GIS overlay analysis for hazard assessment of drought in Iran using Standardized Precipitation Index (SPI)
- DOI : 10.5141/JEFB.2012.038
- Author: Asrari Elham, Masoudi Masoud, Hakimi Somaye Sadat
- Organization: Asrari Elham; Masoudi Masoud; Hakimi Somaye Sadat
- Publish: Journal of Ecology and Environment Volume 35, Issue4, p323~329, 01 Dec 2012
The Standardized Precipitation Index (SPI) is a widely used drought index to provide good estimations of the intensity, magnitude and spatial extent of droughts. The objective of this study was to analyze the spatial pattern of drought by SPI index. In this paper, the patterns of drought hazard in Iran are evaluated according to the data of 40 weather stations during 1967-2009. The influenced zone of each station was specified by the Thiessen method. It was attempted to make a new model of drought hazard using GIS. Three criteria for drought were studied and considered to define areas of vulnerability. Drought hazard criteria used in the present model included: maximum severity of drought in the period, trend of drought, and the maximum number of sequential arid years. Each of the vulnerability indicators were mapped and these as well as a final hazard map were classified into 5 hazard classes of drought: one, slight, moderate, severe and very severe. The final drought vulnerability map was prepared by overlaying three criteria maps in a GIS, and the final hazard classes were defined on the basis of hazard scores, which were determined according to the means of the main indicators. The final vulnerability map shows that severe hazard areas (43% of the country) which are observed in the west and eastern parts of country are much more widespread than areas under other hazard classes. Overall, approximately half of the country was determined to be under severe and very severe hazard classes for drought.
drought , GIS , hazard map , Iran , Standardized Precipitation Index
Within Iran, drought is one of the main natural hazards affecting the economy and the environment (Bruce 1994, Obasi 1994, Wilhite 2000). Droughts cause crop losses (Austin et al. 1998, Leilah and Al-Khateeb 2005), urban water supply shortages (DeGaetano 1999), social alarm (Morales et al. 2000), degradation and desertification of land (Nicholson et al. 1998, Pickup 1998, Evans and Geerken 2004), and forest fires (Flannigan and Harrington 1988, Pausas 2004). Drought is a complex phenomenon which involves different human and natural factors which contribute to the risk of,and vulnerability to drought. Although the definition of drought may be very complex (Wilhite and Glantz 1985), it is usually related to a long and sustained period in which water is scarce (Dracup et al. 1980, Redmond 2002). Drought can essentially be considered as a climatic phenomenon (Palmer 1965, Beran and Rodier 1985) related to an abnormal decrease in precipitation (Oladipo 1985, McKee et al. 1993).
Crucially, efforts toward the development of methodologies to quantify different aspects related to droughts have been made. Further efforts have been made to develop drought indices, which allow for the earlier identification of droughts, their intensity and potential surface extents of the drought. During the twentieth century, several drought indices were developed, which were based on different variables and parameters (Heim 2002). Drought indices are very important for monitoring droughts continuously in time and space, and early warning systems for droughts are based primarily on the information that drought indices provide (Svoboda et al. 2002).
The majority of drought indices have a fixed time scale. For example, the Palmer Drought Severity Index (PDSI) (Palmer 1965) has a time scale of about 9 months (Guttman 1998), though does not allow for the identification of droughts within shorter time scales. Moreover, this index has many other problems related to its calibration and spatial comparability (Karl 1983, Alley 1984, Guttman et al. 1992). To solve these problems, McKee et al. (1993) developed the Standardized Precipitation Index (SPI), which can be calculated for different time scales in order to forecast droughts based on the monitoring of different usable water resources. Moreover, the SPI is applicable to any time scale and is not specific to any one location (Hayes et al. 1999, Lana et al. 2001, Wu et al. 2005).
The SPI was published 1993 following a careful developmental procedure (Redmond 2002), and due to its robustness it has already been widely used to study droughts in different regions, including the USA (Hayes et al. 1999), Italy (Bonaccorso et al. 2003), Hungary (Domonkos 2003), Korea (Min et al. 2003), Greece (Tsakiris and Vangelis 2004), Spain (Vicente-Serrano and Begueria 2003, Lana et al. 2001), and Iran (Noruzi 2007). SPI has also been included in drought monitoring systems and management plans (Wu et al. 2005). In general, different studies have indicated the usefulness of the SPI to quantify different drought types (Edwards and McKee 1997, Hayes et al. 1999, Komuscu 1999). The long time scales (over 6 months) are considered as hydrological drought indicators (river discharges or reservoir storages) (McKee et al. 1993, Hayes et al. 1999).
