Weather and climate events, driven by interacting physical processes across multiple spatial and temporal scales and leading to hazards were defined by [1
] as compound events. Hazards like floods, wildfires, heat waves, and droughts are examples of compound events with multiple drivers, such as run-off, groundwater content, temperature and precipitation [1
]. According to [2
], there is a high probability of increasing extreme precipitation and coastal/river flood risk in Europe and, without adaptive measures, this would substantially increase flood damages. Floods are a threat not only to the safety of the population but also to national security. A MunichRE survey [3
] of the number of meteorological, hydrological and climatological events that caused damage in Europe between 1980 and 2015 reported that the highest damage was caused by weather events such as convective, local, tropical and subtropical storms. According to this study, the number of extreme meteorological phenomena in Europe has increased in the last 35 years. Hydrological phenomena, such as floods and landslides, are the second most damaging natural disaster group.
Southeastern Europe has high sensitivity to extremes in the hydrology cycle, and specifically dry and wet spells. A recent example was the 2011–2012 winter, which was one of the coldest winters in decades. It was followed by an extremely hot and dry summer in 2012 with the worst drought in 40 years and wildfires. In 2014 precipitation above the norm was registered in Southeastern Europe, resulting in wide-spread floods. An extremely wet spring season was observed in the western and central Balkan peninsula. Flood casualties in May 2014 were reported in the Czech Republic, Romania, Slovakia, Croatia, Serbia and Bulgaria as well as economic loses estimated at over 1.1 billion Euro in Serbia alone. Significant precipitation between 13 and 17 May 2014, due to a stationary cyclone, lead to increased river levels, resulting in catastrophic floods. For Romania, this was the third catastrophic flood reported since late April 2014 with 125 villages and nearly 10,000 inhabitants affected. In Croatia, along the rivers Sava, Bosna and Una, tens of thousands of people were evacuated due to the floods [4
]. Floods triggered about 2000 landslides on the Balkan peninsula, some of which inflicted damage to large cities and villages. As a result of muddy landslides, several rivers merged with the catchments of the Sava and Morava rivers. The water volumes of several rivers increased, leading to swampy waters inundating surrounding valleys and causing high economic losses and 49 casualties. The disaster began on 13 May 2014 when the cyclone Ivet formed and then moved from the Mediterranean to Southeastern Europe. Storing enough moisture and warm air from the Adriatic Sea, the cyclone reached the Balkan peninsula, where the air mass experienced strong orographic uplift when passing the Dinaric mountains. That led to extremely large precipitation amounts for the next 4 days. The most significant precipitation in Serbia was recorded from 14 May to 16 May in Belgrade, where for 72 h the precipitation amount reached almost 190 mm. Daily precipitation in 15 May exceeded the historic records of 107.9 mm in Belgrade, 108.2 mm in Veljevo and 110 mm in Loznitsa. On the same day in Belgrade, the 1897 monthly record of 175 mm was exceeded, and a new record of 278 mm was set [5
]. The authors of [7
] studied the climatic and meteorological factors that influence the catastrophic floods in the Balkans, focusing on large-scale circulation. They reported that the cyclone Ivet was unusually stationary, bringing extreme precipitation for several consecutive days, and linked this to a quasi-stationary Rossby wave train. In Bulgaria, according to the National Statistical Institute, in 2014 360 floods were registered, with considerable economic losses in the cities of Varna, Dobrich, Mizia, Burgas, Primorsko and Haskovo [8
]. The authors of [9
] analyzed the meteorologic and hydrologic conditions for 12 major flood events in 2014 in Bulgaria. They reported that four events (24–25 April, 30–31 July, 1–2 August and 23–25 October) were associated with Mediterranean cyclones approaching from the west, southwest or south. One event on 4–6 June 2014 was related to an Atlantic ocean air mass. Three events on 19 April, 15 June and 19–20 June were related to continental air masses from the north, northeast or northwest. The events on 27 January, 28–31 May, 2–7 September and 4–6 December were a result of the combined influence of Mediterranean, Atlantic, continental and polar air masses. As shown by Stoycheva et al. [9
], in 2014 floods in Bulgaria started as early as January and continued until December. For example, the annual precipitation in the capital of Bulgaria, Sofia, was 168% higher than the mean of the previous 20 years.
