This study was conducted in the Kashafrood Basin (35° 40′ -36° 03′ N, 58° 02′ -60° 08′ E; Fig. 1), Khorasan Province in northeastern Iran. The Kashafrood Basin includes the Mashhad-Fariman, Mashhad-Chenaran, and Chenaran-Ghoochan plains, and each of them has a weather station. The study area is characterized by a cool and dry climate. Physiographic details of the three weather stations are included in Table 1.

Table 1

Table 1

Table 1 Characteristics of the three weather stations Station Latitude Longitude Elevation (m) Average Tmax(° C) Average Tmin (° C) Precipitation (mm) Climate Mashhad 36° 16′ N 59° 38′ E 999 21.6 8.3 256 Semi-arid Ghoochan 37° 04′ N 58° 30′ E 1287 19.4 6.1 308 Semi-arid Golmakan 36° 29′ N 59° 17′ E 1176 20.4 6.7 208 Arid Note: Tmax, maximum temperature; Tmin, minimum temperature.

Table 1 Characteristics of the three weather stations

Fig. 1

	Figure Option ViewDownloadNew Window
	Fig. 1 Location of Kashafrood Basin (left map) in the Khorasan Province (lower-right map) of Iran (upper-right map)

2.2.1 SPI (Standardized Precipitation Index)

The SPI is the most popular drought index (Karabulut, 2015) and is a widely recognized index for characterizing meteorological droughts (Hayes et al., 1999; Deo, 2011). McKee et al. (1993, 1995) defined SPI suitable for different timescales (1, 3, 6, 12, 24 and 48 months), and the output values ranged from -2.0 to 2.0. Because precipitation data may be fi tted by a gamma distribution, the SPI is calculated as following using a probability density function of the gamma distribution:

$g(x)=\frac{1}{{{\beta }^{\alpha }}\Gamma \left( \alpha \right)}{{x}^{\alpha -1}}{{\text{e}}^{\frac{-x}{\beta }}}\text{ }$ (x> 0), (1)

where $\Gamma \left( \alpha \right)$ is the gamma function; x (mm) is the amount of precipitation (x> 0); α is the shape parameter (α > 0); and β is the scale parameter (β > 0). More details can be found in Edwards and McKee (1997) and Dogan et al. (2012).

2.2.2 PNI (Percent of Normal Index)

The PNI was described by Willeke et al. (1994) as a percentage of normal precipitation. It can be calculated for different time scales (monthly, seasonally, and yearly). PNI has been found to be rather effective for describing drought for a single region or/and for a single season (Hayes, 2006). PNI is calculated as following:

$PNI=\frac{{{P}_{i}}}{P}\times 100, $ (2)

where P_i is the precipitation in time increment i(mm), and P is the normal precipitation for the study period (mm).

2.2.3 DI (Deciles Index)

The DI was defined a ranking of the precipitation in a particular time interval over the entire historic period (Gibbs and Maher, 1967). Specifically, monthly historical precipitation data are sorted from lowest to highest and divided into ten equal categories or deciles. So, precipitation in a given month can be placed into the historical context by decile.

2.2.4 EDI (Effective Drought Index)

The EDI is calculated in daily time step and its values are standardized in a similar way with that for calculating SPI values. The EDI was originally developed by Byun and Wilhite (1999) to overcome some limitations of other indices. The value of EDI generally ranges from -2.5 to 2.5. Near normal conditions are indicated when EDI ranges from -1.0 to 1.0, while extreme drought conditions are indicated when EDI is less than or equal to -2.0. Effective precipitation should be calculated firstly before obtaining the EDI.

$E{{P}_{i}}=\sum\limits_{n=1}^{i}{\left( \frac{\left( \sum\nolimits_{m=1}^{n}{{{P}_{m}}} \right)}{n} \right), }$ (3)

where EP_i is the effective precipitation (mm), which represents the valid accumulations of precipitation; P_m is the precipitation over the previous m days (mm); and n is the duration of the preceding period (day). When i=365, then EP₃₆₅ shows available precipitation accumulated over 365 days. More details about the calculation of EDI can be found in Byun and Wilhite (1999).

