Droughts and Climate Change in Bulgaria: Assessing Maize Crop Risk and Irrigation Requirements in Relation to Soil and Climate Region Z

Bulgarian Journal of Agricultural Science, 21 (No 1) 2015, 35-53 Agricultural Academy

DROUGHTS AND CLIMATE CHANGE IN BULGARIA: ASSESSING MAIZE CROP RISK AND IRRIGATION REQUIREMENTS IN RELATION TO SOIL AND CLIMATE REGION Z. POPOVA*1, M. IVANOVA1, L. PEREIRA2 , V. ALEXANDROV3, M. KERCHEVA1, K. DONEVA1 and D. MARTINS2 1Agricultural Academy, “N. Poushkarov” Institute of Soil Science, Agrotechnology and Plant Protection, BG - 1080 Sofia, Bulgaria 2Technical University of Lisbon, Institute of Agronomy, Tapada de Ajuda CEER-Biosystems Engineering, 1349-017 Lisboa, Portugal 3Bulgarian Academy of Sciences, National Institute of Meteorology and Hydrology, BG - 1784 Sofia, Bulgaria

Abstract POPOVA, Z., M. IVANOVA, L. PEREIRA , V. ALEXANDROV, M. KERCHEVA, K. DONEVA and D. MARTINS, 2015. Droughts and climate change in Bulgaria: assessing maize crop risk and irrigation requirements in relation to soil and climate region. Bulg. J. Agric. Sci., 21: 35–53

This study aims at assessing maize cropping risk due to observed trends for drought aggravation for the maize crop season which is associated with possible climate change trends relative to precipitation, temperature and reference evapotranspiration (ETo) at selected weather stations in Bulgaria. Water balance and relative yield computations were performed with the model WinISAREG after its calibration and validation for maize at various locations of Bulgaria and using long-term experimental data. Rainfed maize is associated with a great yield variability (29 < Cv < 72%) due to inter-annual and spatial climatic variability during the maize season. The largest yields variability were found for Sandanski and Plovdiv when a soil with low total available water (TAW = 116 mm m-1) is considered. The least variable yields are those for Sofia when TAW = 180 mm m-1. Basing upon economic considerations, relative yield decreases (RYD) were computed with the threshold of 60 and 48% of the potential maize productivity in Plovdiv and Sofia. Maize production is risky in 32% of years in Plovdiv when TAW is large, which is the double of risk in Sofia. If TAW is medium the risky years double and reach 50% of years in Varna. A relationship between the SPI-2 index computed for “July-Aug” with the simulated RYD of rainfed maize was found. It is quite significant in Plovdiv where R2 > 91% was found. Results are less good for Sandanski and Sofia (73 < R2 < 83%). Results indicate that when rainfed maize is grown on soils of large TAW, maize development is less affected by the water stress. Economical losses are produced when high peak season (July-Aug) SPI-2 is less than +0.2 in Sandanski, -0.50 in Plo- vdiv and -0.90 in Sofia. In North Bulgaria the respective SPI threshold ranges between -0.75 at Lom and -1.5 at Pleven. When irrigation is considered, a relationship was also found relating the net irrigation requirements (NIRs) with SPI-2 “July-Aug”. Consequently, the corresponding economic thresholds relative to the yield losses were computed for all locations this allowed to estimate the NIR thresholds that may lead to favourable cropping returns. The derived reliable relationships and related SPI-2 thresholds relative to “July-Aug”, under which soil moisture deficits lead to severe impact of drought on rainfed maize yield for the studied climate regions and soil groups were used for mapping the risk of rainfed maize yield decreases as a func- tion of drought intensity, which combined with maps relative to NIR to overcome drought effects on maize production. Key words: Drought Intensity-Yield Relationships, Isareg Model, Net Irrigation Requirements, Rainfed Maize Vulnerability, Relative Yield Decrease, SPI-index

E-mail: *[email protected], [email protected] 36 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins

Introduction and Objective This study aims at assessing maize cropping risk and sup- porting drought risk management due to observed trends for Climate uncertainty affects performance and manage- drought aggravation for the maize crop season, which is asso- ment of agriculture. Extreme weather events, as drought, lead ciated with negative trends relative to precipitation, air tem- to substantial increase in agricultural risk and unstable farm perature and reference evapotranspiration (ETo) by applica- incomes. The necessity to develop methodologies and simu- tion of the validated WinISAREG model and the SPI-2 index lation tools for better analysing, forecasting and managing at selected locations of South and North Bulgaria. the risk of agricultural drought is evident after the extremely dry 2000, 2007 and 2012 when the average maize grain Materials and Methods yield in Bulgaria dropped to less than 1.8 t ha-1 (Statistical Yearbooks). The recent study by Olesen et al. (2011) on the Climate data and analyses impacts of climate change in European agriculture supports The study was performed for various locations in Bul- the hypothesis that Bulgarian agriculture is becoming more garia: Lom, Pleven, Silistra and Varna as to represent the vulnerable to droughts and climate variability. This study northern regions, and Sofia, Plovdiv, Stara Zagora and San- shows that most negative effects are expected for the Pan- danski in the southern regions (Figure 1). Different climates nonian zone – that includes Bulgaria, Hungary, Serbia, and are therefore considered: a moderate continental climate in Romania where increased heat waves and drought events are Sofia, Pleven, Lom and Silistra, a transitional continental cli- expected. mate at Stara Zagora and Plovdiv, a northern Black Sea cli- Guttman (1998) recommended the Standardized Pre- mate in Varna, and a transitional Mediterranean climate in cipitation Index, SPI (McKee et al., 1993; 1995) because it Sandanski. is standardized and contains a probabilistic interpretation, Monthly precipitation and reference evapotranspiration hence can be used in risk assessment and decision-making. (ETo) relative to the period 1951 to 2004 at selected locations Although analysing precipitation characteristics is funda- of northern and southern Bulgaria are presented in Figure 2. mental when studying drought risk, if there is the need to per- Precipitation represents wet, average and dry years, i.e., when form comparisons among areas having different precipitation the probability for being exceeded is respectively 10, 50 and regimes or climate characteristics, adopting a standardized 90%. The precipitation during the maize cropping season variable like the SPI is preferred to capture dryness and wet- (“May-Sept”) shows a great inter-annual variability and a ness conditions (Bordi et al., 2009). non-negligible seasonality, with less precipitation during the In our recent studies trend tests were applied to monthly months of July, August and September, when maize flower- precipitation, maximum and minimum temperature and to ing and yield formation occur (Figures 2a, 2c, 2e and 2g). the Standardized Precipitation Index with 2-month time step There is also an evident spatial variability, with larger pre- (SPI-2) relative to the period of 1951-2004. Negative trends cipitation in Sofia and the northern locations. were identified for precipitation and SPI-2 at various locations, mainly in the Thrace Plain, indicating that dryness is likely to be increasing in Bulgaria (Popova et al., 2013a; Pop- ova et al., 2013b). In previous studies the WinISAREG model (Pereira et al., 2003), an irrigation scheduling simulation tool for simulating the soil water balance and evaluating the respective impacts on crop yields, was validated using independent data sets relative to long term experiment with early and late maize hybrids (Popova et al., 2006b; Popova, 2008; Popova and Pereira, 2011). Popova and Kercheva (2005) used the CERES models to assess impacts of drought on maize and wheat and identified the role of soil characteristics on those impacts. Popova and Pereira (2008) combined the use of the soil water balance model ISAREG with the Stewart’s yield model (Stewart et al., 1977) to assess impacts of climate variability on irrigation scheduling and management in the Thrace Plain Fig. 1. Experimental fields ofISSAPPNP and of Bulgaria. meteorological stations of NIMH in Bulgaria Droughts and Climate Change in Bulgaria 37

140 6 ) 1 - m 120 5 y m a , d

n 100

o 4 m t i

80 m t a i (

p 3 i M c 60 e P r -

P 2 40 o T 20 E 1 0 0 (a) (b)

