Spatio-Temporal Assessment of Meteorological Droughts Effect on Vegetation Droughts in Khorasan Razavi Province, Iran
Pouyan Dehghan Rahimabadi
1
(
University of Tehran
)
Sahar Nasabpour Molaei
2
(
University of Tehran
)
Esmail Heydari Alamdarloo
3
(
Department of Arid and Mountainous Reclamation Region, Faculty of Natural Resources, University of Tehran, Tehran, Iran
)
Setareh Bagheri
4
(
University of Tehran
)
Hossein Azarnivand
5
(
University of Tehran
)
الکلمات المفتاحية: Vegetation, climate, Trend, indicator, drought,
ملخص المقالة :
Vegetation cover is one of the living components of terrestrial ecosystems and plays an important role in many ecosystem processes that are strongly influenced by climatic events. Thus, meteorological droughts can significantly affect the vegetation cover, especially in arid and semi-arid regions where vegetation is more sensitive to environmental conditions. This study was conducted with the aim of analyzing the effects of meteorological droughts on the vegetation cover in different Land Use Land Cover (LULC) types in Khorasan Razavi province, Iran. For this purpose, the correlation and Linear Regression (LR) between Standardized Vegetation Index (SVI) and meteorological drought indices including Precipitation Evapotranspiration Index (SPEI) and Standardized Precipitation Index (SPI), with 3, 6 and 12-month time scales, were investigated for the period of 2001-2020. Based on the results, it was found that SVI values were negative in the years 2001, 2006, 2008, 2011, 2014 and 2015, in all LULC types, while in 2010, moderate rangeland experienced the most severe drought. The decreasing trend of SVI (increasing vegetation drought) was mostly observed in the southern parts of the province. The correlation between SVI and 6-month SPEI occupied a wider area than the other time scales (23.07%). The highest correlation between SVI and 12-month SPI was distinguished in dense forest, sparse forest, and poor rangeland, and occupied a wider area across the province (24.08%). Moreover, the highest (1.13) and lowest (0.75) changes in the regression coefficient of variations of SVI with multitemporal SPEI and SPI were belonged to moderate forest and agricultural land, respectively. Based on the results of this study, SPEI and SPI showed completely different values in various LULC types. Therefore, any types of indicators should be separately considered to study the terrestrial ecosystems in order to better identify areas affected by meteorological drought.
Spatio-Temporal Assessment of the Effect of Meteorological Droughts on Vegetation Droughts in Khorasan Razavi Province, Iran
Abstract
Vegetation cover is one of the living components of terrestrial ecosystems and plays an important role in many ecosystem processes that are strongly influenced by climatic events. Thus, meteorological droughts can significantly affect the vegetation cover, especially in arid and semi-arid regions that where vegetation is more sensitive to environmental conditions. This study was conducted with the aim of The aim of the present study is to analyze analyzing the effects of meteorological droughts on the vegetation cover in different Land Use Land Cover (LULC) types in Khorasan Razavi province, Iran. For this purpose, the correlation and Linear Regression (LR) between Standardized Vegetation Index (SVI) and meteorological drought indices including Precipitation Evapotranspiration Index (SPEI) and Standardized Precipitation Index (SPI), with 3, 6 and 12-month time scales, was were investigated for the period of 2001-2020. Multitemporal SPEI and SPI were used to determine the sequence of meteorological drought in different LULC types. Based on the results, it was found that SVI values were negative in the years 2001, 2006, 2008, 2011, 2014 and 2015, in all LULC types, while in 2010, moderate rangeland experienced the most severe drought. The decreasing trend of SVI (increasing vegetation drought) was mostly observed in the southern parts of the province. The correlation between SVI and 6-month SPEI occupied a wider area than the other time scales (23.07%). The highest correlation between SVI and 12-month SPI was distinguished in dense forest, sparse forest, and poor rangeland, and occupied a wider area across the province (24.08%). Moreover, the highest (1.13) and lowest (0.75) changes in the slope of changes of SVI with multitemporal SPEI and SPI belonged to moderate forest and agricultural land, respectively. Generally, based on the results of this study, SPEI and SPI showed completely different values in different LULC types in different time scales. Therefore, different types of indicators should be considered to study the terrestrial ecosystems in order to better identify areas affected by meteorological drought.
Key words: Climate, Drought, Indicator, Trend, Vegetation
Introduction
Vegetation cover provides valuable ecosystem services to various terrestrial ecosystems, such as carbon sequestration, livestock grazing and regulation of water resources (Pakeman et al., 2019). It is predicted that future climate warming will have a great impact on terrestrial ecosystems, including rangeland degradation and desertification (Miao et al., 2015), increasing tree mortality, stimulating fires in ecosystems and reducing carbon sequestration in vegetation cover (Allen et al., 2010). Therefore, drought can be an important factor in the quantity and quality of vegetation (Zandi et al., 2021). A shift towards more variable rainfall leads to an overall reduction in the forage biomass resources and an increase in the vegetation cover vulnerability (Williams and Albertson, 2006). Hence, quantifying the vegetation cover response to meteorological droughts is essential for ecological management in terrestrial ecosystems (Wei et al., 2022). However, it is difficult to predict and quantify meteorological drought because it is a complex phenomenon, appears gradually, does not have a specific beginning and is affected by numerous variables. That is why the characteristics of drought, its intensity and frequency are spatially and temporarily different (Iglesias et al., 2009). Different terrestrial and meteorological parameters are important for the drought parameterization. Because the global air temperature is continuously increasing affected by climate change and the spatio-temporal pattern of rainfall and land use is expected to change significantly in the coming decades (Liu et al., 2014). Therefore, it is necessary to develop the more accurate and effective drought monitoring tools. In this case, Remote Sensing (RS) plays an important role in drought studies, due to the spatial and temporal advantages it can provide (West et al., 2019).