2.2.5 CZI (China-Z Index) and MCZI (Modified CZI)

The National Climate Center of China developed the CZI in 1995 as an alternative to the SPI (Ju et al., 1997) when mean precipitation follows the Pearson type III distribution. CZI is calculated as:

$CZ{{I}_{ij}}=\frac{6}{{{C}_{si}}}\times {{\left( \frac{{{C}_{si}}}{2}\times {{\phi }_{\iota j}}+1 \right)}^{{}^{1}/{}_{3}}}-\frac{6}{{{C}_{si}}}+\frac{{{C}_{si}}}{6}$ (4)

where i is the time scale of interest and jis the current month; CZI_ij means the CZI’ s amount of the current month (j) for period i; C_si is the coefficient of skewness; and φ _tj is the standardized variation. Further details can be found in Wu et al. (2001). Furthermore, the MCZI can also be calculated using the formula above but substituting the median precipitation for mean precipitation.

2.2.6 ZSI (Z-Score Index)

The ZSI is occasionally confused with SPI. However, it is more analogous to CZI, but without the requirement for fitting precipitation data to either gamma distribution or Pearson type III distribution. ZSI can be calculated by the following equation:

$ZSI=\frac{{{P}_{i}}-\bar{P}}{SD}, $ (5)

where $\overline{P}$ is the mean monthly precipitation (mm); P_i is precipitation in a specific month (mm); and SD is the standard deviation of any time scale (mm).

2.2.7 RAI (Rainfall Anomaly Index)

The RAI considers two anomalies, i.e., positive anomaly and negative anomaly. First, the precipitation data are arranged in descending order. The ten highest values are averaged to form a threshold for positive anomaly and the ten lowest values are averaged to form a threshold for negative anomaly. The thresholds are calculated by Equations 6 and 7, respectively:

$RAI=3\times \left[ \frac{\left( p-\bar{p} \right)}{\left( \bar{m}-\bar{p} \right)} \right], $ (6)

$RAI=-3\times \left[ \frac{\left( p-\bar{p} \right)}{\left( \bar{m}-\bar{p} \right)} \right], $ (7)

where p is the actual precipitation for each year (mm); $\overline{p}$is the long-term average precipitation (mm); and $\overline{m}$ is the mean of the ten highest values of p for the positive anomaly and the mean of the ten lowest values ofp for the negative anomaly.

Daily precipitation data from 1987 to 2010 were obtained from two sources: three weather stations (Mashhad, Golmakan and Ghoochan) and the AgMERRA gridded dataset (http://data.giss.nasa.gov/impacts/agmipcf/agmerra/). We used SPSS 16.0 software and MATLAB 2013a for data analysis. Furthermore, we used the C# language (Visual Studio 2013 and .NET Framework 4.5.1) to develop a software package for calculating the meteorological drought indices.

We developed the MDM (Meteorological Drought Monitoring) software package for calculating different precipitation-based meteorological drought indices. Normally, if different drought indices can be simultaneously calculated for a given time interval and also for a given region, drier-than-mean or wetter-than-mean conditions can be more confidently defined (Smakhtin and Hughes, 2007). Thus, user-friendly software is a rather useful tool for calculating and comparing multiple locations, different timescales, and different data sources. The MDM software package is currently based on calculations from two sources of data covering the period of 1980-2010. The first is the weather station data file, which includes daily precipitation in Excel format. The second one is a database of daily precipitation from AgMERRA. The user can click the map on the desired point in the package and calculate all indices at 0.25° grid location. There is a complete help instructions in the package describing all setup steps. The detailed description of MDM capabilities and instructions for data analysis are available at https://www.agrimetsoft.com.

We classified dry months as those in which each one of the aforementioned indices falls into one of the three categories: extreme, severe, or moderate. For each year, we counted the frequency of months for the interested location when each one of the indices fell into one of the three drought categories (i.e., extreme, severe, and moderate). For example, the SPI index for 1989 in Ghoochan had 1 month of extreme drought, 0 month of severe drought and 2 months of moderate drought. We then applied multipliers of 3 for extreme drought, 2 for severe drought month, and 1 for moderate drought months to get a total yearly degree of dryness index (DDI) of 5 for SPI at Ghoochan station (i.e., DDI=(1× 3)+(0× 2)+(2× 1)=5). Then we averaged the DDI values from the three weather stations to get an average yearly degree of dryness index of 3 for SPI for station-observed precipitation data in 1989. The same method was used for the AgMERRA derived index outputs, with the exception of the final averaging across sites. The DDI can be calculated by Equations 8 and 9:

$DDI_{y}^{st}=\sum\limits_{int=1}^{{{N}_{int}}}{{{a}_{int}}}\times {{N}_{int\text{, }y}}, $ (8)

$DD{{I}_{y}}=\frac{\left[ \sum\limits_{st=1}^{{{N}_{st}}}{DDI_{y}^{st}} \right]}{{{N}_{st}}}, $ (9)

where DDIst y is the degree of dryness index of the station in each year; a_int is the intensity of drought, with 1 for moderate drought, 2 for severe drought, and 3 for extreme drought; N_{int, y}is the number of dry months for each drought category in each year; DDI_y is the average value of degree of dryness index in each year for all stations; and N_st is the number of stations (N_st=3 in this study).