140 6 ) m 1 120 - 5 m y , a d n

100

o 4 m t i m

t a 80 ( i 3 p i M

c 60 P e - r

o 2 P

40 T E 20 1 0 0 (c) (d)

140 6 ) 1 - m 120 5 y m a , 100 d

n 4 o m t i

80 m ( t a 3 i p M

i 60 P c -

e 2 o r 40 T P 20 E 1 0 0 (e) (f)

140 6 120 m )

1 5 m -

, 100 y n a o

d 4

t i 80 t a m i p m 3 i (

60 c e M r P

P 2

40 - o

T 1 20 E

0 0 (g) (h)

AVERAGE 1.28*STDEV Fig. 2. Average monthly precipitation totals (mm) and reference evapotranspiration ЕТо-РМ (mm day-1) (□); bars represent the 80% confidence interval, 1951-2004; (a) and (b) Sofia; (c) and (d) Plovdiv; (e) and (f) Pleven; (g) and (h) Varna 38 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins

ETo was computed with the PM-ETo equation (Allen et al., follows a regular seasonal distribution, with maxima in July 1998) using only temperature data as described by Popova et and August when precipitation is smaller (Figure 2). al. (2006a). ETo refers to low, average and high climatic de- Probability curves of occurrence for seasonal precipita- mand conditions, when ETo values are exceeded with a prob- tion and reference evapotranspiration ETo relative to the ability of 90, 50 and 10% respectively. ETo shows much less whole “May-Sept” season and the high demand “July-Aug” inter-annual variability than precipitation, as indicated by the period are compared for six locations in Figures 3 and 4. The bars representing the 80% confidence interval, and is higher seasonal precipitation “May-Sept“are the highest and practi- in southern regions reaching 5.6 mm day-1 at Sandanski and cally identical at Sofia and Pleven, varying between 120 and 5.3 mm day-1 at Plovdiv and Stara Zagora (Figure 2d). 600-700 mm (Figure 3a). As it may be observed compar- It drops to 4.9-4.3 mm day-1 in the northern locations of ing the probability curves in the figure, the curves for Sofia Lom, Silistra and Varna (Figure 2h). Peak Season ETo reach- and Pleven are above all others that mean that the seasonal es minima of 4.4-4.0 mm day-1 in Sofia field (Figure 2b). ETo drought is expected to be mitigated there. Comparing the six

700 400 ) m ) 600 m m ( m R (

, 300 500 R , I ) V I X - - 400 V I

V n (

200 i o n t

a 300 i o i t t a i p i t c e i p r

200 c P e

r 100 P 100

0 0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 (a) Probability of precipitation R (V-IX), P R (%) (b) Probability of precipitation R (VII -VIII), P R (%)

Pleven Silistra Sofia Plovdiv Sandanski Varna Fig. 3. Comparison of precipitation (mm) probability of exceedance curves for six climate regions relative to: а) Cropping Season (V-IX) and b) High Demand Season (VII-VIII), 1951-2004

900 400

850 ) )

800 m 350 m m ( m (

750 I I X V - - V I

V M 700 300 P - M o P - T o

E 650 T E 600 250

550

500 200 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 (b) (a) Probability of exceedance ETo -PM (V-IX), P (%) ETo Probability of exceedance ETo-PM (VII-VIII), PETo (%)

Pleven Silistra Sofia Plovdiv Sandanski Varna Fig. 4. Comparison of reference evapotranspiration ЕТо-РМ (mm) probability of exceedance curves at six climate regions relative to: а) Cropping Season (V-IX) and b) High Demand Season (VII-VIII), 1951-2004 Droughts and Climate Change in Bulgaria 39 curves it may be concluded also that seasonal drought sever- ule using selected irrigation thresholds, executing the water ity should increase for rainfed maize in the following order: balance without irrigation, and computing the net irrigation Sofia, Pleven, Silistra, Plovdiv, Varna and Sandanski. requirements and others. Yield impacts of water stress are Regarding the peak demand period “July-Aug” differenc- assessed with the Stewart’s one-phase model [(1-Ya/Ymax) es in precipitation totals are about half of those related to the = Ky (1-ETa/ETmax)] (Stewart et al., 1977; Doorenbos and whole season (Figure 3b). However inter-regional differences Kassam, 1979), whose Ky value of 1.6 was calibrated as re- become much smaller in the dry years (P > 90%) and aug- ferred in the above mentioned studies. ment over the average and wet years. Monthly precipitation and ETo data series (1951 to 2004), The inter-seasonal variability of reference evapotranspira- with ETo computed as described by Popova et al. (2006a) are tion ETo-PM “May-Sept”, compared with precipitation, is much relative to the eight locations referred above. lower (Figure 4). ETo-PM “May-Sept” is the largest in Sandan- Soil characteristics ski (740 to 840 mm), about 110 mm less at Pleven and Plovdiv Available soil water holding capacity values, herein the and the lowest in Sofia (Figure 4a). During the high demand TAW values are relative to the three main soil groups occur- season “July-Aug” the differences in ETo between locations are ring in the regions represented by each of the eight locations. half than those relative to ETo “May-Sept” (Figure 4b). TAW values were determined in former studies (Stoyanov, 2008; Boneva, 2012). In southern Bulgaria the most com- Simulation soil water balance and yield model mon soils are the Chromic Luvisols and Cambisols, that have Maize is one of the main summer crops in Bulgaria. Dur- predominantly medium TAW (136 mm m-1), and Vertisols of ing the period 1960-1990 it had been under irrigation over large TAW (170 ≤ TAW ≤ 180 mm m-1). In the plains of north- more than 100 000 ha. In the last 25 years however, regard- ern Bulgaria, TAW ranges from 157 to 180 mm m-1 in Cher- less of the occurrence of severe droughts and dramatic yield nozems soils or TAW > 170 mm m-1 in Vertisols. In the ter- losses referred above, maize is grown under rainfed condi- races along the rivers small TAW (≤ 116 mm m-1) are found, tions (Statistical Yearbooks). Crop parameters required for which correspond to coarse-textured Luvisols and Fluvisoils - modelling consist of crop coefficients (Kc), water depletion Soil map of Bulgaria (Koinov et al., 1998; Boneva, 2012). fractions for no stress (p) and the water yield response factor (Ky), as defined by Allen et al. (1998). Crop parameters were Calibration/validation of crop parameters obtained when calibrating the WinISAREG and Stewart’s Crop data originated from long term field experiments, models using field data. The calibration and validation of the mainly reported by Varlev et al. (1994), Eneva (1997) and WinISAREG and the Stewart’s models for Pustren and Zora Varlev and Popova (1999). The parameters Kc, p and Ky (Stara Zagora), Tsalapitsa (Plovdiv) and Bojurishte (Sofia) were those obtained from the above referred model calibra- were described respectively by Popova et al. (2006b), Popova tion. The related parameterization and data on crop growth (2008), Popova and Pereira (2011), Ivanova and Popova (2011) stages and root depths were extended to other locations using and Popova (2012). Calibration and validation of both mod- data obtained by Rafailov (1995 and 1998), Varlev (2008) and els were performed using data relative to different irrigation Stoyanov (2008), which referred to various maize hybrids management conditions and for rainfed maize (Eneva, 1997; (Popova, 2012; Popova et al., 2012), Varlev et al., 1994; Varlev and Popova, 1999). Combining both the ISAREG and the Stewart’s models, The WinISAREG model, described by Pereira et al. (2003) it was possible to estimate crop water and irrigation require- uses the soil water balance approach proposed by Dooren- ments and the yield impacts of water stress for each year of bos and Pruitt (1977) and the updated methodology proposed the series 1951-2004, i.e., the relative yield decrease (RYD) by Allen et al. (1998) to compute crop ET and irrigation re- due to water stress. Computations were performed for all quirements. Data required to perform the soil water balance eight locations and using the available TAW data and for all with ISAREG consist of: (1) weather data on precipitation years of the weather data series. It resulted three RYD series, and reference evapotranspiration ETo; (2) soil water data, one for each soil type – low, medium and high TAW - for the total available soil water (TAW, mm m-1), i.e., the differ- each location. Empirical curves relating RYD with the prob- ence between soil water storage at field capacity and wilting ability of their occurrence (PRYD) were therefore built. The point for a soil depth of 1.0 m (Allen et al., 1998), and (3) crop corresponding net irrigation requirements (NIR) were also data relative to the crop development stages and correspond- estimated for all of years of the series and related empirical ing dates, crop coefficients, root depths and the soil water probabilities of exceedance curves (PNIR) were also built. depletion fractions for no stress. The model allows various RYD thresholds representing the values when yields be- simulation options including to simulate an irrigation sched- come insufficient to achieve a positive farm return were identi- 40 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins fied for all locations. Thresholds base upon the assumption that selected weather stations. Results of the tests show that, rela- rainfed maize cultivation is profitable if the harvested yield is tive to ETo for maize crop season (V-IX), a significant trend above 4500 kg; however, this value changes with production was observed for the period 1970-2004 (Figure 5). costs and commodity prices and needs to be updated for prac- The detected increase of seasonal ETo is 1.0 mm yr-1 at So- tical use. These thresholds varied from one location to another fia and Silistra, up to 2.0-2.3 mm yr-1 at Stara Zagora, Pleven, because potential yield productivity was different among all Lom and Sandanski and reaching a maximum of 2.6 mm yr-1 locations. For example, the RYD thresholds indicative of eco- at Plovdiv and Varna. In all stations the months of July and nomic losses corresponds to 60% at Plovdiv and to 48% at So- August contribute most for seasonal ETo increase (Table 1). fia: for Tsalapitsa, Plovdiv, the average potential yield for the If one considers the whole period (1951-2004), the magnitude period 1971-1991 using tardy maize hybrids (Н708, 2L-602 of the trend is about half at Lom, Varna and Sandanski and -1 and ВС622) is Ymax = 11 228 kg ha while for Gorni Lozen, So- only ¼ at Pleven and Plovdiv thus indicating that the aggrava- -1 fia, maxY = 8460 kg ha was observed for the same period with tion of climate conditions is higher in recent times (Table 2). semi-early maize hybrids (HD-225, SK-48A, Px-20, P37-37). Results relative to seasonal precipitation (V-IX) show that, contrarily to ETo, a significant inter-seasonal variability Results and Discussions combined with a negative trend were observed for the recent 35 years. The magnitude of the trend ranges from -2.8 mm Trends of aggravation of climate conditions during 1970-2004 yr-1 at Stara Zagora and Sofia to -2.3 mm yr-1 (Silistra) and Trend analyses were applied to climate data, mainly refer- -1.8 mm yr-1 (Pleven, Plovdiv) and reaching a minimum of ence evapotranspiration (ETo) and precipitation, relative to -1.0 mm yr-1 at Lom (Table 3). The detected trends in Peak

