderate drought (Sc-PDSI
The calculations were performed separately for each month and each location
(grid point) on the basis of the entire 50 years of available data
(1961–2010). Input data were the computed precipitation needed to end or
ameliorate the drought (0.1
All the procedures followed in the calculation of the climatological probability of recovering from a drought are based on the processes used by Oeztuerk (1981) to compute the probability distribution for precipitation. In the first step the actual precipitation data on “moving windows” of 1, 3, and 6 months are matched with the precipitation needed to recover from the drought in the next 1, 3, and 6 months. In the second step the cumulative probability (CDF) of the computed precipitation needed to end or ameliorate the drought is derived.
Previous studies of drought in the Carpathian region were based on the
analysis of intensity, duration and spatial extent, either at national level
(e.g., Palfai, 1990; Szinell et al., 1998; Szalai, 2000; Popova et al.,
2006; Trnka et al., 2009; Cheval, 2013) or at inter-regional level (e.g.,
Bartholy et al., 2013; Spinoni et al., 2013). Our results show that the
incidence of drought in this region is rather high. During the period
1961–2010, every part of the region experienced on average between 0.5 and 6
drought months per year, (Sc-PDSI
The spatial and temporal analysis of the results for precipitations needed
to recover from drought and their climatological probability is related to
three levels of severity: moderate drought when
PDSI originally was designed to measure the soil moisture departure as a difference between a climatological moisture supply which in our case is the actual precipitation and the precipitation needed to maintain a normal soil moisture level (CAFEC precipitation; Palmer, 1965). In this study, other means of moisture supply such as precipitation in form of snow water equivalent are not considered. Since the regional spatial variation of precipitation in this region is mainly determined by the mountain orography and the large scale atmospheric processes (UNEP/DEWA, 2007), it is expected (in a temperate-continental climate) that moisture supply is more significant in the high altitudes while the moisture demand is higher in the low altitudes (higher rate of evapotranspiration due to higher temperatures). With increasing continental conditions from west to east and temperature decreasing from north to south, a higher moisture demand in the south and southwest, and higher moisture supplies in the northern, western and southwestern parts of the region are expected.
Based on these general climatological characteristics, the PDM will produce
the highest
Incidence (% surface of the region) of different severity levels of drought per month.
For both the required precipitation and the probabilities of recovery from drought a spatial pattern linked with the atmospheric circulation patterns responsible for the climate variability in the Carpathian region can be noticed. The southern and southwestern Carpathians and the western Carpathians act like a barrier for the main sources of moisture (Mediterranean and North Atlantic air masses; Busuioc and von Storch, 1996; Busuioc, 2001). These systems are causing first, high precipitation amounts and a pronounced annual precipitation cycle, as it is the case of North Atlantic circulation in the western, northern and northwestern part of the Carpathian region. Secondly, is causing a highly variable precipitation intensity and a relatively constant distributed precipitation regime through the year (by creating a second precipitation peak in autumn), as it is the case of Mediterranean cyclones in the southwestern and southern part of the Carpathian region. The cyclonic presence and trajectories have been the subject of extensive climatological research (e.g., van Bebber, 1891; Radinovic, 1987; Katsoulis, 1998; Flocas, 1998; Maheras, 2001). Often these studies establish a connection between the advance of the cyclones from the Mediterranean area and intense precipitation events. High amounts of precipitation with genesis in the Mediterranean space (Gulf of Genoa) are produced on the cyclonal trajectory V (from the Tyrrhenian Sea to Ukraine). Most important for the Carpathian region are the trajectory Vc that crosses from west to east (the south of the Carpathian region) in spring and very rarely in summer and trajectory Vb (important for the western part of the Carpathian region, passing over the Pannonian Plain) towards Poland. For both trajectories, the cyclones circulate especially in autumn, winter and spring with the largest probability of occurrence in April and a secondary maximum in early autumn. The cyclone circulation diminishes and migrates southwards in December-January, due to the intensification of both the Azores and Siberian anticyclones (Maheras, 2001).