The performance of AgMERRA datasets was evaluated using five widely-used statistical indices, i.e., relative absolute bias (ABIAS), mean errors (ME), mean absolute error (MAE), Pearson’ s correlation coefficient (r), and coeffi cient of determination (R²). Specifically, the ABIAS was computed to describe the absolute magnitude of systematic bias of the difference between the station-observed precipitation data and the AgMERRA precipitation data (Eq. 10). The ME was selected to represent the average difference between the station-observed precipitation data and the AgMERRA precipitation data (Eq. 11). The MAE was used to determine the average magnitude of the error (Eq. 12). Pearson’ s correlation coefficient (r) was used to measure the degree of agreement between the two sources of data (Eq. 13). The coeffi cient of determination (R²) described the proportion of the total variance in the station-observed precipitation data that can be explained by the AgMERRA precipitation data (Eq. 14). It ranges from 0 to 1 with higher values indicating stronger agreement.

$ABIAS=\frac{\sum\limits_{i=1}^{N}{\left| A{{g}_{i}}-S{{t}_{i}} \right|}}{\sum\limits_{i=1}^{N}{S{{t}_{i}}}}\times 100%, $ (10)

$ME=\frac{\sum\limits_{i=1}^{N}{\left( A{{g}_{i}}-S{{t}_{i}} \right)}}{N}, $ (11)

$MAE=\frac{\sum\limits_{i=1}^{N}{\left| A{{g}_{i}}-S{{t}_{i}} \right|}}{N}, $ (12)

$r=\frac{\left( \sum\limits_{i=1}^{N}{Ag\times St} \right)-\left( \frac{\sum\limits_{i=1}^{N}{Ag\times \sum\limits_{i=1}^{N}{St}}}{N} \right)}{\sqrt{\left( \sum\limits_{i=1}^{N}{\mathop{Ag}^{2}-{{\frac{\left( \sum\limits_{i=1}^{N}{Ag} \right)}{N}}^{2}}} \right)}\times \sqrt{\left( \sum\limits_{i=1}^{N}{\mathop{St}^{2}-{{\frac{\left( \sum\limits_{i=1}^{N}{St} \right)}{N}}^{2}}} \right)}}$, (13)

${{R}^{2}}={{\left. \left\{ \frac{\sum\limits_{i=2}^{N}{\left( S{{t}_{i}}-\overline{St} \right)\times \left( A{{g}_{i}}-\overline{Ag} \right)}}{{{\left[ \sum\limits_{i=1}^{N}{{{\left( S{{t}_{i}}-\overline{St} \right)}^{2}}} \right]}^{0.5}}\times {{\left[ \sum\limits_{i=1}^{N}{{{\left( A{{g}_{i}}-\overline{Ag} \right)}^{2}}} \right]}^{0.5}}} \right. \right\}}^{2}}, $ (14)

where N is the total sets of AgMERRA precipitation data or station-observed precipitation data; Ag_i and St_i are the AgMERRA precipitation data (mm) and station-observed precipitation data, respectively; and $\overline{Ag}$ and $\overline{St}$ are the average values of AgMERRA and station-observed precipitation data (mm), respectively.

We first calculated the statistical indices (i.e., ABIAS, ME, MAE, R² and r) to compare the AgMERRA precipitation data with the station-observed precipitation data at three stations in the Kashafrood Basin. These results are presented in Table 2 at monthly time scale. Figure 2 shows the relationships between average monthly precipitation over each selected grid box for the three stations (Mashhad, Ghoochan and Golmakan) and the corresponding values from the three stations. These results show that there is a good agreement between the station-observed precipitation data and the AgMERRA precipitation data, with R²=0.9025 for Mashhad, R²=0.8437 for Ghoochan, and R²=0.6924 for Golmakan. As shown in Table 2, the ME values ranged from -2.20 to 0.39 mm; the ABIAS values ranged from 21.0% to 37.5%; and the r values ranged from 0.85 to 0.96. It means that the AgMERRA precipitation data are quite consistent with the station-observed precipitation data. So, the AgMERRA precipitation data can be acceptable for monitoring meteorological droughts.