800 800 y = 0.46 x - 216.08 y = 1.05 x - 1501 RMSE= 34.9 mm R2 = 0.14 RMSE= 28.3 mm ) 700 ) 700 m m m m ( ( y = 2.59 x - 4454.1 M M 2

P P R = 0.48 - 600 - 600

o o RMSE= 27.4 mm T T E E

500 500 0 3 6 9 2 5 8 1 4 7 0 3 6 9 2 5 8 1 4 0 3 6 9 2 5 8 1 4 7 0 3 6 9 2 5 8 1 4 5 5 5 5 6 6 6 7 7 7 8 8 8 8 9 9 9 0 0 5 5 5 5 6 6 6 7 7 7 8 8 8 8 9 9 9 0 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 а) b) Year Year

800 y = 0.55x - 388.26 800 RMSE=38.1 mm y = 1.60x - 2552.9 ) ) 700 700 R2 = 0.52 m m RMSE=24.1 mm m m ( (

M M P P

- y = 2.57x - 4478.3 - 600 y = 2.30 x - 3880.6 600

o 2 o R2 = 0.33 R = 0.58 T T

E RMSE=22.4 mm E RMSE= 33.5 mm

1.05-30.09. AVG on 3 years Trend line (1951-2004) Trend line (1970-2004) Fig. 5. Seasonal ETo-PM “May-September” (mm) (o) at: a) Sofia; b) Plovdiv; c) Pleven and d) Varna; comparison of trendlines relative to 1951-2004 and 1970-2004 Droughts and Climate Change in Bulgaria 41

Demand season (VI-VIII) are responsible for the seasonal When the whole period (1951-2004) is considered, the precipitation decrease (V-IX). magnitude of the trend for Peak season precipitation, com-

Table 1 ЕТо Trend Analysis (b refers to slope coefficient), 1970-2004 Seasonal ET Peak Season ET High Peak Season Periods 0 0 (V-IX) (VI-VIII) (VII-VIII) b, ΣET , b, ΣET , b, ΣETcrop= Stations -1 0 -1 -1 0 -1 -1 KcΣET , mm yr mm 35 yr mm yr mm 35 yr mm yr mm 35 yr0 -1 Plovdiv 2.59 90.65 1.96 68.6 1.48 67.3 Pleven 2.30 80.5 1.83 64.05 1.43 65.1 Varna 2.57 89.95 1.78 62.3 1.31 59.6 Sandanski 2.37 82.95 1.78 62.3 1.30 59.2 Stara Zagora 1.90 66.5 1.57 54.95 1.26 57.3 Lom 2.27 79.45 1.67 58.45 1.24 56.4 Silistra 1.03 36.05 0.85 29.75 0.74 33.7 Sofia 1.05 36.75 0.86 30.1 0.70 31.9

Table 2 ЕТо Trend Analysis (b refers to slope coefficient), 1951-2004 Seasonal ET Peak Season ET High Peak Season Periods 0 0 (V-IX) (VI-VIII) (VII-VIII) b, ΣET , b, ΣET , b, ΣETcrop= Stations -1 0 -1 -1 0 -1 -1 KcΣET , mm yr mm 54 yr mm yr mm 54 yr mm yr mm 54 yr0 -1 Plovdiv 0.46 24.84 0.36 19.44 0.17 11.9 Pleven 0.55 29.7 0.43 23.22 0.21 14.7 Varna 1.60 86.4 1.02 55.08 0.66 46.3 Sandanski 0.82 44.28 0.61 32.94 0.31 21.8 Stara Zagora -0.05 -2.7 0.06 3.24 -0.06 -4.2 Lom 0.94 50.76 0.71 38.34 0.38 26.7 Silistra 0.81 43.74 0.55 29.7 0.36 25.3 Sofia 0.06 3.24 0.06 3.24 0.01 0.7