Meteorological droughts monitoring and the evaluation of its impact on vegetation cover is essential for proper management in terrestrial ecosystems. Traditionally, this goal has been achieved by the direct use of meteorological drought indices which is calculated from in situ observation data collected at a limited number of meteorological stations. While these indices are accurate, they are not capable of providing spatially accurate drought estimates due to geographic and economic constraints when establishing in situ stations (Ozelkan et al., 2016). However, todays, there are different indicators for drought diagnosing and monitoring in real time (Ghazaryan et al., 2020), announcing the beginning or end of the drought (Tarnavsky and Bonifacio, 2020), studying drought levels and drought response measures (Zargar et al., 2011), analyzing the quantitative effects of droughts on environmental variables at a various spatial and temporal scales (Tarnavsky and Bonifacio, 2020) and finally announcing drought conditions (Adedeji et al., 2020; Tsakiris and Vangelis, 2020). These indicators help drought warning, monitoring and contingency planning.
Several methods have been used to monitor and evaluate meteorological droughts and their effects on vegetation cover, through the development of various climatic and vegetation indices based on RS techniques. Ariapour et al. (2013) estimated the changes in vegetation cover and LULC using Landsat TM and ETM+ images in Sabzevar city. The results indicated the appropriateness of RS technology in order to accurately estimate the area of LULC and vegetation changes was confirmed by these researchers. Ezzine et al. (2014) investigated the consistency of three drought indices in different LULC classes during a 15-year period (1998-2012). Their results showed a stronger relationship between Standardized Precipitation Index (SPI) and Standardized Vegetation Index (SVI) than SPI and Standardized Water Index (SWI) in autumn and winter seasons. Su et al. (2017) proposed a framework of standard drought indicators to describe all aspects of drought, duration, onset, intensity, expansion and simultaneous examination of three types of drought such as agricultural, meteorological and hydrological droughts in order to harmonize the drought monitoring and reduce the contradictions caused by the lack of a global standard for measuring and describing drought. Safari Shad et al. (2017) investigated the drought in Isfahan province using SPI, TCI, VCI and NDVI. The results of the correlation among different drought indices showed that VCI and TCI have the highest and lowest correlations with meteorological droughts, respectively. Heydari Alamdarloo et al. (2020) investigated the effect of climate fluctuations on vegetation dynamics in Northwest of Iran. They reported that NDVI had the best correlation with temperature in 8-12 °C class with R2 value of 0.36 and precipitation in 213-300 mm classes with a R2 value of 0.38. Ebrahimi Khusfi and Zarei (2020) assessed the relationship between meteorological droughts and vegetation degradation using SPI and NDVI in the Meyghan plain in Markazi province. They reported that the decrease in precipitation increased the sensitivity of the low density vegetation cover (poor rangeland) to meteorological droughts compared to dense vegetation cover (agricultural land).
Considering the above and emphasizing the dangers of droughts, it is necessary to realize the relationship between vegetation cover and meteorological drought to understand the effects of water scarcity on vegetation production and initiate prevention. Since Khorasan-Razavi province, in northeast of Iran, has been introduced as one of the drought-prone regions in the country, which suffers a lot of damage due to this phenomenon every year (Erfanian and Alizadeh, 2009)Hence, this study aimed to investigate the meteorological drought in Khorasan Razavi province, Iranthis province, using SPEI and SPI and their effect on vegetation cover in different LULC types. In this study both SPEI and SPI were used to assess the effect of each of them on vegetation cover and separate and compared their effects.
Materials and Methods
Study area
The case study is Khorasan Razavi province (33⁰ 52' - 37⁰ 52' N; 56⁰ 19' - 61⁰ 16' E), located in the northeast of Iran, with an area of 127222.67 Km2. About three fourth of the province is dominated by arid and semi-arid climates. The average annual rainfall is 250 mm, which changes from 116 mm/year in the north to 313 mm/year in the south and the annual mean temperature is about 14.5 °C (Ziyaee et al., 2018). Moreover, The the altitude varies between 231 and 3308 meters above sea level (Fig. 1).