Climate oscillation indices have a direct relation to flood occurrence through their influence on precipitation patterns. It is well known that the North Atlantic oscillation (NAO) plays an important role in the inter-annual variability of the European climate [10
]. For example, the authors of [11
] investigated the effect of the NAO over the Iberian Peninsula and showed that the winter NAO index was negatively correlated with rainy spells corresponding to large amounts of precipitation, in both frequency and intensity, over the west, southwest, and interior of the Iberian Peninsula. It was also negatively correlated with the winter precipitation maximum, and positively correlated with the winter maximum dry spells. This is consistent with the findings of [12
], that Southern Europe flood and drought periods are controlled by the negative and positive NAO phases, respectively. Fluctuations in atmospheric circulation related to the NAO index are an important driver of climate variability over the Mediterranean region [13
] as well, in particular for weather extremes, such as flood and drought. For example, the authors of [14
] analyzed relationships between NAO phases and drought indices and the influence of the NAO on droughts in the entire Mediterranean region between 1901 and 2006. They found that during periods of positive NAO phase, droughts were recorded in Southern Europe (the Iberian Peninsula, Italy and the Balkans) and areas of Turkey and Northwest Africa.
Observations from the Gravity Recovery And Climate Experiment (GRACE) mission are a valuable data set of the terrestrial water storage (TWS). Terrestrial water storage is one of the main components of the hydrological cycle and by definition consists of all forms of water above and beneath the surface of the Earth. TWS variability is the sum of changes in snow, soil moisture, and ground water. Terrestrial water storage anomaly (TWSA) is a very important part of physically based model equations that represent water fluxes and storage processes. In [15
], the main applications of GRACE data for groundwater monitoring on regional to global scales were reviewed. This study presented different approaches for estimating groundwater storage variations by using GRACE data along with hydrological model outputs, in situ data and other remotely sensed observations. Another use of GRACE data was presented in [16
], where GRACE and reanalysis flux rate data were used to study water storage acceleration globally and locally. Regions of different size, climate conditions, and El Ni
o-Southern Oscillation (ENSO) influence were considered. The authors of [16
] found a positive trend and a robust negative acceleration, that was to some extend reconstructed from reanalysis data and that was also present in non-ENSO signals. Correlation coefficients between GRACE and reanalysis fluxes time series ranging from 0.5 to above 0.9 were reported in [16
]. The authors of [16
] also found that dry periods leading to water storage increased and wet periods of negative storage rates may dominate long-term trends. GRACE data can also be incorporated into groundwater, hydrological and land surface models to improve the simulations of the terrestrial water cycle. The authors of [17
] used 13 years of GRACE observations to assess the accuracy of four global numerical model realizations that simulated the continental branch of the global water cycle, including snow, surface and subsurface water contents. Monitoring of TWSA is of interest in Southeastern Europe, since it is an important component of the hydrological cycle and is related to compound events that are a common phenomenon in the region. The first compound event in Southeastern Europe analyzed using TWSA was the 2007 heat wave in Bulgaria. Time series analysis of TWSA, temperature, precipitation and integrated water vapor (IWV) showed both positive anomalies of temperature and IWV and negative precipitation and TWS anomalies during the 2007 heat wave [18
]. It was found that the July TWSA values were strongly correlated with the end of winter precipitation means. The authors of [18
] demonstrated the potential of GRACE TWS data in studying dry extremes in the hydrology cycle in Bulgaria.