Table 2

Table 2

Table 2 Statistical indices between the AgMERRA precipitation data and the station-observed precipitation data for Mashhad, Ghoochan, Golmakan stations and for the Kashafrood Basin Region ABIAS (%) MAE (mm) ME (mm) r Ghoochan 28.4 7.17 -0.67 0.92 Golmakan 37.5 6.45 0.39 0.85 Mashhad 25.2 4.50 -2.20 0.96 Kashafrood Basin 21.0 4.20 -0.83 0.95 Note:ABIAS, relative absolute bias; MAE, mean absolute error; ME, mean errors; r, Pearson’ s correlation coefficient.

Table 2 Statistical indices between the AgMERRA precipitation data and the station-observed precipitation data for Mashhad, Ghoochan, Golmakan stations and for the Kashafrood Basin

Fig. 2

	Figure Option ViewDownloadNew Window
	Fig. 2 Relationships of the average monthly precipitation data between the AgMERRA and the weather stations (Mashhad (a), Ghoochan (b) and Golmakan (c)), as well as relationship of the average precipitation data between the AgMERRA and the weather stations over the whole Kashafrood Basin (d)

The eight drought indices (i.e., SPI, PNI, DI, EDI, CZI, MCZI, RAI, and ZSI) were calculated for the three stations (Mashhad, Ghoochan, and Golmakan) from 1987 to 2010. For Mashhad station, there were close relationships between the AgMERRA-derived drought indices and the station-derived drought indices, all correlation coefficients being larger than 0.82 (Table 3; Fig. 3). The Pearson’ s correlation coefficients for SPI, PNI and DI were more or less identical (0.91, 0.89 and 0.89, respectively). It should be noted that similar conclusions were drawn by Keyantash and Dracup (2002), who found that SPI and DI are the two most robust indices for monitoring meteorological drought in Oregon. Quiring (2009) recommended SPI, DI, and PNI as the best indices of meteorological drought. In our study, the trends of these three indices (SPI, DI, and PNI) were very similar (Fig. 3). The values of CZI, ZSI, and RAI were nearly the same for the observation period. In Mashhad, the trends of MCZI were somewhat different from those of CZI (Fig. 3). Morid et al. (2006) also found that MCZI was a poor detector of meteorological drought. It should be particularly pointed out that 2008 was the driest year in Mashhad station during the study period (1987-2010), as suggested by all indices.

Table 3

Table 3

Table 3 Pearson’ s correlation coefficients between the AgMERRA-derived drought indices and the station-derived drought indices for the three stations Station SPI PNI DI EDI CZI MCZI RAI ZSI Ghoochan 0.87 0.88 0.88 0.89 0.76 0.77 0.76 0.79 Golmakan 0.50 0.50 0.50 0.65 0.33 0.31 0.30 0.31 Mashhad 0.91 0.89 0.89 0.91 0.82 0.84 0.82 0.83 Note: SPI, Standardized Precipitation Index; PNI, Percent of Normal Index; DI, Deciles index; EDI, Effective Drought Index; CZI, China-Z Index; MCZI, Modified CZI; RAI, Rainfall Anomaly Index; ZSI, Z-score Index.

Table 3 Pearson’ s correlation coefficients between the AgMERRA-derived drought indices and the station-derived drought indices for the three stations

Fig. 3

	Figure Option ViewDownloadNew Window
	Fig. 3 Comparison of eight drought indices derived from the AgMERRA precipitation data and from the station-observed precipitation data at an annual time scale for Mashhad station. SPI, Standardized Precipitation Index; PNI, Percent of Normal Index; DI, Deciles index; EDI, Effective Drought Index; CZI, China-Z Index; MCZI, Modified CZI; RAI, Rainfall Anomaly Index; ZSI, Z-score Index.