The purpose of this study is to establish a spatial pattern for drought using a multi-temporal assessment of SPI in Iran. For this purpose, different aspects of drought hazard, namely, the maximum severity of drought in the period, trend of drought, and the maximum number of sequentially arid years have been prepared in the GIS, deploying the new model. It is the first attempt of its kind in Iran, and preparing such hazard maps may prove to be useful for regional planners, and policy makers for agricultural and environmental strategies, not only in Iran but also in other countries facing similar problems of water shortage.
Iran was selected as a study area for a test assessment of drought vulnerability. It covers an area of 1,648,195 km2, which lies between the latitudes of 25°14′ and 39°42′ N and the longitudes of 44°10′ and 63°11′ E. The population of the country has increased from 34 million in 1978 before of the revolution to 68 million in 2006, with an effective doubling of the population in less than thirty years. The elevation varies from sea level to around 5,500 m in the Damavand Mountains, and the climate differs widely but most parts of the country are arid or semi arid, with a mean annual rainfall of 50-2,000 mm. The average precipitation in Iran is 245 mm per year, and the main period of precipitation is during the winter (60% of total rainfall).
The meteorological data used in this study, consisting of monthly precipitation and temperature measurements for 40 synoptic stations distributed fairly evenly throughout the country (Fig. 1), were obtained from the Iran Meteorological Organization (IMO). In the present work, to determine the adequate quantity of stations with suitable scatter formula 1 was used. An exhaustive list of the selected stations is given in Table 1.
N: minimum of adequate station number (in this study: N = 40)
CV%: average of coefficient of variations of annual precipitation for synoptic stations of Iran
E%: acceptable faults (%) for the determination of correct number, for this work E% is considered to be 15%
SD: standard deviation of annual precipitation for synoptic stations of Iran
annual precipitation average for synoptic stations of Iran
To determine the common duration of the suitable statistical period for all the stations, formula 2 (Mahdavi 2002) was used. Through this formula we determined that 37.5 years is the least number of years which are needed for the current study. The duration of the data used in this study includes that from 1 January 1967 to 31 December 2009 for all stations.
N: minimum necessary annual data (in this paper: N = 37.5 years)
t: t student with the freedom degree of n-6
R: ratio of return period precipitation of 100 years to 2 years
In the next stage, annual precipitation and SPI were calculated for each year of each station using the following equation:
Pi: total precipitation in each year;
P: average precipitation in the period
SD: standard deviation of annual precipitation in the period
To check the normality of the data for each station, MINITAB.14 was used.
P-values of normality test within the software were determined. P-values > 0.05 indicate that the distribution of data for the period of record is normal, while amounts less than this indicate that the distribution of the data is not normal. In the current assessment 90% of stations were determined to have normal data which was acceptable for further statistical assessment.
The assessment of hazard of drought has been attempted by first identifying the main criteria of drought in the study area, and then by establishing the thresholds (class limits) of severity for criteria, and finally, by analyzing the hazard through X analysis. Recommendations appearing in some literature (e.g., Zehtabian and Jafari 2002, Masoudi et al. 2007, Zareiee 2009b) as well as the statistically suitable parameters of the region, such as average and standard deviation for the trend data, have also been taken into consideration while fixing the thresholds of the five classes of severity (ratings scores between 1 to 5) for each indicator. Three criteria (Table 2) have been processed in the GIS to arrive at the hazard map for each criterion.
Criteria used for drought hazard in the present model include: maximum severity of drought in the period, trend of drought, and the maximum number of sequential arid years. The amounts of SPI ≤ -0.5 were considered in order to represent drought conditions and dry years. These thresholds help in the evaluation of secondary and tertiary criteria. To determine trends of hazards for each station or its Thiessen polygon, the period of data recording was divided into two equal periods, and in each period the percentage of dry years was calculated. Then trend of hazard was calculated using following equation:
In order to ensure that the effect of all criteria gets projected in the final hazard map, the overlays of the individual hazard criterion maps, as derived from three criteria, were analyzed step by step. The severity of hazard assigned to each polygon has been assessed using the mean of all the attributes (rating scores) of criteria used in the GIS. The following equation was applied to the GIS in order to assess the hazard map of meteorological drought:
The hazard score in each polygon denotes the cumulative effect of all the criteria for qualifying the five severity classes (Table 3). This facilitated the production of final hazard map which shows the different degrees of drought hazard.
