Atmospheric circulation types and winter daily precipitation in Iran

Authors


Correspondence to: T. Raziei, Soil Conservation and Watershed Management Research Institute (SCWMRI), P. O. Box 13445-1136, Tehran, Iran. E-mail: tayebrazi@scwmri.ac.ir

ABSTRACT

The relationship between daily large-scale atmospheric circulation types (CTs) and wintertime daily precipitation over Iran during the period 1965–2000 is investigated. Twelve atmospheric CTs identified in a previous study, which applied the K-means clustering technique to the rotated principal components (RPCs) of the 500 hPa geopotential height fields, are also considered in this study. The leading spatial modes of daily precipitation variability over Iran are extracted by a PC analysis, with Varimax rotation, applied to the APHRODITE (Asian Precipitation-Highly Resolved Observational Data Integration Towards Evaluation of the Water Resources) gridded precipitation dataset; six Iranian sub-regions with independent precipitation variability are identified. The relationships between the CTs and the daily precipitation are investigated by computing: (1) the spatial patterns of the performance index (PI) for each CT and (2) the cross tabulations between the frequencies of occurrence of the CTs and the RPC scores of the daily precipitation, associated with each of the six sub-regions. Results suggest that two particular CTs affect the precipitation occurrence over most of the country, while the remaining ten provide more regional or negligible contributions to precipitation. The more (less) influencing CTs in each precipitation sub-region are then identified and a characterization of the main large-scale atmospheric features governing the winter precipitation fields is provided.

1. Introduction

Space and time variability analysis of regional precipitation is crucial for water resource management, at a regional level, and for numerous environmental and socio-economic systems, particularly for the agricultural sector. The variability of the precipitation field depends on many factors, including the thermodynamic structure of the atmosphere, the orography and the interaction with the large-scale atmospheric circulation. Several studies have been carried out not only to identify large-scale atmospheric circulation types (CTs) leading to precipitation, but also to understand how their variability can affect the frequency and amount of precipitation in a given region (Corte-Real et al., 1998; Romero et al., 1999a; Wibig, 1999). Many authors have studied the possible linkages between the CTs and several surface climate variables, such as precipitation and temperature (Kilsby et al., 1998; Romero et al., 1999b; Chen and Hellström, 1999; Kidson, 2000; Xoplaki et al., 2000; Santos et al., 2005). As an example of regional assessment, Corte-Real et al. (1998) classified the main atmospheric patterns associated with precipitation in Portugal using the principal component analysis (PCA) and the cluster analysis (CA). Examples of more recent studies on the link between large-scale atmospheric circulation and precipitation are given by Santos et al. (2007), Casado et al. (2010), Polo et al. (2011) and Espinoza et al. (2012).

In Iran, precipitation variability is mainly controlled by the complex orography (Zagros and Alborz mountain chains in the west and north, respectively; see Figure 1(a)) that enhances (mitigates) large-scale precipitation on their windward (leeward) side. Further, the importance of the orography in controlling the spatial variability of precipitation over the country is still more pronounced under strong convective activity. Several authors have analysed the link between atmospheric indices and surface climate variables in Iran (Nazemosadat and Cordery, 2000; Nazemosadat and Ghasemi, 2004; Ghasemi and Khalili, 2006), and trends in extreme precipitation events (Zhang et al., 2005; Rahimzadeh et al., 2009). Alijani (2002) investigated the relationship between variations in the monthly mean 500 hPa flow patterns and precipitation/temperature over Iran during the winter months from 1961 to 1990. The S-mode PCA with Varimax rotation was used to reduce the dimensionality of the monthly geopotential height data of each winter month to a few significant factors. The factor scores for each month were then correlated with the monthly Z-scores of precipitation/temperature anomalies observed at several synoptic stations. Results highlighted the key role played by the troughs and ridges located near Iran on the atmospheric conditions of the country.

Figure 1.

(a) Topographic map of Iran and spatial pattern of: (b) total annual mean precipitation (mm), (c) total winter mean precipitation (mm) and (d) percentage of winter to the annual total precipitation over Iran (%).

Besides those efforts, only a few attempts have been devoted to the study of the relationship between large-scale CTs and daily precipitation, mainly due to the lack of long historical records of daily precipitation at the stations. Raziei et al. (2012a), using the APHRODITE (Asian Precipitation-Highly Resolved Observational Data Integration Towards Evaluation of the Water Resources) gridded precipitation dataset for the period 1961–2004, investigated the influence of large-scale atmospheric circulation on seasonal regimes of daily precipitation over Iran. The leading regional spatial modes of daily precipitation for each season (excluding summer) were computed applying the S-mode PCA with Varimax rotation. The dynamical features associated with each regional precipitation regime were then identified by composing only daily atmospheric fields (500 hPa geopotential height and relative vorticity) with rotated PC (RPC) scores ≥ 1.5 (strong positive phase) of the precipitation field, thus excluding days with small and local precipitation amounts. The results suggested that the spatial distribution of precipitation over Iran is largely governed by the geographical location of both the mid-tropospheric trough over the Middle East and the Arabian anticyclone. Moreover, Raziei et al. (2012b) have recently identified the daily atmospheric CTs in the Middle East and studied their relationship with the occurrence of meteorological dry/wet spells during winter in western Iran. Daily large-scale weather conditions during the period 1965–2000 have been classified into 12 CTs by applying the PCA to the 500 hPa geopotential height fields, coupled with the non-iterative K-means clustering technique. The linkage between the monthly frequencies of daily CTs and the frequencies of occurrence of dry/wet events (identified by applying the standardized precipitation index on a 1 month time scale to precipitation data recorded at 140 stations) was assessed by a correlation analysis. Stepwise multiple linear regressions were used to evaluate the predictive potential of the CTs. The results suggested that some CTs are skilful predictors of the winter dry/wet events in western Iran.