Table 3 Precipitation Trend Analysis (b refers to slope coefficient), 1970-2004 Seasonal R Peak Season High Peak Season Periods (V-IX) (VI-VIII) (VII-VIII) b, ΣR, b, ΣR, b, ΣR, Stations mm yr-1 mm 35 yr-1 mm yr-1 mm 35 yr-1 mm yr-1 mm 35 yr-1 Plovdiv -1.84 -64.4 -1.48 -51.8 -1.04 -36.4 Pleven -2.14 -74.9 -1.82 -63.7 -0.72 -25.2 Varna 1.70 59.5 0.33 11.55 0.52 18.2 Sandanski -0.04 -1.4 -0.20 -7 -0.68 -23.8 Stara Zagora -2.82 -98.7 -2.41 -84.35 -1.52 -53.2 Lom -1.06 -37.1 -1.09 -38.15 0.09 3.2 Silistra -2.29 -80.15 -1.79 -62.65 -1.29 -45.2 Sofia -2.69 -94.15 -2.61 -91.35 -2.04 -71.4 42 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins paring with 1970-2004, is much smaller at Sofia, Stara Zago- tions (Pleven). Moreover, both figures show that cultivation ra, Pleven and Silistra (Figure 6a and 6c). in soils with high TAW (180 mm m-1) leads to much less RYD Globally the regions of Stara Zagora, Plovdiv, Sofia, Silis- relative to soils with low TAW. tra and Pleven, with a decrease in peak season precipitation The economical RYD thresholds for Pleven and Plovdiv (VI-VIII) and an increase in evapotranspiration demands are 67 and 60% respectively, correspond to the average yield -1 could have consequences in the occurrence of drought for potential for the period 1971-1991 of Ymax = 13 790 kg ha -1 summer crops in recent times (1970-2004). at Pleven and Ymax = 11 230 kg ha at Plovdiv. For Pleven, the RYD threshold refers to only 12% of years if a soil with Drought and climate impacts on yield and risky years high TAW is considered, and to 30% of the years when maize Results in Figure 7 refer to the empirical probability of ex- is cropped in a soil with low TAW (Figure 7b). Differently, ceedance curves of the relative yield decrease at Plovdiv and for Plovdiv the threshold corresponds to 30 and 68% of the Pleven, respectively in southern and northern Bulgaria, where years, respectively (Figure 7a). These examples clearly show RYD is affected by the soil water holding capacity, consider- the combined effects of climate and soil type. ing the late maize hybrids H708, 2L602, BC622. Results in The empirical probability of exceedance curves of RYD Fig.7a indicate that when soils have low water holding capac- relative to six climate regions - Silistra, Pleven and Varna ity, as 116 mm m-1, rainfed maize yields are not only affected in northern Bulgaria, and Sofia, Plovdiv and Sandanski in by droughts but also by climate variability. When comparing South - are compared in Figure 8a for rainfed maize culti- Figures 7a and 7b, it becomes evident that the vulnerability vated in soils of medium TAW (136 to 157 mm m-1). In Fig- is much higher for southern (Plovdiv) than for northern loca- ure 8b the probability curves for the same locations are com-

600 600 y = -2.61 x + 5360.3 500 y = -0.50 x + 1165.5 500 y = -0.91 x + 1929.8 y = -1.48 x + 3075.2 RMSE= 74.0 mm RMSE= 78.0 mm ) ) RMSE= 70.7 mm RMSE= 71.0 mm m 400 m 400 m m ( (

n 300 n 300 o o t i t i t a t a i 200 i 200 p p i i c c e e

r 100 r 100 P P 0 0 0 3 6 9 2 5 8 1 4 7 0 3 6 9 2 5 8 1 4 0 3 6 9 2 5 8 1 4 7 0 3 6 9 2 5 8 1 4 5 5 5 5 6 6 6 7 7 7 8 8 8 8 9 9 9 0 0 5 5 5 5 6 6 6 7 7 7 8 8 8 8 9 9 9 0 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 a) Year b) Year

600 600 y = 0.08 x + 34.9 500 y = -1.82 x + 3809.4 500 y = 0.33 x - 530.3 RMSE= 77.2 mm y = 0.46 x - 789.6

) RMSE= 76.2 mm ) RMSE= 54.6 mm RMSE= 57.0 mm m 400 m 400 m m ( (

n 300 n 300 o o t i t i t a t a

i 200 200 i p p i i c c e 100 e 100 r r P P 0 0 0 3 6 9 2 5 8 1 4 7 0 3 6 9 2 5 8 1 4 0 3 6 9 2 5 8 1 4 7 0 3 6 9 2 5 8 1 4 5 5 5 5 6 6 6 7 7 7 8 8 8 8 9 9 9 0 0 c) d) 5 5 5 5 6 6 6 7 7 7 8 8 8 8 9 9 9 0 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 Year Year

1.06-31.08. AVG on 3 years Linear (1951-2004) Linear (1970-2004) Fig. 6. Precipitation total for peak demand period “June-August” (mm) (o) at a) Sofia; b) Plovdiv; c) Pleven and d) Varna; comparison of trendlines for 1951-2004 and 1970-2004 Droughts and Climate Change in Bulgaria 43 pared when the soil has a large TAW. It can be noticed that the same for soils with medium and high TAW. The dif- the northern locations have the respective probability curves ference is that RYD increase when soil TAW decreases. grouped below the Plovdiv and Sandanski curves, while So- Comparing all six RYD curves, it may be concluded that fia behaves differently of other southern locations and the the vulnerability increases in the following order: Sofia, respective curve is the lowest because precipitation is higher Pleven, Silistra, Varna, Plovdiv and Sandanski thus indi- there. Consequently the vulnerability to droughts and cli- cating that the increase of vulnerability depends mainly mate variability is smaller there. The RYD curve for San- upon the amount of precipitation during the maize crop danski is above all others, which means that vulnerability is season (Figure 3). highest in this southern region where climate approaches the When TAW = 180 mm m-1 only 10% of the years are at Mediterranean climate. risk of economic losses at Pleven and Silistra. When TAW As it may be observed comparing Figure 8a and 8b, is medium (157 mm m-1) the risky years are 18 and 35% in the relative position of the empirical probability curves is those two sites and reach 50% in Varna (Figure 8b). RYD (%)

100

80 a) 70

60 100

90 50

RYD (%) 80 40 ) 70 % (

D 60 30 Y

R 32 % risky years (17/54) for TAW=116 50 20 19 % risky years(10/54) for TAW=136-157 40 13 % risky years(7/54) for TAW=180 30 10

b) 20 0 10 Year 1958 2000 1993 1963 1965 2003 1988 1952 1961 1990 1994 1996 1962 1974 1980 1956 1989 1986 1964 1951 1997 1972 1981 1968 1960 1978 1971 1983 1995 1955 1991 1979 1970 1977 1957 1975 2002

Year 0 8 0 3 3 5 3 8 2 1 0 4 6 2 4 0 6 9 6 4 1 7 2 1 8 0 8 1 3 5 5 1 9 0 7 7 5 2 5 0 9 6 6 0 8 5 6 9 9 9 6 7 8 5 8 8 6 5 9 7 8 6 6 7 7 8 9 5 9 7 7 7 5 7 0 9 0 9 9 9 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 PRY DP(RYD%) (%)1 2 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1.4 3 5 7 9 11 12 14 16 18 20 22 24 25 27 29 31 33 35 36 38 40 42 44 46 47 49 51 53 55 57 59 60 62 64 66 68 70 71 73 75 77 79 81 82 84 86 88 90 92 94 95 97 99 b) 1,43 5 7 911121416182022242527293133353638404244464749515355575960626466687071737577798182848688909294959799 RYD Threshold late hybrids TAW=180 mm m-1 TAW=136-157 mm m-1 TAW=116 mm m-1 Fig. 7. Probability of exceedance curves of relative yield decrease RYD (%) for rainfed maize comparing soil groups of small, medium and large total available water (TAW) at: a) Plovdiv, South Bulgaria; b) Pleven, North Bulgaria, Ky=1.6, 1951-2004 44 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins

However, the RYD threshold values need to be updated Irrigation requirements in relation to climate and every year to reflect the actual economic farming conditions soil characteristics and commodity prices. Probability of exceedance curves of net irrigation require- As referred above the RYD threshold is the value when ment (NIR, mm) for maize were built using WinISAREG yield becomes insufficient to achieve a positive farm return. over the period 1951-2004 (Popova, 2012; Popova et al., 2012; Thus, relationships between the yield threshold values У (t 2013a). Results relative to Pleven (Figure 10a) show that for ha-1) and the grain price G (BGN t-1) are derived at differ- soils of large TAW (180 mm) NIR vary 0 mm in wet years, -1 ent levels of production expenses PE (BGN ha ), as shown i.e., when the probability of exceeding is PNIR > 95%, 80-180 in Figure 9a. Since the grain price G ranges predominant- mm in average demand years (40% < PNIR < 75%), and 300- -1 ly between 200 and 400 BGN t over the last 20 years, the 330 mm in very dry years (PNIR < 5%). In soils with medium yield threshold values vary between 2 and 5.5 t ha-1 (Figure TAW (157 mm), NIR increase and reach 370 mm in the very 9a). When the yield threshold values are expressed in relative dry year. Differently, NIR in Plovdiv and Stara Zagora (Fig- terms, as RYD or [у = У/Уmах = 1-RYD], relationships are ure 10b) are about 100 mm larger than in Pleven, while in universal and could be compared for the main agro-climatic Sandanski NIR are about 210 mm higher. regions and the different levels of PE (BGN ha-1) (Figure 9b and 9c). Trend analyses of yields and irrigation requirements Rainfed maize is associated with great yield variability An empirical trend analysis was performed for both RYD with the coefficient of variation relative to 1951-2004 in the and NIR (Figures 11 and 12) which shows that they both may range 29 < Cv < 72% (Table 4). The coefficient of variability increase for the last years of the observation period (1951- ranges 29-42% in Sofia, with Cv = 29% for soils with larger 2004). Then, RYD for Plovdiv region may increase by 0.35% TAW. Higher Cv are for the other southern regions, namely in year-1 or a decrease of about 40 kg ha-1 year-1 may be expected Sandanski (Cv = 72%), Plovdiv (Cv = 69%) and Stara Zagora if irrigation is not applied (the black trendline, Figure 11b). (Cv = 59%). The variability of rainfed maize in the Danube Similarly, RYD for the region of Stara Zagora may increase Plain (Pleven, Varna and Silistra) is much lower than in the by 0.14% year-1 (or a decrease by 15 kg ha-1 year-1results not Thracian Lowland. However, in the region of Lom the yield shown). However, significant trends were not found for other variability is also high (35 < Cv < 55%). regions and further analyses are required. If one considers

100 100

90 90

80 80

70 70

60 60 ) ) % % ( (

50 50 D D Y Y R 40 R 40 18 % risky years for Pleven 30 35 % risky years for Silistra 30 12 % risky years for Pleven 39 % risky years for Sofia 10 % risky years for Silistra 20 59 % risky years for Plovdiv 20 20 % risky years for Sofia 82 % risky years for Sandanski 29 % risky years for Plovdiv 49 % risky years for Varna 63 % risky years for Sandanski 10 10 14 % risky years for Varna

0 0 a) 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 b) P RYD (%) P RYD (%) Pleven RYD Threshold Plovdiv/Silistra RYD Threshold Sofia RYD Threshold Pleven Silistra Sofia Plovdiv Sandanski Varna Fig. 8. Comparison of relative yield decrease (RYD, %) probability of exceedance curves, Ky = 1.6, relative to six climate regions and two soil groups of: a) medium TAW (136-157 mm m-1) and b) large TAW (180 mm m-1) rainfed maize, 1951-2004 Droughts and Climate Change in Bulgaria 45

-1 16 PRODUCTION EXPENCES 800 lv ha Y = 1400 x-1 o n

100 1.00 i t

1 14 -1 - Y = 1200x -1 90 0.90 a c , % a 600 lv ha r e h u 12 -1 -1 80 0.80 , f , t Y = 1000x 800 lv ha e e u u

v a l 70 0.70 10 1000 lv ha-1 y = 97.6 x -1.0

-1 d

v a l Y = 800x v a l

d 60 -1.0 0.60 -1 o l 1200 lv ha y = 84.5 x d o l 8 s h -1 50 -1.0 0.50 o l s h Y = 600x -1 e y = 74.7 x e 1400 lv ha r r s h h h 6 40 -1.0 0.40 e r t t y = 60.7 x d h l D e 30 0.30 t i 4 Y D Y

R 20 0.20 Y

2 R

10 0.10 - 1 0 0 0.00 0 200 400 600 800 0 100 200 300 400 500 600 700 800 y = Grain price, lv t -1 а) Grain price, lv t -1 b)

PRODUCTION EXPENCES 1200 lv ha -1 100 1.00 o n i 90 0.90 t , % a c e r

u 80 0.80 , f e u v a l 70 -1.0 0.70 d y = 142.5 x o l 60 -1.0 0.60 v a l d

s h y = 123.3 x o l e

r 50 -1.0 0.50 h

y = 108.9 x s h e t 40 -1.0 0.40 r D y = 88.6 x h t Y 30 0.30 D R Y

20 0.20 R - 10 0.10 1

0 0.00 y = 0 100 200 300 400 500 600 700 800 c) Grain price, lv t -1 Fig. 9. Relationships between yield threshold values and grain price G (BGN t-1) at different levels of production expenses РЕ (BGN ha-1) when yield is expressed in absolute terms Y (t ha-1) (a) or in relative terms, as relative yield decrease RYD (%) (b) or a fraction y=Y/Ymax=1-RYD (c)

Table 4 Variability of rainfed maize grain yield characterized by the mean value, kg·ha-1, and the coefficient of variation Cv, %, for the considered climate regions and soil groups in Bulgaria, 1951-2004 South Bulgaria North Bulgaria Climate Moderate Transitional Transitional Moderate North Black region Continental Continental Mediterranean Continental Sea Sofia Field Thracian Lowland Sandanski Danube Plain Location Sofia Plovdiv Stara Zagora Sandanski Pleven Lom Silistra Varna semi early (HD- 225, SK-48A, Maize hybrid Px-20, P37-37) Late (H708, 2L602, BC622)

TAW mean C v mean C v mean C v mean C v mean C v mean C v mean C v mean C v -1 -1 -1 -1 -1 -1 -1 -1 (mm·m-1) (kg·ha ) (%) (kg·ha ) (%) (kg·ha ) (%) (kg·ha ) (%) (kg·ha ) (%) (kg·ha ) (%) (kg·ha ) (%) (kg·ha ) (%) Small 116 4421 42 3894 69 3723 59 2292 72 6419 50 4187 55 4866 46 4349 50 Medium 136 4920 37 4550 59 4299 52 2906 59 7237 44 4827 47 5616 40 5156 42 180 5896 29 5915 43 4250 41 8867 34 6094 35 7118 30 6810 30 Large 173 5483 41 46 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins the present observation period 1970-2004 the magnitude of Trend test application to net irrigation requirements NIR for the trend, comparing with 1951-2004, is double in the Thrace Plovdiv region and 1951 to 2004 shows an increase by 1.5 mm (red dashed line, Figure 11b). Moreover, the maximum RYD year-1, i.e. 80 mm in 54 years (Figures 12b). For the region of increase (by 0.90% year-1) is found not only for the southern Stara Zagora NIR increase by 0.5 mm year-1, i.e. 27 mm. Rela- (Stara Zagora) but also for the northern (Pleven) regions (Fig- tive to the present observation period 1970-2004, the trend anal- ure 11c), where a decrease of 100-120 kg ha-1 year-1, or 4.2 t yses show that NIR may increase by 3.5 (Plovdiv, Pleven) to 4.0 ha-1, may be expected thus indicating that the aggravation of mm year-1 (Stara Zagora), i.e. by 120-140 mm 35 years-1, and by climate conditions in recent times is related to important eco- 2.5 mm year-1, i.e. 90 mm 35 years-1, at Sofia (Figures 12a), Lom nomic losses for the rainfed maize crop system. The detected and Silistra that is double than in Varna (results not shown). trend is also substantial (0.60-0.65% yr-1) at Sofia, Silistra and The results indicate that due to the detected climate Lom being half than in Pleven when expressed in kg ha-1 yr-1 change impacts, compared to the available irrigation sched- (Figures 11a and 11d). uling estimation in references (Zahariev et al, 1986), not less

500 450 400 350 m ) (

300 s R

I 250 N 200 150 100 50 0 3 8 5 1 7 4 6 1 4 6 2 9 0 4 1 3 2 4 9 1 4 7 2 8 8 1 6 4 9 3 5 0 7 5 2 8 91 79 63 9 8 6 52 6 85 8 62 9 90 9 67 0 7 5 98 9 86 8 8 6 5 5 8 5 5 8 99 0 9 7 6 7 60 7 83 7 66 8 6 7 95 5 7 7 57 7 0