The annual cycles of the moisture supply and demand follow a continental pattern in the Carpathian region (imposed by the North Atlantic circulation) with a maximum of supply and demand at the beginning of the summer (May-June-July) and end of summer (July-August) respectively a minimum in the winter months (December-January-February). Regardless of the annual cycles of precipitation, the months with the higher probability of substantial excess of precipitation from the normal (April-May in spring and October-November in autumn) will be related with the cyclonic presence from the Mediterranean area. Nevertheless the joined influence of the circulations moving either from the Atlantic or the Mediterranean Sea is a common characteristic of the Carpathian region (Busuioc and von Storch, 1996; Busuioc, 2001).
As shown in Fig. 4 the incidence of drought events (Sc-PDSI
One of the characteristics of these drought events is the strong prevalence
(% from the area) of extreme droughts (Sc-PDSI
Percentage above the normal precipitation needed to end a drought – Pp (%) – in the next 1, 3, and 6 months for the drought events with the highest incidence (% surface from the region)
Probability (%) of ending (left) or ameliorating (right) moderate (top), severe (centre) and extreme drought (bottom) events with the highest incidence in the Carpathian region in the next 1, 3, 6 months.
As shown in Table 2 for selected drought events between 200 % to more than
480 % of the normal 1-monthly precipitation would have been required for
recovery (i.e., bringing Sc-PDSI to a level of
In order to get a better idea of the climatological probabilities to recover
from such droughts, we analyzed the first 25 most significant events
(droughts occurring on > 75 % of the area) for different
drought intensity levels. Figure 5 shows the required precipitation in percent of the climatologically expected precipitation and the associated
probabilities for different drought intensities and precipitation
accumulation periods. It can be seen that moderate severity droughts
(
Percentage (%) above the normal precipitation needed to end
a (top) moderate (
Most of these values indicate the improbability of ending or ameliorating the drought in a short period of time, as their climatological probability is (extremely) low. If we settle a limit of 50 % probability, above which the amount of precipitation could be considered more likely than not (IPCC, 2007), none of the drought events could have been ended in the next month. However, a few of the moderate droughts (July 1990, December 1986, December 2000, January 1991) and of the severe droughts (July 1990, December 2000, January 1991) could have been ameliorated with 48 to 140 % above the normal precipitation in 1 month (top right). On the other hand, in a 6-month period almost all drought events considered could most probably have been ended with amounts 10 to 80 % above the normal precipitation (188 % for the extreme drought of April 1991). Only the severe drought from January and February 1964 and the extreme drought from July 2007 could not have been ended even in a 6-month period, making them the most excessive droughts of the studied period in the Carpathian region. Nevertheless, they could have been ameliorated with 45 to 65 % above the normal 6-monthly precipitation. In a 3-month period, only one drought event of extreme intensity (July 1990, requiring 136 % above the normal precipitation), 12 events of the moderate and 6 events of severe drought could have been ended with high probability. All the other events could only have been ameliorated with a range of 15 to 140 % (August 1992) above the normal 3-monthly precipitation.
Climatological probability (%) of receiving the precipitation
needed to end a (top) moderate (
Figure 6 presents the averaged positive deviations (in percent) from the normal precipitation needed to recover from a drought computed for the period 1961–2010. The Pannonian Basin, the Transylvanian Plateau and the external Carpathian foothills and plains in the southern and eastern part of the region require the highest relative amount of precipitation to recover from drought. In these regions, moderate droughts and extreme droughts needed between 250 and 300 % (sometimes up to 600 %) above the normal precipitation to end a drought in the next month, a decrease which is noticed with increasing altitude. The topographic pattern is lost when the moisture supply is required for a larger time window. This is due to the general climate characteristics that overwrite the variability introduced by the local physical conditions. Also, the longer time intervals require less relative amounts of precipitation to recover from drought (i.e., from 20 up to 40–60 % for all the drought intensities).