This study aims to complement the aforementioned outcomes (Raziei et al., 2012b) by analysing the relationship between the previously identified CTs and the winter daily precipitation in Iran. In fact, as meteorological dry/wet events are mainly controlled by daily precipitation, which is partially driven by large-scale dynamics, an analysis of the link between the CTs and daily precipitation over Iran is considered pertinent. Nevertheless, as the meteorological drought and wetness are defined with respect to climatological means, and in Raziei et al. (2012b) they were monitored on a monthly time scale, it should be expected that the daily CTs influencing dry/wet spells should differ from those affecting the daily precipitation events. Furthermore, the analysis is here extended to the whole of Iran and is based on the APHRODITE gridded precipitation dataset for the period 1965–2000. To investigate the relative contribution of the large-scale CTs to the precipitation events in different areas across Iran, a regionalization of winter daily precipitation is carried out. For this purpose, the PCA with Varimax rotation is applied to the daily precipitation fields. Additional analyses are based on: (1) the estimation of the relative contribution of each CT to the precipitation totals, through the calculation of the performance index (PI) and (2) the computation of contingency tables between the frequencies of occurrence of the CTs and the RPC scores associated with each precipitation sub-region.

Section '2. Data and methods' provides a description of data and methods used for the analysis. In Section '3. Results', the main results are presented and discussed, with particular emphasis given to the spatial modes of winter daily precipitation (Section '3.1. Spatial modes of winter daily precipitation') and the relationship between the CTs and daily precipitation (Section '3.2. Relationships between atmospheric CTs and daily precipitation'). Last, the main conclusions and an outlook for future investigations are provided in Section '4. Conclusions'.

2. Data and methods

2.1. Data

In this study, the winters (here defined as December–March, DJFM) in the period 1965–2000 are considered. The DJFM period exhibits the largest atmospheric variability over the extra-tropical Northern Hemisphere, and largely represents the rainy season in Iran. The observational daily precipitation dataset is provided by the APHRODITE project for the Middle East (Yatagai et al., 2008, 2009). Its state-of-the-art-gridded daily precipitation datasets, with high-resolution grids over Asia, are based on interpolations of data collected from the rain-gauge observation networks over the region, using the Shepard's (1968) algorithm. The ratio between the daily precipitation and its corresponding climate-mean value is interpolated at a 0.05° grid resolution. Each gridded ratio is then multiplied by the respective gridded climate-mean value on a daily basis. Subsequently, the 0.05° analysis is re-gridded to 0.25 or 0.5° grids. The data release for the Middle East APHRO_V0902, covering the domain (15°–65°E, 25°–45°N), with a spatial resolution of 0.5° latitude × 0.5° longitude, is used here. Only grid points over Iran (638) were extracted for the current analysis. More specifically, the dataset for Iran is based on 337 meteorological stations, provided by the Islamic Republic of Iran Meteorological Organization (IRIMO) to the APHRODITE project, with each having at least 5 years of available data. These Iranian stations are composed of 154 World Meteorological Organization (WMO) stations and 183 non-WMO stations, spread throughout the country and covering different time sections, with the longest time period starting in 1961 up to 2004 (44 years). Further details can be found in Yatagai et al. (2008, 2009).

The already identified 12 CTs (Raziei et al., 2012b) are used here. The CTs were computed using daily means of the 500 hPa geopotential height for DJFM, covering Iran and the Middle East (10°–80°E, 20°–60°N), with a 2.5° latitude × 2.5° longitude spatial resolution. These data were retrieved from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis archive (Kalnay et al., 1996; Kistler et al., 2001). Raziei et al. (2012a) found that the Arabian anticyclone largely affects the moisture transports over Iran. Thus, the domain selected to represent the CTs in this study extended 10° of latitude further south compared to that in Raziei et al. (2012b); this larger domain indeed allows a more effective detection of the Arabian anticyclone in each CT. Furthermore, the composite maps of the NCEP/NCAR 500 hPa relative vorticity and 850 hPa moisture transports are also calculated so as to identify the main large-scale atmospheric features associated with the CTs. The latter fields are intentionally shown on a spatial domain narrower than for the CTs (i.e. 20°–75°E, 10°–50°N) to emphasize the moisture transports over the study area. Last, in order to assess the CTs that contribute more (less) to the precipitation occurrence over Iran, the relative vorticity and omega-vertical velocity fields, within the longitudinal section (18°–78°E) and averaged over the latitude band (25°–40°N), are discussed.