Year 5 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 9

P (%) 1 1 2000 1 NIRs 1 1 2003 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1.4 3 5 7 9 11 12 14 16 18 20 22 24 25 27 29 31 33 35 36 38 40 42 44 46 47 49 51 53 55 5759 60 62 64 66 68 70 71 73 75 77 79 81 82 84 86 88 90 92 94 95 97 99

-1 -1 a) TAW=157 mm m TAW=180mm m NIRs threshold for TAW=157 mm m-1 NIRs threshold for TAW=180 mm m-1

500

450

400

350

300 m ) (

s 250 R I 200 N 150

100

0 0 10 20 30 40 50 60 70 80 90 100 P NIRs (%) b) Pleven Silistra Sofia Plovdiv Sandanski Varna

Fig. 10. Probability of exceedance curves for net irrigation requirements (NIRs, mm): a) at Pleven region as influenced by soil total available water TAW and b) for climate regions and soils of medium water holding capacity (TAW= 136-157 mm m-1), 1951-2004 Droughts and Climate Change in Bulgaria 47

100 100 y = 0.37 x +49 y = 0.73 x + 35 90 y = 0.66 x + 18 90 80 80 70 70 )

60 ) 60 % ( %

(

50 50 D D Y

40 Y 40 R 30 R 30 20 20 10 10 0 0

a) Year b) Year

100 100 90 y = 0.89 x + 29 90 y = 0.61 x + 38 80 80 70 70 ) ) %

60 % 60 ( (

50 D 50 Y Y R

40 R 40 30 30 20 20 10 10 0 0

c) Year d) Year

Fig. 11. Trendline of relative yield decrease RYD,%, for rainfed maize at: a) Sofia; b) Plovdiv; c) Pleven; d) Silistra; late hybrids (H708, 2L-602 and ВС622), soils of medium water holding capacity (TAW=136-157 mm·m-1), 1951-2004 vs. 1970-2004

400 400 y = 1.49 x - 186 350 350 y = 3.44 x + 108 300 y = 2.53 x + 65 300 250 250 m ) m ) ( (

s 200 s 200 R R I I N 150 N 150 100 100 50 50 0 0

a b) a) Year Year

400 400 350 y = 3.52 x + 112 350 300 300 y = 2.26 x + 131 m )

250 ( 250

m ) ( s

s R 200

200 I R I N 150 N 150 100 100 50 50 0 0

c) Year d) Year

Fig. 12. Trendline of net irrigation requirements NIRs, mm, of maize at: a) Sofia; b) Plovdiv; c) Pleven; d) Silistra; late hybrids (H708, 2L-602 and ВС622), soils of medium water holding capacity (TAW=136-157 mm·m-1), 1951-2004 vs. 1970-2004 48 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins that two additional irrigation events should be required in the Positive SPI values indicate greater than median precipitation, dry years of the present climate. and negative values indicate less than median precipitation.

It is important to relate RYD with the SPI-2Jul-Aug because Using the SPI-2 as water stress indicator for rainfed maize the SPI is standardized and therefore is well comparable and water management among locations and from a year to another (Guttman, 1998; A variety of indices exist to support operational drought Lana et al., 2001; Bordi et al., 2009). In addition, it contains a management (Pereira et al., 2009; Mishra and Singh, 2010; probabilistic interpretation. Thus, when using SPI values, one NDMC, 2013). The Standard Precipitation Index, SPI (Mc- knows when precipitation at a given location is near normal, Kee et al., 1993; 1995), the mostly used in South East Europe or is anomaly in excess or deficit, and may easily compare (Gregoric, 2012), is a standardized index which computation is among locations with different climatic characteristics. SPI- based on the long-term precipitation record cumulated over a 2Jul-Aug provides therefore a quick information for manage- selected time scale, shorter when meteorological or agricultural ment, namely when mapped at country level. droughts are considered, longer when the analysis aims at water SPI-2Jul-Aug were related to RYD for all eight climate regions supply management. That long-term precipitation record is fit- under study. Examples of the linear relationships obtained are ted often to the gamma distribution that is transformed through shown in Figures 13a and 13b. Statistical results are given in Ta- an equal-probability transformation into a normal distribution. ble 5 and they include the determination coefficient (R2), the re-

Table 5 Linear regression parameters relative to relationships between RYD (%) with the SPI-2July-Aug across the considered soil groups and climate regions, Bulgaria, 1951-2004 Soil Groups according to TAW Small: Medium: Large: Region 116 mm m-1 136-157 mm m-1 173-180 mm m-1 RYD % RYD % RYD % Intercept a 79.6 74.1 62.1 Sandanski Slope coefficient b -15 -15.7 -16.1 R2 (%) 75 77 78 Intercept a 67 61.81 51.3 Stara Zagora Slope coefficient b -19.9 -20.5 -20.5 R2 (%) 80 82 83 Intercept a 65.2 59.4 47.2 Plovdiv Slope coefficient b -24.8 -24.7 -23.6 R2 (%) 92 92 91 Intercept a 57.7 51.3 38.5 Lom Slope coefficient b -23.8 -23.6 -22.1 R2 (%) 86 86 86 Intercept a 48.4 42.6 31.2 Sofia Slope coefficient b -21 -20.5 -18.8 R2 (%) 76 75 73 Intercept a 56.1 49.1 35.9 Silistra Slope coefficient b -20.5 -20.3 -19.3 R2 (%) 86 86 86 Intercept a 53.5 47.6 35.7 Pleven Slope coefficient b -23.3 -23 -21.6 R2 (%) 82 81 79 Intercept a 63.6 56.9 43 Varna Slope coefficient b -18.1 -17.6 -16.5 R2 (%) 82 81 80 Droughts and Climate Change in Bulgaria 49 gression coefficient (b) and the intercept (a) when relationships ness (Figure 13). Results show that their values change little are computed for low, medium and high TAW soils. among the regions and, for each location, among soil groups.

The determination coefficients are generally high, which Differently, the intercept (value of RYD when SPI-2Jul-Aug = means that a large fraction of the RYD variation is explained 0) depends upon the soil group: the intercept decreases from by the SPI-2Jul-Aug, i.e., by the dryness conditions during the low to high TAW since dryness has larger influence on RYD maize peak demand period. Changes of R2 for different soils when TAW is small. are negligible. These results confirm the preceding analysis When monitoring precipitation, the adoption of SPI-2July-Aug and the possibility of using SPI-2Jul-Aug as an indicator of water may be useful to manage the risk of economic losses with rain- deficit not depending on the soil type. The regression coef- fed maize by advising farmers to adopt supplemental irrigation ficients may be assumed as indicators of the linear yield de- if water is available at farm (Figures 13c and 13d). The related crease of RYD when SPI-2Jul-Aug decreases from its maximum farm advising may be regionally oriented and take into consid- (wet conditions) to low (and negative) values referring to dry- eration the dominant soil type (as for the RYD and NIR further

y = -23.6 x + 47.2 2 y = -93.8 x + 189.9 R = 0.91 2 ) R = 0.89 % ( 67 e Threshold 2012 ) 270 Threshold 2012 i z m a 60 Threshold 1995-2005 m m

Threshold 1995-2005 ( 235

d e e f i z a i m m a

D r N I R s R Y

0 0 -2.00 -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50 -2.00 -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50 a) c) SPI ( 2) for "July - Aug" SPI ( 2) for "July - Aug"

RYD/NIRs threshold relative to 2012

y = -21.6 x + 35.7 y = -84.3 x + 148.4 R2 = 0.79 R2 = 0.75 Threshold 2012 ) 73 ) %