Figure 7 shows the corresponding probabilities. The probability of ending or
ameliorating an extreme drought (Sc-PDSI
As presented at the beginning of this section, the succession, intensity and the predominance of the cyclonic circulation may lead to a seasonal variability of the precipitation needed to recover from drought and climatological probability. The soil moisture supply and demand follow the annual cycle of precipitation and temperature (imposed by the general atmospheric circulation) but they are reflected differently at the level of the month with the highest and lowest probabilities of recovering from drought. More likely to recover from drought are the months with higher probability of substantial excess of precipitation from the normal and especially for the regions with a constant precipitation regime throughout the year. The more the precipitation regime presents a pronounced peak the more the preferred recovery moths are variable.
The months with the highest (top), and lowest (bottom) probability
of having (left) a moderate drought (
Corresponding to these characteristics, in almost the entire Carpathian region, the preferred months for ending a drought event are the months of April and May as in Fig. 8, corresponding to the months with the largest probability of receiving high precipitation amounts compared with the normal and a maximum activity of Mediterranean cyclones. The least preferred months for ending a drought event are the months of January and February for southern, southwestern, and northwestern regions, and October, November, and December for the northeastern part. This corresponds to the months with the largest probability of receiving low precipitation amounts compared with the normal and a minimum activity of the Mediterranean cyclones. This situation can be observed in Fig. 8 where we present the months with the highest and lowest probability for ending droughts at different intensities during the next month in Fig. 8a, the next 3 months in Fig. 8b and the next 6 months in Fig. 8c.
Moderate drought events in April appear to have the highest probability for being ended in the next month. Also, severe and extreme drought events in April and May (for northern and northeastern regions) are characterized by highest probabilities of being ended in following month. The late summer (July-August) and early autumn (September-October) drought events are ended with highest probability in the southern, western and northwestern parts of the Carpathian region as seen in the Fig. 8a, top.
In Fig. 8b top, we show that the drought events with the highest probability of being ended in 3 months are the droughts from the end of winter (January-February) in the western, southern and northwestern regions for the moderate droughts and spring droughts (April-May) in northern and northeastern regions, especially for the extreme drought events. The late autumn drought events (October-November) present the highest probability of being ended in the next 6-month period as seen in Fig. 8c, top.
Concerning the lowest probabilities for ending a drought event, the worst months for ending the droughts are the winter months, corresponding to the driest period of the annual precipitation cycle and the minimum activity of the Mediterranean cyclones in the Carpathian region. This makes drought events between October (in the northern, northeastern area of the Carpathian region) and February (in the southern and eastern part of the region and Pannonian Basin) the least probable to be ended in the next month as seen in the Fig. 8a, bottom.
In Fig. 8b bottom, we show that the drought events with the lowest probability of being ended in the next 3-month period are the droughts from the December in the northern and northeastern regions and autumn droughts in the other regions.
The least probable to be ended in the next 6 months are the summer droughts (June-July-August), especially in the southern and southwestern regions, while the winter droughts are the least probable to be ended in the north and northeastern part of the region as seen in Fig. 8c.
As shown, in Carpathian region, the water deficit occurs throughout the whole year. As the agriculture is an important economic sector in the Carpathian region, the drought impact could be essential. Most crops may experience water stress (deficit) at various stages in their growth cycle. The sequences of vegetative growth with their key physiological phases (i.e., crop phenology) and their sensitivity to water deficit can be used to highlight the importance of seasonal analysis of drought occurrence. Winter crops (i.e., winter wheat) are planted in Carpathian region in September through October and harvested July through August of the next year, while the spring crops (i.e., corn, spring wheat, sunflower, and potatoes) are planted April through May and harvested August through September or even October (potatoes) of the same year (UNEP/DEWA, 2007).