2.2. Spatial modes of daily precipitation and CTs

The PCA is a variable-reduction statistical procedure (Richman, 1986) that has been extensively used to identify spatial and temporal modes of precipitation variability in different regions (Fernandez Mills et al., 1994; Romero et al., 1999a; Serrano et al., 1999; Quadrelli et al., 2001). The Varimax rotation, on the other hand, allows the identification of regions featuring different temporal variability within the area of interest (Rencher, 1998).

Daily precipitation data for DJFM is used as an input for the S-mode (multiple stations over time) PCA to extract the leading spatial modes of precipitation variability. As the assumptions in PCA require normally distributed variables (Fovell and Fovell, 1993), before performing the PCA of precipitation, the square-root transformation of daily precipitation data is applied to reduce the positive skewness of the daily totals (Lanzante and Harnack, 1982; Richman and Lamb, 1985; Romero et al., 1999a; Neal and Phillips, 2009). Following North's rule-of-thumb (North et al., 1982), the sampling errors of the eigenvalues of each mode are estimated to identify the ‘well-defined’ (non-degenerate) modes to be retained for subsequent rotation. As only the first six modes are ‘well-defined’ at a 95% confidence level, the Varimax rotation is only applied to these six modes. The resulting RPC scores represent the temporal variability of the winter daily precipitation in the sub-regions depicting the highest loadings.

As mentioned above, 12 CTs from Raziei et al. (2012b) are used in the current analysis. In that study, the CTs have been classified applying the PCA with Varimax rotation to the 500 hPa geopotential height fields, coupled with the non-iterative K-means clustering technique. In this approach, extreme RPCs (in their positive and negative phases) were considered as potential groups and their centroids were computed by averaging all days that fulfilled the extreme score criterion (Esteban et al., 2005, 2006). According to that method, CT+ and CT− denote the CTs deriving from the positive and negative phases of the RPCs, respectively.

As discussed in Raziei et al. (2012b), there are several techniques proposed in the literature for the CT classification (Yarnal, 2001; Huth et al., 2008; Philipp et al., 2010). The method followed here is based on the S-mode PCA, acting as a data pre-processing tool prior to CA. Recently, the method, among others, was evaluated by the COST Action 733 ‘Harmonisation and Applications of Weather Type Classifications for European regions’, which was classified as a PCA-based method (Philipp et al., 2010). This technique is also suitable for the purpose of this study and is in conformity with Raziei et al. (2012b). The S-mode PCA with Varimax rotation is the basic approach for both the CT identification and the precipitation regionalization. This warrants consistency between the methodologies used for analysing the two atmospheric fields (500 hPa geopotential height and precipitation). Contrary to Alijani (2002), where factor scores indicating significant correlations with precipitation were considered in the composite maps of the 500 hPa geopotential height, the non-iterative K-means clustering is used here for the compositing, following the extreme score approach, successfully tested by Esteban et al. (2005, 2006) and Philipp et al. (2010).

2.3. Relationships between CTs and daily precipitation

To establish the contribution of the large-scale circulation to the precipitation totals in Iran, the spatial patterns of the PI associated with CTs are considered. The index, introduced by Zhang et al. (1997), compares the mean daily precipitation for the i-th CT with the climatological daily mean precipitation in the following manner:

equation image(1)

where ni is the number of days of type i, Ri is the respective total amount of precipitation and R is the total precipitation received in the entire period of n days. Thus, the index considers the probability of precipitation occurrence conditional to the presence of a given CT and takes the intensity of precipitation for that CT into account. A PI value lower than 1 means that the corresponding CT does not considerably contribute to precipitation, while the contrary holds for a PI greater than 1. A PI value of approximately 1 means that a given CT leads to precipitation amounts near the climatological mean. As the index is normalized with respect to the daily climatology, the basic requirement for the index application is the availability of long time series of daily precipitation (at least 30 years of data records for locations with substantial time fluctuations) at a given location for climate-mean computations. Further, according to the index definition, a given CT is considered favourable to precipitation only if it yields more precipitation than the climatological mean. Although this CT characterization meets the purpose of the study (i.e. isolation of the CTs that most contribute to precipitation in Iran), it does not allow the identification of the CTs that can generate either precipitation extremes or long-lasting precipitation events with contributions below the climate-mean.

The relationship between the CTs and daily precipitation over Iran is further assessed using contingency tables (cross tabulations). These tables are joint frequency distributions of two or more categorical variables, being widely used in the analysis of the relationship between discrete variables (Von Storch and Zwiers, 1999; Wilks, 2006). The main purpose of contingency table analysis is to determine whether the row classification (first variable) is independent of the column classification (second variable). The null hypothesis (H0) states that the two variables are not related, i.e. the distribution of data among the categories of the first variable is not affected by the classification of the second variable (Helsel and Hirsch, 2002).