For Ghoochan station, there was a good correspondence between the AgMERRA-derived drought indices and the station-derived drought indices, all correlation coefficients being larger than 0.76 (Table 3; Fig. 4). Through comparing MCZI and CZI indices, we found that MCZI represented the range of wet years better than CZI, while CZI represented the dry years better than MCZI. Shahabfar and Eitzinger (2013) described MCZI as a best performer during rainy seasons in mountainous and semi-mountainous areas. In general, SPI, PNI, DI, CZI and ZSI showed similar trends (Fig. 4). Wu et al. (2001) and Morid et al. (2006) obtained similar outputs from SPI, CZI, and ZSI. In our study, the precipitation recorded by AgMERRA and the derived drought indices for years of 2001 and 2002 were lower than the precipitation recorded at weather stations and the derived drought indices. All indices indicated that 2001 and 2008 were the driest years in Ghoochan station during the study period (1987-2010).

Fig. 4

	Figure Option ViewDownloadNew Window
	Fig. 4 Comparison of eight drought indices derived from the AgMERRA precipitation data and from the station-observed precipitation data at an annual time scale for Ghoochan station

In contrast to Mashhad and Ghoochan stations, the correlations (> 0.31) between the AgMERRA-derived drought indices and the station-derived drought indices for Golmakan station (r< 0.65; Table 3; Fig. 5) were not robust at all. For example, 1992 and 1993 were two years when the station-derived drought indices showed an extreme drought while the AgMERRA-based drought indices showed a normal condition. This discrepancy was likely caused by the 300-m elevation difference between the Golmakan station and the nearest pixel of AgMERRA. For Golmakan, SPI, PNI, DI, and EDI showed a good agreement between the AgMERRA-based drought indices and the station-derived drought indices in presenting wet and dry spells. The values of CZI, MCZI, RAI, and ZSI had similar trends (Fig. 5), with 1993 being the driest year in Golmakan station during the study period (1987-2010).

Fig. 5

	Figure Option ViewDownloadNew Window
	Fig. 5 Comparison of eight drought indices derived from the AgMERRA precipitation data and from the station-observed precipitation data at an annual time scale for Golmakan station

Pearson’ s correlation coefficients among all the indices derived from the AgMERRA precipitation data are shown in Table 4 and the correlations from the station-observed precipitation data are shown in Table 5. The correlation coefficient between EDI and RAI was the lowest (0.60 for station-observed precipitation data and 0.74 for AgMERRA precipitation data). The second lowest correlation was found between EDI and ZSI (0.68 for station-observed precipitation data and 0.76 for AgMERRA precipitation data). In general, the correlation coefficients among the drought indices obtained from the AgMERRA precipitation data are higher than those obtained from the station-observed precipitation data. In both the AgMERRA and station-observed datasets, the SPI, PNI, and DI and also CZI and MCZI, and ZSI and RAI showed higher correlation coefficients. These results agreed with the findings of Wu et al. (2001), Dogan et al. (2012) and Asefjah et al. (2014). Given the strong correlations between PNI, DI and SPI (R²=0.99), we only used SPI to define the number of dry months during the study period. Similarly, we only used CZI in the following analysis due to the high correlation between CZI and MCZI (R²=0.99).

Table 4

Table 4

Table 4 Pearson’ s correlation coefficients between drought indices derived from the AgMERRA precipitation data on an average annual basis SPI PNI DI EDI CZI MCZI RAI ZSI SPI 1.00 0.99 0.99 0.84 0.94 0.95 0.94 0.95 PNI 0.99 1.00 0.99 0.85 0.94 0.95 0.94 0.95 DI 0.99 0.99 1.00 0.85 0.94 0.95 0.94 0.95 EDI 0.84 0.85 0.85 1.00 0.78 0.81 0.74 0.76 CZI 0.94 0.94 0.94 0.78 1.00 0.99 0.98 0.98 MCZI 0.95 0.95 0.95 0.81 0.99 1.00 0.96 0.97 RAI 0.94 0.94 0.94 0.74 0.98 0.96 1.00 0.99 ZSI 0.95 0.95 0.95 0.76 0.98 0.97 0.99 1.00

Table 4 Pearson’ s correlation coefficients between drought indices derived from the AgMERRA precipitation data on an average annual basis