Early drought in the growing season (the end of autumn in October and November for winter crops and the end of spring in late April and May for spring crops) are affecting wheat germination and crop establishment (Bouaziz and Hicks, 1990). The water stress during the vegetative stages (the months of April and May for winter crops and late May and June for spring crops) may affect the leaf index development (Rickman et al., 1983). Soil water deficit increased towards harvesting (early summer for winter crops and late July or beginning of autumn in August for spring crops) is likely to produce a severe reduction in grain growth and quality which eventually cause reduction in final yields. On the other hand, it has been noted that water deficit in the maturity (anthesis) and harvesting period accelerates development (Simane et al., 1993) and significantly contribute to grain yield (Palta et al., 1994).
The main characteristics of the spatial and temporal variability of the precipitation needed to end or ameliorate a drought in the Carpathian region are presented in this study. Sc-PDSI was used as a drought indicator, and the Palmer drought model was considered to calculate moisture supply and demand. The incidences of drought in the region is considerable. During the study period (1961–2010), the region experienced, on average, drought events from at least 0.5 months to 4–6 months per year for moderate droughts and less than a month per year for extreme droughts.
The amount of the precipitation needed to end a drought in the next month, reached, on average, between 200 and 480 % above the normal 1-monthly and up to 50 % above the 6-monthly total of the normal precipitation. It was also shown that most of the drought events, irrespective of their intensity, are extremely unlikely (< 5 %) to be ended in the next month.
Regionally, the Pannonian Basin, the Transylvanian Plateau and the external Carpathian foothills and plains in the southern and eastern part of the region require the highest quantity of precipitation to recover from drought and present the lowest climatological probabilities. In almost the entire Carpathian region, the preferred months for ending a drought event are the months of April and May corresponding to months with the largest probability of receiving high precipitation amounts. Often during this period of the year a connection between the cyclones from the Mediterranean area and intense precipitation events is established. The worst months for ending the droughts in the Carpathian region are the late autumn and winter months, corresponding to the driest period of the annual precipitation cycle and the minimum activity of the Mediterranean cyclones in the Carpathian region. High precipitation amounts over the northern and northwestern part of the region are causing higher moisture supply and higher climatological probabilities to recover from drought. On the other hand for the eastern and northeastern area of the Carpathian region in October and for the southern and eastern part of the region and Pannonian Basin in February, the smallest climatological probabilities occur to recover from drought.
The early drought events for winter crops (in October and November) cause a high stress effect especially on germination and early crop establishment. On the other hand the water stress is much lower for the spring crops from April and May and for the same physiological phase due to a high cyclonic activity.
The summer droughts have a low probability to be ended, especially in the southern and southwestern regions, if in the maturity (anthesis) and harvesting periods of the crops, the water deficit can cause less damage in crop development, thereby producing even an increase in the grain yield.
The computation of the self-calibrating PDSI was done in four steps: (a) computation of the soil water budget (Thornthwaite, 1948), (b) normalization with respect to demand, (c) normalization with respect to location, and (d) computation of the drought severity.
The computation of the soil water budget was done considering the following assumptions: the soil is divided in two layers, the AWC value is site dependent – representative of the soils type, the top layer contains 25.4 mm of available moisture at field capacity, the moisture stored in the soil layers changes according to the priority conditions imposed by the top layer on supply and demand. Rainfall surplus is first added to the top layer until this layer is saturated and only then it passes to the second layer while on the other hand moisture is withdrawn from the top layer first, before removing from the second soil layer.
Following these rules eight hydrological parameters of the water balance are
computed: the actual evapotranspiration (ET), the soil water recharge (
Normalization with respect to demand (or moisture departure for the month –
Accumulated z-index (mm) versus duration (months) with the intercept of the most extreme drought/wet spell.
Duration factors
This was done by converting the moisture departure for the
month (
Using the climate characteristic coefficient (
Once
Differently from the original computation (the original PDSI is computed
using the duration factors
The most extreme wet/dry period is defined, in this study, as events with
duration greater or equal to 3 consecutive months and with the highest
intensity of
This study was supported, in part, by the results of the CARPATCLIM project
(