In this study, the occurrence frequencies of the different CTs are cross-tabulated with the RPCs of daily precipitation, under the assumption that both variables are independent. For this purpose, the scores of each RPC are categorized into ‘rainy/no-rainy’ classes, defined as values above/below zero, respectively. To test the null hypothesis, the Pearson chi-squared test of independence is computed by directly comparing the observed frequencies with the expected frequencies. The test provides information about the probability of association between the two variables (CT and precipitation RPC). In addition, the standardized residuals of these tables (positive/negative standardized differences between observed and expected frequencies) determine the direction of this association (Stahl and Demuth, 1999). The sign of the residuals indicates whether a CT is a precipitation-generating pattern or a mostly dry pattern (‘rainy’ or ‘no-rainy’ CT, hereafter). A positive residual ≥ 2 means that the observed frequency is greater than the expected frequency of precipitation occurrence (i.e. the corresponding table cell cannot be suitably fitted by the model of independence; Helsel and Hirsch, 2002) and the CT is classified as a ‘rainy’ CT. The same holds for a negative residual ≤− 2, which now identifies a ‘no-rainy’ CT. Six contingency tables are thereby constructed in order to determine the link between the occurrence of the 12 CTs and precipitation in the six sub-regions. For the sake of succinctness, the results of the six contingency tables concerning the standardized residuals are compactly presented in a single pivot table.

3. Results

The spatial patterns of the annual and winter mean precipitation fields are discussed before defining the sub-regions of winter precipitation variability. A description of the relationships between each CT and regional daily precipitation is presented in the subsequent sections.

3.1. Spatial modes of winter daily precipitation

Figure 1(b) and (c) show the spatial patterns of the total annual and winter (DJFM) precipitation in Iran, respectively, averaged over the studied period. The map of the total annual mean precipitation (Figure 1(b)) mainly reflects the influences of both orography and regional seas. The highest values are observed in the Caspian Sea region and mountainous areas of northern and western Iran, while the lowest values are recorded in central, eastern and southern Iran. The highest precipitation in the Caspian Sea region is due to the proximity of the sea to the high Alborz Mountains that enhances the land–sea interaction (condensation barrier effect), causing high and regularly distributed precipitation over the region all through the year. Over the mountainous areas of western Iran there is a second maximum of annual precipitation, again related to the effect of orography in enhancing the precipitation processes, particularly windward of the Zagros Mountain. On the lee side of the Zagros and Alborz mountain chains, there is a vast area (central-eastern Iran) where precipitation amounts are quite low (arid or semi-arid region).

The spatial pattern of winter precipitation in Figure 1(c) resembles that of annual precipitation, but here the high amount of precipitation over the Zagros mountain chain in western Iran is very pronounced. The percentage of winter precipitation (Figure 1(d)) is lower than 40% of the total annual precipitation in the north, increasing southwards, where it reaches about 90% in the coastal areas of the Persian Gulf. Most parts of the country receive 50–90% of total annual precipitation in winter, suggesting that winter is the rainiest period in most of the country.

By applying the PCA with Varimax rotation to daily precipitation fields, grid points characterized by similar temporal variability are grouped into a single sub-region, denoted by high positive values in the rotated loading (R-Loading). Since, by definition, the rotated loadings are correlations between daily precipitation series and the corresponding RPC scores, a threshold of 0.6 on the rotated loadings seems to be suitable for isolating sub-regions with similar temporal variability (orange and red areas in the maps). Results show that Iran is divided into six sub-regions of independent precipitation variability. The six R-Loadings, which cumulatively explain 61.5% of the total variance of daily precipitation fields (Table 1), are depicted in Figure 2; loadings represent the correlation coefficients between the square-rooted precipitation anomaly and the corresponding RPCs.

Figure 2.

Leading six RPC modes (R-Loadings 1–6) of daily winter (DJFM) precipitation in Iran.

Table 1. Explained variances (EV) and cumulative explained variances (CEV) of the six Varimax RPCs of winter (DJFM) daily precipitation in Iran
EV (%)CEV (%)Number of RPC
11.711.71
11.423.12
10.433.53
10.043.54
9.452.95
8.761.56

The first mode (R-Loading 1) accounts for 11.7% of the total variance and isolates western-central Iran. R-Loading 2 represents 11.4% of the total variance and depicts relatively high positive values over central-eastern Iran. R-Loading 3, explaining 10.4% of the total variance, identifies a sub-region in north-eastern Iran, while R-Loading 4 (10.0% of the total variance) outlines north and north-western Iran as another distinct sub-region. R-Loading 5, with 9.4% of the total variance, identifies a sub-region in southern Iran. Finally, R-Loading 6 explains 8.7% of the total variance and characterizes the precipitation variability in south-eastern Iran. The corresponding RPCs represent the time variability of winter daily precipitation in each sub-region (not shown) and are used in the contingency analysis.

3.2. Relationships between atmospheric CTs and daily precipitation

Figure 3 displays the composites of the geopotential and relative vorticity at 500 hPa for the 12 CTs. The relative frequency of occurrence (%), the mean and maximum lifetimes (d) of each CT are listed in Table 2. Composite maps of moisture transports at 850 hPa for each CT are shown in Figure 4. Finally, Figure 5 shows the spatial patterns of PI for each CT over Iran, while Table 3 contains the standardized residuals obtained from the contingency tables between the frequency of occurrence of the CTs and the precipitation in the six identified sub-regions (the percentage of rainy days in each CT and sub-region is also listed within parenthesis).