The aim of this paper is to investigate the cross-correlations between the multiple drivers of the hydrology cycle in Bulgaria by using as a case study the wet extreme in 2014. In particular, the GRACE time series analysis of wet extremes is applied for the first time in Southeast Europe and its added value to the widely used temperature and precipitation time series is demonstrated. TWSA has a potential to be one of the key flood early warning indicators advancing the seasonal flood predictions in the region. In Section 2
the data sets and the analysis methods are presented. Section 3
presents the time series decomposition of temperature, precipitation and TWS, as well as cross-correlations with four climate indices for the study period of 2003–2014. In Section 3.3
we evaluate atmospheric reanalysis (ERA5) temperature and precipitation time series over the 2003–2014 period, as ERA5 is a valuable tool for regional studies. There are two reasons for this. Firstly, Southeastern Europe topography makes the modeling of the interaction between air mass and topography challenging. Secondly, precipitation is generally weakly modeled as it depends on the interaction between atmospheric dynamic and hydrologic cycle. Conclusions are presented in Section 4
One of the most noticeable consequences of climate change will be the impact on the hydrology cycle, i.e., on the water cycling between ocean, atmosphere and land [45
]. The extremes in the hydrology cycle, such as dry and wet spells, are already taking place on a regional scale, and understanding their characteristics has been possible in the last two decades by using time series from satellite missions such as GRACE and GRACE-Follow on. The first demonstration of regional GRACE-derived TWS depletion was the 2003 heat wave in Western Europe [46
]. Drought occurrence and severity were quantified by Thomas et al. [47
] by calculating the magnitude of the deviation of regional and monthly TWSA from the GRACE climatology. A recent work by Boergens et al. [48
] studied the 2018–2019 droughts in Central Europe and reported TWS deficits of 73% and 94% of the mean amplitude of seasonal water storage variations, respectively. The study also reported that the water deficits in 2018 and 2019 were the largest observed in GRACE and GRACE-Follow on time series, and importantly those deficits were not expected to recover within one year. According to the European State of the Climate report, compiled by the Copernicus Climate Change Service, in 2018 and partly in 2019 Southeast Europe experienced higher than average summer precipitation [49
]. The recent and past wet summers in Southeast Europe motivated this quantification of the cross-correlation and possible time lags between different drivers. For the selected case study of the year 2014 we showed that the TWSA annual cycle had a major deviation from the 12-year averaged annual cycle. This is an important finding, which can be used as a criteria for an early warning of wet extremes in Bulgaria. Furthermore, the cross-correlation between TWSA and precipitation had a clearly detectable time lag. Periods with increasing or decreasing precipitation were followed by 3 to 6 months increasing or decreasing TWSA trends in Bulgaria, respectively. This time lag is consistent with a global study by Humphrey et al. [38
], which reported a range between 1 and 5 months depending on the region. Our work is the first step towards the estimation of flood potential in Bulgaria. Reager and Famiglietti [51
] proposed a flood index by determining repeated maxima in the TWSA, beyond which additional precipitation will have flood generation potential provided increases in runoff or evaporation are not possible. The establishment of the European Gravity Service for Improved Emergency Management (EGSIEM, [52
]) prototype was a major development that will support the operational implementation of regional flood early warning systems across Europe. Thus our study is timely and highly relevant to the EGSIEM uptake in Bulgaria and Southeast Europe.
Time series analyses of temperature, precipitation and TWS and their correlations with four climate oscillation indices were performed for the 2003–2014 period in order to assess the 2014 wet extremes in Bulgaria. The year 2014 started with negative precipitation anomalies in January and February, which in combination with positive temperature anomalies indicated a drier and warmer winter followed by five consecutive wetter than usual months from March to July and then an even wetter September with precipitation over 200% of the 12-year monthly mean. Anomalies of TWS, derived from three GRACE time series, for the 12-year period had an annual cycle and were mostly positive in the first half of the year from February to June and negative in the second half from July to December. By contrast, the 2014 TWSA was negative from January to April and positive from May to August 2014. Cross-correlations between the long-term trends for the 2003–2014 showed: (1) weak to moderate negative correlations of temperature long-term component with precipitation and TWSA long term components and (2) moderate positive correlations between precipitation long-term component and TWSA long-term component. For the 12-year period, temperature had a positive linear trend of 0.1 C/year, the precipitation trend was negative (1 mm/year) and TWSA decreased in the range of 2–4 mm/year. The seasonal precipitation cycle peaked in late spring and early summer (May and June). The TWSA peak was in early spring (March and April) while the temperature peak was in summer, i.e., it was out of phase with the TWSA peak.
The long-term variability components of temperature and precipitation from SYNOP and ERA5 showed very good agreement. The year 2014 stood out as the year with the highest increase in the long-term component of the precipitation over the studied 12-year period. For April and September 2014 large subseasonal precipitation residuals were observed, which coincided with a significant number of registered flood events, with a secondary peak observed for July 2014. The MOI and NAO index were weakly correlated with observed precipitation and weakly to moderately correlated with the TWSA. MOI was positively correlated with the precipitation and TWSA, while NAO was negatively correlated with the same parameters. An above-average MOI was more likely to be followed by an above-average TWSA one month later, while an above-average NAO is more likely to be followed by a below-average TWSA one month later.
As demonstrated in this study, the 2014 wet extremes in Bulgaria were closely related to the largest positive precipitation and TWS anomalies recorded for the 12-year period (2003–2014). The presented case study gives a valuable insight about the deviation of the components of annual hydrology cycles during extremely wet years such as 2014. The observed time lags between precipitation and TWSA may be used as a guidance in operational forecasting of wet extremes in Bulgaria. The GRACE time series are essential to the understanding and interpretation of the hydrology cycle, and analysis of the GRACE-Follow on time series will be performed in the near future.