Table 5

Table 5

Table 5 Pearson’ s correlation coefficients between drought indices derived from the station-observed precipitation data on an average annual basis SPI PNI DI EDI CZI MCZI RAI ZSI SPI 1.00 0.99 0.99 0.84 0.94 0.96 0.87 0.92 PNI 0.99 1.00 0.99 0.86 0.92 0.95 0.85 0.90 DI 0.99 0.99 1.00 0.86 0.92 0.95 0.85 0.90 EDI 0.84 0.86 0.86 1.00 0.73 0.76 0.60 0.68 CZI 0.94 0.92 0.92 0.73 1.00 0.99 0.96 0.98 MCZI 0.96 0.95 0.95 0.76 0.99 1.00 0.93 0.97 RAI 0.87 0.85 0.85 0.60 0.96 0.93 1.00 0.99 ZSI 0.92 0.90 0.90 0.68 0.98 0.97 0.99 1.00

Table 5 Pearson’ s correlation coefficients between drought indices derived from the station-observed precipitation data on an average annual basis

As shown in Table 6, DDI values from EDI and RAI suggested that the most severe drought year was the year 2001 over the entire study period (1987-2010). However, DDI values from SPI, CZI, and ZSI suggested that 2008 was the most severe drought year over the same period. In fact, the lowest average annual precipitation values recorded at these three stations were 173 mm in 2001 and 123 mm in 2008. It should be emphasized here that due to a significant reduction in rainfall, about 95% of rain-fed farms suffered severely from droughts in 2001 and also in 2008 in Iran (FAO, 2008; Ministry of Jihad-e-Agriculture Iran, 2009; USDA Foreign Agricultural Service, 2010). According to the reports from Ministry of Jihad-e-Agriculture (2009), about 2.5× 10⁶ hm² of irrigated agricultural land, 4× 10⁶ hm² of rain-fed agricultural land, and 1.1× 10⁶ hm² of gardens were affected by drought of 2001. In general, the data in Table 6 showed that 2001 and 2008 were the two driest years during the study period (1987-2010). However, the DDI values from EDI and RAI seem to be more sensitive to the station-observed droughts than the DDI values from other drought indices. The reasons may include: EDI is more sensitive to subtle changes in rainfall (Kim et al., 2009; Dogan et al., 2012; Deo et al., 2017) and RAI can better identify the positive anomaly or negative anomaly.

Although most of the drought indices are strongly cross-correlated and exhibit rather comparable seasonal and annual DDI trends, selecting a single most appropriate index of meteorological drought is still a diffi cult task and the difficulty arises from various sources including the spatial variability of climates and the temporal scales of the intended applications (Morid et al., 2006; Shahabfar and Eitzinger, 2013). Therefore, a variety of indices should always be examined to select the best or better drought indices for a specific case study. Another consideration in drought index selection is the differences in index formulations, as SPI uses the gamma distribution in its structure while EDI dose not.

Table 6

Table 6

Table 6 Average yearly Degree of Dryness Index (DDI) for five drought indices derived from the AgMERRA precipitation data and from the station-observed precipitation data across the Kashafrood Basin Year SPI CZI ZSI EDI RAI WS AgM WS AgM WS AgM WS AgM WS AgM 1987 2 5 2 4 1 3 29 29 11 13 1988 2 1 1 1 0 1 0 0 12 7 1989 3 6 2 5 1 3 23 24 19 19 1990 2 4 3 2 1 2 29 34 17 15 1991 1 1 1 0 1 0 6 8 12 11 1992 2 2 1 2 1 1 8 0 15 12 1993 3 2 3 2 2 2 12 0 17 10 1994 0 0 1 0 0 0 26 21 17 14 1995 3 4 2 4 1 3 13 25 14 18 1996 1 3 1 2 1 2 20 33 12 13 1997 1 3 1 3 0 2 10 20 12 11 1998 2 3 1 2 1 2 0 0 11 14 1999 1 0 0 1 0 0 4 6 7 10 2000 4 6 3 5 2 4 24 29 15 14 2001 4 3 4 3 2 2 36 35 20 16 2002 1 1 1 0 1 0 10 8 14 12 2003 0 1 0 0 0 0 0 0 11 9 2004 1 1 1 0 0 0 0 0 11 7 2005 2 1 1 1 1 1 0 0 14 9 2006 2 2 2 2 1 1 24 10 14 11 2007 3 2 2 1 1 1 3 3 14 12 2008 5 3 5 4 3 3 30 29 18 18 2009 2 2 1 2 0 1 7 9 9 8 2010 4 3 2 3 1 2 8 17 14 13 Note: WS, weather station; AgM, AgMERRA.

Table 6 Average yearly Degree of Dryness Index (DDI) for five drought indices derived from the AgMERRA precipitation data and from the station-observed precipitation data across the Kashafrood Basin