Figure 3.

Composite maps of the 500 hPa geopotential height (black contours every 30 gpm) and relative vorticity (dashed lines with blue shading for negative vorticity, solid lines with red shading for positive vorticity; contours every 0.5 × 10−5 s−1) for the 12 CTs identified in Raziei et al. (2012b).

Figure 4.

Composite maps of moisture transports (shading, contours every 4 g s−1) and corresponding streamlines at 850 hPa (solid contours) for the 12 CTs in Figure 3.

Figure 5.

Spatial patterns of PI associated with the 12 CTs in Figure 3. Units are dimensionless.

Table 2. Relative frequency of occurrence (%), mean and maximum lifetimes (d) of CTs obtained from the NCEP/NCAR reanalysis
 CT1−CT1+CT2−CT2+CT3−CT3+CT4−CT4+CT5−CT5+CT6−CT6+
Occurrence frequency (%)8.57.37.58.87.79.67.68.59.68.98.57.3
Mean lifetime (d)4.63.83.84.23.73.62.73.13.22.92.82.4
Maximum lifetime (d)24241114121510119151112
Table 3. Standardized residuals from the contingency tables between the frequencies of CTs and the precipitation occurrences in the six sub-regions of Iran
CTCentral-western (RPC1)Central-eastern (RPC2)North-eastern (RPC3)North-western (RPC4)South-western (RPC5)South eastern (RPC6)
  1. Values within parenthesis are the percentages of ‘rainy’ days in each CT and sub-region. The ‘rainy’ CTs for each region are in bold. All residuals are statistically significant at the 0.1% level.

CT1−0.9 (36.7%)4.0 (43.1%)− 0.1 (35.8%)− 2.9 (29.1%)0.9 (31.5%)4.1 (40.2%)
CT1+1.3 (38.1%)− 3.0 (22.2%)− 3.4 (24.7%)1.5 (43.8%)− 4.3 (16.3%)− 6.4 (9.7%)
CT2−− 0.0 (33.8%)− 0.3 (30.5%)− 1.7 (30.5%)− 0.5 (36.6%)0.3 (30.2%)− 2.3 (21.8%)
CT2+− 1.9 (28.2%)0.1 (31.9%)4.0 (48.6%)0.6 (40.5%)2.0 (34.7%)1.4 (32.6%)
CT3−− 1.6 (28.8%)− 2.2 (24.6%)− 3.9 (23.4%)− 5.3 (20.5%)− 1.8 (23.7%)− 0.2 (28.2%)
CT3+− 2.7 (26.2%)3.6 (41.4%)7.0 (56.9%)3.2 (48.1%)1.1 (31.9%)0.7 (30.7%)
CT4−− 4.2 (20.3%)− 4.3 (18.2%)− 6.6 (14.5%)− 5.0 (21.5%)− 4.7 (15.2%)− 2.7 (20.9%)
CT4+2.0 (39.8%)1.7 (36.3%)6.6 (56.9%)9.6 (69.4%)8.0 (51.5%)3.4 (38.2%)
CT5-− 6.2 (16.2%)1.0 (34.3%)2.3 (43.1%)− 4.4 (25.2%)− 5.5 (14.5%)0.7 (30.7%)
CT5+9.9 (63.0%)− 1.2 (28.0%)− 3.5 (25.7%)5.4 (55.3%)4.1 (40.4%)− 0.1 (28.5%)
CT6-− 0.3 (33.1%)− 0.2 (30.9%)− 0.8 (33.9%)− 1.8 (32.5%)− 3.0 (20.7%)− 0.8 (26.6%)
CT6+3.4 (44.8%)− 0.0 (31.3%)− 1.5 (31.3%)− 1.0 (35.1%)2.7 (37.3%)1.3 (32.6%)

CT1− features a weak trough (positive vorticity) over Iran/Pakistan/Afghanistan, with prevailing mid-tropospheric north-westerly flow over Iran that generally produces stable and cold/dry weather during winter in the country (Figure 3(a)). Its frequency of occurrence is 8.5%, with mean and maximum lifetimes of 4.6 and 24 d, respectively, making it one of the most persistent (long lasting) CTs over Iran (Table 2). The PI pattern associated with CT1− (Figure 5(a)) shows that it does not contribute to above average precipitation amounts over northern and western Iran. However, PI values greater than 1 characterize central-eastern and southern regions, suggesting that CT1− significantly contributes to precipitation in that sector of the country. This is supported by Figure 3(a), showing a core of maximum relative vorticity over Pakistan/Afghanistan, accompanied by low-tropospheric moisture transports blowing from the Persian Gulf (Figure 4(a)). In addition, the standardized residuals computed from the contingency tables for the six precipitation sub-regions corroborate the following finding (first row of Table 3): CT1− is a ‘rainy’ CT for the central-eastern and south-eastern sub-regions (residuals greater than 2 and 40% of rainy days), while it is a ‘no-rainy’ CT for the north-western sub-region (residual lower than − 2 and 29.1% of rainy days).