Some studies previously carried out in Iran and throughout the rest of the world have based their estimation on the ‘present state’ of hazard of drought during a specific year, and using some indices like SPI and PNPI (e.g., Ensafi Moghaddam 2007, Raziei et al. 2007). Such Indicator maps or information based solely on the present state of hazard derived from small number of recent years data are inadequate for the representation of areas which are more vulnerable to hazard (Masoudi 2010). The adequate representation of such areas requires a combination of more indices of hazard, like the maximum number of sequential years of hazard in a period, and also important index of trends showing different aspects of hazard. This kind of classification using different criteria is the first attempt of its kind to define areas with a higher risk of drought. GIS analysis not only facilitated model development but also allowed for the evaluation of spatial
correlations and the production of hazard maps.
Table 4 describes the hazard criteria maps used in the model; ‘maximum severity of drought in the period’ shows the most hazardous of three criteria used in the model. This indicator is assessed based on the worst droughts or the least amount of SPI in a year, which has occurred during the period of study (1967-2009) for each station. Ninety percent of the area in this hazard map (Fig. 2) is categorized as being under severe or very severe risk of drought, indicating that most parts of the country have experienced significant droughts in the period of study. The areas least prone to drought are coastal areas as well as some territories to the north and to the south. These results are in good agreement with other results regarding drought assessment in different regions of Iran (Ensafi Moghaddam 2007, Raziei et al. 2007, Sarhadi et al. 2008).
But the most parts of hazard map (Fig. 3) showing ‘maximum number of sequential arid years in the period’ is under slight and moderate hazard classes (61%) compared to severe and very severe hazard classes, indicating period of droughts doesn’t continue so long (more than three years) in the most parts of country. It seems impacts of drought regarding to this condition are observed more in the south-eastern parts and parts in the north and the west compared to the central parts. This aspect of drought have been used alone to show vulnerability to drought in regions, showing importance of this criteria in the hazard assessment (Feiznia et al. 2001, Zehtabian and Jafari 2002).
While the drought hazard map (Fig. 4) based on the ‘% of increasing trend’ appears to be the least hazardous among three criteria used in the model. Fifty eight percent of the area in this hazard map is categorized as having slight or no hazard classes. However, the percentage of land falling under the category of “None class” is 16%, and the area of land classified as hazard of “none” was reduced in the second data period, as compared to the first. This indicates a trend of elevating drought conditions in the country, confirming studies of the region which have indicated that climate changes is resulting in drier conditions (Zareiee 2009a, Asrari and Masoudi 2010, Masoudi and Afrough 2011). In all of the generated maps, hazardous conditions are observed more in the north-western parts of the country.
On the other hand the final hazard map of the country (Fig. 5) shows four different hazard classes. From Fig. 6, a general conclusion can be derived that in Iran an almost equal proportion of land (47%) is under severe or very severe classes of drought, compared to the areas under
slight or moderate risk of drought (53%). Hazardous lands are observed more in western, north-eastern and southeastern parts of country, while northern parts show more risk compared to southern parts. This pattern is observed in another study which reports that climate changes is leading to drier conditions, especially in northern parts of country (Zareiee 2009b). One of the most risk prone zones is in the north-western parts of country where the impact of climate change and the occurrence of drought conditions such as drying of the biggest lake of country, Orumieh Lake, can be strongly observed.
[Fig. 1.] Locations of weather stations of this study.
[Table 1.] Name of the selected stations over the study area
[Table 2.] Criteria used for the hazard assessment of drought using SPI
[Table 3.] The severity classes of hazard map produced in the GIS
[Table 4.] Percentage of areas under each hazard class, based on three criteria used in the model of drought
[Fig. 2.] Hazard map of “maximum severity of drought in the period.”
[Fig. 3.] Hazard map of “maximum number of sequential arid years in the period.”
[Fig. 4.] Hazard map of “% of increasing trend in the period.”
[Fig. 5.] Hazard map of drought vulnerability.
[Fig. 6.] Percent age of areas under hazard classes of drought vulnerability.