( Threshold 1995-2005 295

67 m Threshold 2012 e

m 270 i z (

a Threshold 1995-2005 e m i z

a d e f m

i n a N I R s D r R Y

0 0 -2.50 -2.00 -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50 -2.50 -2.00 -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50 b) SPI ( 2) for "July - Aug" d) SPI ( 2) for "July - Aug"

RYD/NIRs threshold relative to 2012

Fig. 13. Relationships between seasonal SPI2 “July-Aug” and relative yield decrease of rainfed maize RYD or Net irrigation requirements NIR for: a) and c) Plovdiv and b) and d) Pleven; soils of large TAW (180 mm m-1), late maize hybrids 50 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins mapping in Figures 16 and 17). It needs to reflect in addition high TAW negative economic impacts occur only in severe- the actual economic farming conditions using the relationships ly/extremely dry peak demand periods when SPI-2July-Aug < in Fig.9b, as it is in the case of risen grain price (from 200 to -1.5 while for Sandanski such impacts occur for SPI-2July-Aug < 400 BGN t-1) and production expenses (from 800 to 1200 BGN +0.10. This also indicates that the region of Sandanski is ex- ha-1) during the extremely dry 2012 (Figures 14a and 14b). tremely vulnerable to water deficits in rainfed maize systems

The SPI-2July-Aug threshold, when computed for the values or, in other words, that rainfed maize is not viable there. This of RYD that do not produce economic losses, may be used as is due to the predominant climate, of Mediterranean type, indicators of dryness that affects yields. These values are rep- where rainfall in summer is low, much less than in all other resented in Figure 14 for all locations and soil groups includ- regions of Bulgaria. ing, when the change in economic condition was considered Results show that in the Thrace Plain (Plovdiv and Stara (Figure 14b). It can be observed that for Pleven and soils with Zagora), for soils with low or medium TAW, rainfed maize

1.5

0.5 g u A -

l 0 u J 2 -

I -0.5 P S -1

-1.5

-2 Lom Pleven Silistra Varna Sofia Sandanski Plovdiv Stara Zagora Northern Locations Southern Locations

1.5

0.5 g u A - l

u 0 J 2 -

I -0.5 P S -1

-1.5

-2 Lom Pleven Silistra Varna Sofia Sandanski Plovdiv Stara Zagora Northern Locations Southern Locations

116 mm m-1 136-157 mm m-1 173-180 mm m-1

Fig. 14. Threshold values of SPI2 “July-Aug” indicative of economic risk for rainfed maize in various regions and soil groups having small (116 mm m-1), medium (136-157 mm m-1) and large (173-180 mm m-1) TAW relative to the economical conditions of: a) 1990-2005 and b) the very dry 2012 Droughts and Climate Change in Bulgaria 51

is vulnerable to dryness even when SPI-2July-Aug are not nega- lower than -0.50. In North Bulgaria conditions are more fa- tive, which indicates a high vulnerability to water stress in vourable but the risk of economic losses is high for soils with that area (Figure 14a). However, that vulnerability is lower low TAW mainly along the Black Sea coast (Varna) and in for soils with high TAW when SPI-2July-Aug thresholds are not the region of Lom.

a) b) c)

Fig. 15. Spatial distribution of seasonal SPI2 “July-Aug” relative to the year of: a) extreme (2000), b) average (1970) and c) moderate (1981) irrigation demand, Bulgaria

a) b) c) Fig. 16. Spatial distribution of relative yield decrease (RYD, %) for rainfed maize relative to the year of: a) extreme (2000), b) average (1970) and c) moderate (1981) irrigation demand, Bulgaria

Fig. 17. Spatial distribution of net irrigation requirements (NIR, mm) for maize relative to the year of: a) extreme (2000), b) medium (1970) and c) moderate (1981) irrigation demand, Bulgaria 52 Z. Popova, M. Ivanova, L. Pereira , V. Alexandrov, M. Kercheva, K. Doneva and D. Martins

Mapping RYD of rainfed maize and creases from maximum values (wet conditions) to low nega- irrigation requirements to cope with drought tive ones, again not depending upon the soil type. Differently,

The results of the study are used for mapping the hazards the intercept (value of RYD when SPI-2July–Aug = 0) reflects the of drought and identification of drought prone territories in soil group and decreases from low to high soil TAW.

Bulgaria. Maps of the SPI2 “July-Aug”, illustrate the dryness con- Moreover, it was possible to define the SPI thresholds for ditions during the maize Peak demands period relative to the each region and soil TAW relative to the peak period of July- very dry (2000), the average demand (1970) and the moderate- August, SPI-2July-Aug when decreased yields lead to economic ly dry (1981) year of presented in Figures 15a, 15b and 15c. losses. Results show that vulnerability to water stress de- The latter maps, The map of Soil Geographical Regions crease when soils have a higher soil water holding capacity, (Koinov et al., 1998) and the derived relationships between and that vulnerability is higher in southern regions than in simulated RYD (%) and the indicator of water deficit for northern ones. SPI-2July-Aug value may support advising farm- rainfed maize SPI2 “July-Aug” (Table 5) are used to predict the ers about the risk of economic losses and to adopt then sup- distribution of hazardous yield losses for maize in Bulgaria in plemental irrigation. the same years (Figures 16a, 16b and 16c). The derived reliable relationships and specific thresholds

The elaborated maps correspond to soils of medium TAW of seasonal SPI2 “July-Aug”, under which soil moisture deficit (136-157 mm m-1) that are widespread in a different degree leads to severe impact of drought on rainfed maize yield for over the main geographical regions of Bulgaria. Thus they the main climate regions and soil groups in Bulgaria, are rep- express the hazard of drought and further the irrigation man- resentative of a wider area of Moderate Continental, Transi- agement to support farmers (Figures 17a, 17b and 17c). tional Continental, Transitional Mediterranean and Black Sea climate in SEE. They are used for elaboration of RY decreas- Conclusion es and irrigation requirements’ maps, as well as for identification of drought prone territories and coping with drought at This study, relative to rainfed maize crop, was applied to regional and national level. eight Bulgarian climate regions and three soil groups, and the Further studies are desirable in terms of analyzing the period 1951-2004. constraints of irrigation as an adaptation measure to cope Relative to climate change, significant negative trends for with droughts and climate change. precipitation during Peak Demand Season “June-Aug” were identified in The Thrace region, which are combined with the Acknowledgements respective positive trends for ETo June-Aug. Authors gratefully acknowledge the financial support of An analysis relative to the present climate period 1970- Drought Management Centre for South East Europe Project, 2004 shows a trend for aggravation of drought not only in South East Europe Transnational Cooperation Programme The Thrace but also in the northern locations and Sofia field, co-funded by the European Union. thus confirming that agricultural lands in Bulgaria experi- ence an increased vulnerability to water stress. Rainfed maize is associated with great yield variability References (29% < Cv < 72%). Allen, R. G., L. S. Pereira, D. Raes and M. Smith, 1998. Crop It was possible to define probability curves relative to the Evapotranspiration – Guidelines for Computing Crop Water Re- RY decreases and to the required irrigation NIR to avoid quirements. FAO, Rome, Irrigation and Drainage Paper, 56. these yield deficits. Therefore, based on economical consid- Boneva, K., 2012. Study on soil characteristics related to calibra- erations, thresholds defining the risk associated with water tion and use of “crop-water-yield” simulation models. In: Z. deficits could be computed. Popova (Ed.) Risk assessment of drought in agriculture and Linear relationships were found relating RYD with the irrigation management through simulation models, “N. Poush- karov” Institute of Soil Science, Sofia, pp. 141-165 (Bg). SPI-2July–Aug. The determination coefficients were generally high, thus indicating that a large fraction of the RYD varia- Bordi, I., K. Fraedrich and A. Sutera, 2009. Observed drought and wetness trends in Europe: an update. Hydrol. Earth Syst. tion is explained by the SPI-2 , i.e. by the dryness condi- July–Aug Sci., 13: 1519-1530. tions during the maize peak demand period. These results al- Çakir, R., 2004. Effect of water stress at different development low considering the SPI-2July–Aug as an indicator of water defi- stages on vegetative and reproductive growth of corn. Field cit for rainfed maize, which is not depending of the soil type. Crops Research, 89: 1-16. The regression coefficients may be assumed as indicators of Doorenbos, J. and A. H. Kassam, 1979. Yield Response to Water. the linear yield decrease of RYD when SPI-2July–Aug also de- FAO, Rome, Irrigation and Drainage Paper, 33. Droughts and Climate Change in Bulgaria 53