CT1+ is characterized by a cyclonic curvature over the eastern Mediterranean (positive vorticity) and by an anticyclonic curvature (negative vorticity) over central-southern Iran and southern Arabian Peninsula, causing mid-tropospheric southwesterly winds that can favour precipitation occurrences over Iran (Figure 3(b)). The frequency of occurrence of this CT is 7.3%, with mean and maximum lifetimes of 3.8 and 24 d, respectively, making it one of the most persistent CTs (Table 2). The PI pattern for CT1+ (Figure 5(b)) only shows important contributions to precipitation occurrences over north-western Iran, leaving most of the country without precipitation. This outcome is in agreement with previous findings using weather station data (Raziei et al., 2012b). This is caused by the presence of a cyclonic circulation over the eastern Mediterranean (Figure 3(b)) and by a low-level anticyclonic circulation over eastern Saudi Arabia, leading to moisture transports from southern Red Sea/Arabian Sea to northwestern Iran (Figure 4(b)). Results obtained for the standardized residuals in Table 3 (second row) suggest that CT1+ is a ‘no-rainy’ CT for almost all sub-regions (negative residuals lower than − 2). Positive residuals, although lower than 2, are found only for central-western and north-western sub-regions, which is in agreement with the PI map that shows values slightly greater than 1 in those regions.

CT2− shows a strong ridge over the north-east of the Caspian Sea that appears as a blocking system commonly associated with the Siberian high pressure in the lower troposphere; nonetheless, the Siberian high pressure only marginally affects weather in Iran, as the country is out of its direct influence in this case. The relative frequency of occurrence of this CT is 7.5%, with mean and maximum lifetimes of about 3.8 and 11 d, respectively. The PI pattern for CT2− (Figure 5(c)) suggests a low potential for precipitation throughout Iran, mostly due to the absence of strong vorticity and moisture transports over the country (Figures 3(c) and 4(c)). In this case, the positions of both the mid-tropospheric trough over the eastern Mediterranean and the Arabian anticyclone, typical pre-conditioning factors for widespread precipitation over Iran, seem to be unfavourable to moisture transports towards the country. Table 3 (third row) supports this finding showing standardized residuals lower than 2 for almost all sub-regions. In particular, a residual value lower than − 2 is found for the south-eastern sub-region (21.8% of rainy days), which characterizes CT2− as ‘no-rainy’ for the area.

CT2+ reveals a strong trough in the north-eastern sector of the selected domain and a high geopotential gradient over Iran that yields strong mid-tropospheric westerly winds over Iran (Figure 3(d)). The CT2+ has a relatively high frequency of occurrence compared to the other CTs (8.8%) and mean and maximum lifetimes of 4.2 and 14 d, respectively. The map of PI for CT2+ (Figure 5(d)) suggests only modest contributions to precipitation over northern, eastern and southern Iran. The position of the Arabian anticyclone seems to favour moisture transports over southern Iran (Figure 4(d)). The standardized residuals in Table 3 (fourth row) show values ≥ 2 in north-eastern and south-western Iran, denoting CT2+ as a ‘rainy’ CT for these sub-regions.

CT3− has a spatial configuration similar to CT2+ (prevalently zonal over Iran), but the cyclonic curvature in the uppermost section of the selected domain is now displaced westwards (Figure 3(e)). The frequency of occurrence of this CT is 7.7%, the mean and maximum lifetimes are 3.7 and 12 d, respectively (Table 2). The PI pattern for CT3− (Figure 5(e)) clearly highlights the prevailing ‘no-rainy’ character of this CT over most of the Iran (PI values lower than 1). The standardized residuals in Table 3 (fifth row) confirm this result: they are negative in all the six sub-regions and lower than − 2 in central-eastern, north-eastern and north-western sub-regions.

CT3+ is characterized by a strong ridge over Eastern Europe and a deep trough over the uppermost east sector of the study domain, with a northeast-southwest tilt (Figure 3(f)). It is, together with CT5−, the most frequent CT (frequency of occurrence 9.6%), with mean and maximum lifetimes of 3.6 and 15 d, respectively. The PI map for CT3+ suggests that it contributes to precipitation in northern and eastern Iran (PI values greater than 1), in accordance with the convergence of moisture flows over these regions (Figure 4(f)). The standardized residuals in Table 3 (sixth row) show values greater than 2 in central-eastern, north-eastern and north-western sub-regions, characterizing CT3+ as a ‘rainy’ CT for these areas.

CT4− features a trough over the central Mediterranean Sea, a trough over Pakistan and a ridge over western Iran, leading to northwesterly flow over the country (Figure 3(g)). The frequency of occurrence of this CT is 7.6%, with mean and maximum lifetimes of about 2.7 and 10 d, respectively (one of the least persistent CTs; Table 2). Its PI map (Figure 5(g)) suggests that it is a ‘no-rainy’ CT for the whole of the country. This finding is supported not only by the westernmost location of the Arabian anticyclone that limits moisture transports towards Iran (Figure 4(g)), but also by the negative standardized residuals (lower than − 2) in all sub-regions, with less than 20% of rainy days (seventh row in Table 3).

CT4+ features a deep trough over the Middle East, triggering southwesterly flow over Iran (Figure 3(h)). Due to the high gradient of the geopotential height over Iran, CT4+ is expected to play a relevant role in Iranian weather. The frequency of occurrence of this CT is 8.5%, with mean and maximum lifetimes of 3.1 and 11 d, respectively (Table 2). The corresponding spatial pattern of PI (Figure 5(h)) displays values greater than 1.5 throughout Iran, suggesting that CT4+ provides precipitation amounts significantly greater than the climatological mean, especially in areas near the Persian Gulf. The positions of both the mid-tropospheric trough and the Arabian anticyclone lead to strong moisture transports blowing from the Mediterranean Sea and the southern water bodies (Persian Gulf, Arabian Sea) towards Iran (Figure 4(h)). The standardized residuals for this CT (Table 3, eighth row) are ≥ 2 in all of the identified sub-regions, with the exception of the central-eastern sub-region, characterizing CT4+ as a ‘rainy’ CT for the entire country (percentage of rainy days between 38 and 69%).

CT5− depicts a trough over Iran, which induces northwesterly flow over the country (Figure 3(i)). Together with CT3+, it is the most frequent CT (frequency of occurrence of 9.6%) and has mean and maximum lifetimes of 3.2 and 9 d, respectively (Table 2). The PI pattern (Figure 5(i)) shows values greater than 1 in the eastern side of the country. While the western side of the country rarely experiences precipitation due to the presence of a ridge west of Iran, the eastern side records some precipitation, favoured by the moisture transports from the Caspian Sea (Figure 4(i)). The standardized residuals in Table 3 (ninth row) show that positive values are associated with the eastern sub-regions, where a value greater than 2 is found for the north-eastern sub-region (percentage of rainy days of about 43%). On the contrary, residual values lower than − 2 are associated with the western sub-regions, denoting CT5− as a ‘no-rainy’ CT for those areas.

CT5+ is characterized by a well-defined trough over the eastern Mediterranean, favouring precipitation over western Iran (Figure 3(j)). The mid-tropospheric trough and ridge axes are located near 30° and 60°E, respectively. The influence of such a large-scale atmospheric configuration, known as the Sudanese low pressure system, on precipitation occurrence over Iran was previously analysed by Mofidi and Zarrin (2005, 2006). The authors found that this system, along with the Arabian anticyclone, favours the transport of moisture from the southern and western water bodies (Red Sea, Arabian Sea, Persian Gulf and Mediterranean Sea) towards Iran, substantially contributing to precipitation over western Iran. The frequency of occurrence of CT5+ is 8.9%, while mean and maximum lifetimes are 2.9 and 15 d, respectively. The PI spatial pattern for this CT shows values ≥ 1.5 in western Iran, suggesting that this CT is particularly important for precipitation occurrence over the western side of the country (Figure 5(j)). This finding is also supported by the standardized residuals ≥ 2 in the three western sub-regions (tenth row of Table 3).

CT6− is characterized by a trough over Eastern Europe, implying mid-tropospheric westerly flow over Iran (Figure 3(k)). The frequency of occurrence of this CT is 8.5% and mean and maximum lifetimes are 2.8 and 11 d, respectively. The mid-tropospheric features and the position of the Arabian anticyclone seem to be unable to carry enough moisture into Iran (Figure 4(k)). Similarly to CT4−, the corresponding PI map shows values ≤1 all over the country (Figure 5(k)), whereas the standardized residuals are negative for all the six sub-regions (11th row in Table 3), but ≤− 2 only in the south-western sub-region.

CT6+ exhibits a ridge and a sharp trough located over the Balkans and the Red Sea (Figure 3(l)), respectively, which resembles CT5+ with the trough axis NE-SW tilted. The frequency of occurrence of CT6+ is low (7.3%) and, on average, it is the least persistent CT (mean lifetime of 2.4 d). This CT may favour precipitation occurrences over central-western Iran through the transport of maritime air masses from the Red Sea and the Persian Gulf (Figure 4(l)). The PI pattern shown in Figure 5(l) depicts values between 1 and 1.5 in western Iran, while the standardized residuals are ≥ 2 in the central-western and south-western sub-regions (Table 3, last row).

Summarizing, it can be concluded that according to Figure 5 and Table 3, the CTs that control winter daily precipitation more intensely over Iran are CT4+ and CT5+, while CT3− and CT4− present the weakest influences. In particular, while the forcing of CT4+ is apparent all over the country, the CT5+ forcing is primarily manifested over western Iran. The large-scale features characterizing these CTs (Figure 3(h) and (j)) suggest that the presence of a deep mid-tropospheric trough, westwards of Iran, and the Arabian anticyclone are pre-conditioning factors for ascending motions and moisture transports from the Mediterranean Sea, the Persian Gulf and the Arabian Sea towards Iran. This finding is in agreement with the results by Alijani (2002) and Raziei et al. (2012a), which suggest that the closer the trough is to Iran, the higher the precipitation amounts tend to be over the country. The role played by the low-level Arabian anticyclone in transporting moisture from southern water bodies into the cyclonic systems over Iran is also stressed in the latter study.