Doorenbos, J. and W. O. Pruitt, 1977. Crop Water Requirements. Slovenian Environmental Agency, Ljubljana. http://www.met. FAO, Rome, Irrigation and Drainage Paper, 24. hu/doc/DMCSEE/DMCSEE_final_publication.pdf Eneva, S., 1997. Irrigation and irrigation effect on field crops. Popova, Z., and M. Kercheva, 2005. Ceres model application for Problems of crop production science and practice in Bulgaria. increasing preparedness to climate variability in agricultural plan- Agricultural University of Plovdiv, (Bg). ning - Risk Analyses. Phys Chem Earth, Parts A/B/C, 30: 117-124. Gregoric, G. (Ed.), 2012. Drought Management Centre for South- Popova, Z. and L. S. Pereira, 2008. Irrigation scheduling for fur- East Europe – DMCSEE. Summary of Project Results, Slove- row irrigated maize under climate uncertainties in the Thrace nian Environmental Agency, Ljubljana. http://www.met.hu/doc/ plain, Bulgaria. Biosyst Eng., 99: 587-597. DMCSEE/DMCSEE_final_publication.pdf Popova, Z. and L. S. Pereira, 2011. Modelling for maize irriga- Guttman, N. B., 1998. Comparing the Palmer drought index and tion scheduling using long term experimental data from Plovdiv the standardised precipitation index. J. Am. Water Res. Assoc., region, Bulgaria, Agric. Water Manage., 98: 675-683. 34: 113-121. Popova, Z., S. Eneva and L. S. Pereira, 2006b. Model validation, Ivanova, M. and Z. Popova, 2011. Model validation and crop co- crop coefficients and yield response factors for irrigation schedul- efficients for maize irrigation scheduling based on field experi- ing based on long-term experiments. Biosyst. Eng., 95: 139-149. ments in Sofia. In: 100 Years Soil Science in Bulgaria, Trans. Popova, Z., M. Kercheva and L. S. Pereira, 2006a. Validation of Nat. Conf., Sofia, pp. 542-548 (Bg). the FAO methodology for computing ETo with limited data. Ap- Koinov, V., I. Kabakchiev and K. Boneva, 1998. Atlas of the Soils plication to South Bulgaria. Irrig. Drain., 55: 201-215. in Bulgaria. Zemizdat, Sofia. Popova, Z., M. Ivanova, D. Martins, L. S. Pereira, K. Doneva, Lana, X., C. Serra and A. Burgueño, 2001. Patterns of monthly rain- V. Alexandrov and M. Kercheva, 2013a. Drought and climate fall shortage and excess in terms of the standardized precipitation variability impacts on Bulgarian agriculture: the case of rainfed index for Catalonia (NE Spain). Int. J. Climatol., 21: 1669-1691. maize. In: N. Lamaddalena, M. Todorovic, L. S. Pereira (Eds.) McKee, T. B., N. J. Doesken and J. Kleist, 1993. The relationship Water, environment and agriculture: challenges for sustainable of drought frequency and duration to time scales. Proc. 8th Conf. development (Proceedings of 1st CIGR Inter-Regional Confer- on Applied Climatology, American Meteorological Society, ence on Land and Water Challenges, Bari, Italy, 2013), FULL Boston, pp. 179-184. PAPERS (USB) and Book of abstracts, pp. 191-193. McKee, T. B., N. J. Doesken and J. Kleist, 1995. Drought monitor- Popova, Z., M. Ivanova, D. Martins, L. S. Pereira, M. Kerche- ing with multiple time scales. Proc. 9th Conf. on Applied Clima- va, V. Alexandrov and K. Doneva, 2013b. Droughts and cli- tology, American Meteorological Society, Boston, pp. 233-236. mate change in Bulgaria: assessing trends and related impacts Mishra, A. K. and V. P. Singh, 2010. A review of drought con- in maize cropping system. Agric. Science, 46 (1): 19-30 (Bg). cepts. J. Hydrol., 391: 202-216. Popova, Z., M. Ivanova, L. S. Pereira, V. Alexandrov, K. Do- NDMC, National Drought Mitigation Center, 2013. Drought Ba- neva, P. Alexandrova and M. Kercheva, 2012. Assessing sics. NDMC. http://drought.unl.edu/DroughtBasics/Whatis- drought vulnerability of Bulgarian agriculture through model Drought.aspx (Accessed 05 Nov. 2013) simulations. J. Env. Sci. Eng., B 1: 1017-1036. Olesen, J. E., M. Trnka, K. C. Kersebaum, A. O. Skjelvåg, B. Rafailov, R., 1995. Annual reports of ISS “N. Poushkarov”, Sofia Seguin, P. Peltonen-Sainio, F. Rossi, J. Kozyra and F. Mi- (Bg). cale, 2011. Impacts and adaptation of European crop production Rafailov, R., 1998. Annual reports of ISS “N. Poushkarov”, Sofia systems to climate change. Europ. J. Agron., 34: 96-112. (Bg). Pereira, L. S., I. Cordery and I. Iacovides, 2009. Coping with Wa- Statistical Yearbooks, 1900-2011. National statistical institute, ter Scarcity. Addressing the Challenges. Springer, Dordrech. Sofia, Bulgaria (Bg). Pereira, L. S., P. R. Teodoro, P. N. Rodrigues and J. L. Teixeira, Stewart, J. L., R. J. Hanks, R. E. Danielson, E. B. Jackson, W. 2003. Irrigation scheduling simulation: the model ISAREG. In: O. Pruitt, W. T. Franklin, J. P. Riley and R. M. Hagan, 1977. G. Rossi, A. Cancelliere, L. S. Pereira, T. Oweis, M. Shatanawi, Optimizing Crop Production through Control of Water and Sa- A. Zairi (Eds.) Tools for Drought Mitigation in Mediterranean linity Levels in the Soil. Utah Water Research Laboratory Re- Regions, Kluwer, Dordrecht, pp. 161-180. port PRWG151-1. Utah State University, Logan. Popova, Z., 2008. Optimization of irrigation scheduling, yields Stoyanov, P., 2008. Agroecological potential of maize cultivated and their environmental impacts by crop models. Avtoreferat on typical soils in the conditions of Bulgaria. Habilitation thesis, of DSc Thesis with extended summary in English, “N. Poush- “N. Poushkarov” Institute of Soil Science, Sofia, (Bg). karov” Institute of Soil Science, Sofia, pp. 100. Varlev, I., 2008. Potential, efficiency and risk under maize cultiva- Popova, Z. (Ed.), 2012. Risk assessment of drought in agriculture tion in Bulgaria. Agricultural Academy, Sofia, 128 pp. (Bg). and irrigation management through simulation models. “N. Varlev, I. and Z. Popova, 1999. Water-Evapotranspiration-Yield. Poushkarov” Institute of Soil Science, Sofia, 242 pp. (Bg). Irrig and Drain Inst, Sofia, 144 pp. (Bg). Popova, Z., 2012. Drought vulnerability estimates based on crop- Varlev, I., N. Kolev and I. Kirkova, 1994. Yield-water relationships yield models. In: G. Gregoric (Ed.) Drought management centre and their changes during individual climatic years. In: Proceed- for South-East Europe-DMCSEE. Summary of project results, ings of 17th Europ. Reg. Conf. of ICID, Varna, pp. 351-360.

Received August, 2, 2014; accepted for printing December, 2, 2014.