In Figure 6(a), the longitude-pressure cross section of the relative vorticity and omega-vertical velocity fields associated with CT4+, averaged over the latitude belt of 25°–40°N (i.e. the approximate latitudinal extension of Iran), are shown. It can be noted that the mid-tropospheric trough axis (core of positive relative vorticity) is at about 40°E, whilst the longitudinal belt of 45°–63°E, where Iran is located, presents strong rising motions (negative vertical velocity), which are favourable to precipitation. For CT5+ (Figure 6(b)), the mid-tropospheric trough axis is located further west, at about 30°E, whereas strong ascending motions are verified over western Iran (43°–54°E).

Figure 6.

Longitude–pressure cross sections of the relative vorticity (shading) and omega-vertical velocity (black contour lines) averaged over the latitude band (25°–40°N) for the ‘rainy’ CTs [CT4+ and CT5+ (upper)] and ‘no-rainy’ CTs [CT3− and CT4− (lower)]. Unit for relative vorticity is s−1 (shading every 0.5 × 10−5 s−1, red positive and blue negative values), for vertical velocity Pa s−1.

In order to highlight the difference between the vertical structures of the atmospheric circulation related to the aforementioned CTs and those that less affect precipitation over Iran, the relative vorticity and omega-vertical velocity for CT3− and CT4− are shown in Figure 6(c) and (d). For CT3−, there is a weak cyclonic circulation (positive relative vorticity) in the mid-upper troposphere and a wide area of sinking motion (positive omega-vertical velocity) over Iran (about 45°–63°E) that inhibits precipitation occurrence. For CT4−, however, two cores of positive relative vorticity are found in the mid-troposphere, westwards and eastwards of Iran, as well as a prevailing sinking motion over the broad longitude range 46°–70°E that includes Iran.

4. Conclusions

This study provides an analysis of the relationship between winter daily large-scale CTs and daily precipitation over Iran during the period 1965–2000. Twelve CTs, previously identified by Raziei et al. (2012b), through a K-means clustering technique applied to the PCs of the 500 hPa geopotential height fields, are considered here. Daily precipitation fields are regionalized by the PCA with Varimax rotation: six sub-regions with independent precipitation variability within the country are also isolated. The relationships between the 12 CTs and the regional winter daily precipitation are investigated by computing the spatial patterns of the PI associated with each CT, as well as contingency tables between the frequencies of occurrence of the CTs and precipitation, based on the RPCs of daily precipitation.

The results suggest that CT4+ and CT5+ favour the occurrence of precipitation over most of the country and can be considered the main ‘rainy’ CTs (i.e. the CTs that more significantly contribute to precipitation over large areas of the country). Conversely, CT3− and CT4− are essentially ‘no-rainy’ CTs, with weak control on the precipitation field. The remaining CTs provide more regional or low contributions to precipitation in the study area; as an example, CT1− is a ‘rainy’ CT for the central-eastern and south-eastern sub-regions (Figure 5 and Table 3).

Focusing on the three sub-regions identified in western Iran, the CTs that most contribute to winter precipitation are CT3+, CT4+, CT5+ and CT6+. This result is only partially in agreement with Raziei et al. (2012b; see their Tables 3–V). As expected, since daily precipitation and meteorological dry/wet events are different concepts (different time scales involved), the CTs driving daily precipitation are not necessarily the same leading to dry/wet spells defined on a monthly basis.

Additionally, these outcomes underline the key role played by the Arabian anticyclone in transporting moisture towards Iran, as previously noted by Raziei et al. (2012a). In particular, the longitudinal position of the low-level Arabian anticyclone is crucial. During the occurrence of CT4+, precipitation above normal is observed all over the country (PI spatial pattern in Figure 5(h)) and is related to the position of the Arabian anticyclone over the Indian Ocean, which triggers moisture transports towards Iran. For CT5+, instead, the incursion of moist air is limited to the western half of Iran, owing to the location of the Arabian anticyclone over eastern Saudi Arabia. Furthermore, during the two main ‘no-rainy’ CTs (CT3− and CT4−), the Arabian anticyclone is located in its westernmost position compared to the other CTs. This is in agreement with Raziei et al. (2012a), who found that the westward migration of the Arabian anticyclone from its climatological mean position favours precipitation over western Iran, while its eastward displacement favours precipitation on the eastern side of the country.

In summary, this study isolated the more (less) controlling CTs on the occurrence of winter precipitation over each sub-region of Iran. These results also represent a useful contribution for short-term precipitation predictability and for water resource management too. Spring and autumn frosts and heat waves are also important events that frequently hit many parts of Iran, with detrimental agricultural impacts. As such, the investigation of the linkage between the CTs and these weather extremes will be carried out in upcoming studies.

Acknowledgements

The NCEP/NCAR reanalysis data have been provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado (United States) through their web page http://www.cdc.noaa.gov, while the gridded precipitation data by the APHRODITE's Water Resources project through the web page http://www.chikyu.ac.jp/precip. Part of this work is supported by European Union Funds (FEDER/COMPETE-Operational Competitiveness Programme) and by national funds (FCT-Portuguese Foundation for Science and Technology) under the project FCOMP-01-0124-FEDER-022696. We also thank the anonymous reviewers for their valuable comments.

Ancillary