Fig. 1 Location of Khorasan Razavi province in Iran
Land Use Land Cover
Land Use Land Cover (LULC) was used to represent the effect of meteorological drought on dynamics of different types of vegetation cover. In this study, LULC map was extracted from the LULC map prepared by Natural Resources and Watershed Management Organization of Iran (www.frw.ir). This map was divided LULC into 13 types including dense forest (3.78%), moderate forest (0.12%), sparse forest (1.44%), agricultural land (24.78%), good rangeland (2.49%), moderate rangeland (21.17%), poor rangeland (37.73%), bareland (2.51%), rock (0.48%), saltland (2.62%), sand dune (2.14%), urban (0.41%), wetland (0.33%) (Fig. 2).
Fig. 2 LULC map of Khorasan Razavi province
Methodology
In the present study, Standardized Vegetation Index (SVI) was adopted to represent the vegetation cover dynamics. SVI maps prepared using monthly Enhanced Vegetation Index (EVI) acquired from MOD13A2 product of MODIS for the period of 2001-2020. EVI maps for May were used, because natural vegetation cover is at its peak on this month in Khorasan Razavi province. Additionally, the monthly precipitation and temperature data for 10 meteorological stations existing in the study area within the period of 2000-2020 were used (Table 1). Two of the most widely used meteorological drought indices including Precipitation Evapotranspiration Index (SPEI) and Standardized Precipitation Index (SPI) were calculated to characterize the intensity of meteorological drought. The 3, 6 and 12-month SPEI and SPI were considered as short, medium and long time scales, respectively. Afterwards, coefficient of determination (R2) and slope of the Linear Regression (LR) between SVI and multitemporal SPEI and SPI to assess the relationship between vegetation cover dynamics and meteorological droughts and vegetation cover dynamics. The maximum value of the R2 for each pixel was considered to evaluate the response of SVI to different SPEI and SPI time scales. This means that if a pixel of the SVI maps showed the best correlation with the 3-month SPEI, cumulative precipitation and reference evapotranspiration from February was computed.
Table 1 The properties of the meteorological stations
Row | Station Name | Longitude (E) | Latitude (N) | Altitude (m) | Type |
1 | Golmakan | 36.48 | 59.28 | 1176.0 | synoptic |
2 | Gonabad | 34.35 | 58.68 | 1056.0 | synoptic |
3 | Kashmar | 35.27 | 58.47 | 1109.7 | synoptic |
4 | Mashhad | 36.24 | 59.63 | 999.2 | synoptic |
5 | Neyshabur | 36.27 | 58.8 | 1213.0 | synoptic |
6 | Quchan | 37.10 | 58.45 | 1287.0 | synoptic |
7 | Sabzevar | 36.21 | 57.65 | 962.0 | synoptic |
8 | Sarakhs | 36.54 | 61.15 | 977.6 | synoptic |
9 | Torbat-E-Jam | 35.29 | 60.56 | 950.4 | synoptic |
10 | Torbat-E Heydariyeh | 35.33 | 59.21 | 1451.0 | synoptic |
It should be noted that the response of vegetation cover to meteorological droughts were analyzed and compared in a 20-year period in seven LULC types including dense forest, moderate forest, sparse forest, good rangeland, moderate rangeland, poor rangeland and agricultural land. The bareland, rock, saltland, sand dune, urban, wetland classes were considered non-vegetated vegetation types and masked in the results.
Standardized Vegetation Index
The SVI is used forto monitoring the vegetation cover dynamics. In this study, EVI maps were used to generate SVI maps. EVI is a more robust proxy for biomass due to its improved resistance to soil and atmospheric contamination, compared to NDVI (Huete et al., 2002; Matsushita et al., 2007; Garroutte et al., 2016). The SVI is used for monitoring the vegetation cover dynamics. SVI This index describes the probability of variation of the normal EVI over the time, in a certain interval (Veneros and García, 2022). In this study, This index SVI was computed from the EVI values for each pixel as Equation 1:
SVI = 
(Equation 1)
Where; SVI is Standardized Vegetation Index, EVIi is EVI for month i (May) and EVImean is mean value of EVI in May during 2001-2020 and STD is standard deviation of EVI in May during 2001-2020.
After calculating SVI for years 2001 to 2020, Mann-Kendall test was used to detect the vegetation cover trend.
Mann-Kendall test
The Mann-Kendall test was developed by Mann (1945) and then evolved by Kendall (1975). This test is calculated based on Equation 2:
(Equation 2)
Where; the values of xi and xj are consecutive data, n is the length of the time series, and the sign function can be calculated as Equation 3:
(Equation 3)
The mean of )E (S)) and the variance (Var (S)) are obtained as Equations 4 and 5:
(Equation 4)
(Equation 5)
Where; tp is the number of sequences for pth value and p is the number of sequence values. The second component in the above formula is an adjustment for the sequence or sensitive data. The standardized statistics of ZM test are obtained from the Equation 6.
(Equation 6)
The positive and negative values of the ZM trend reflect increasing and decreasing trend, respectively. The threshold of -1.96 > ZM > +1.96 shows a significant trend at 95% confidence level.
Precipitation Evapotranspiration Index
SPEI is an extended form of SPI. This index is developed to consider precipitation and potential evapotranspiration in determining meteorological drought. Unlike SPI, SPEI shows the main effect of increase in temperature on water demand. Thus, SPEI is an appropriate index to assess the intensity of drought in the increasing global warming (Vicente-Serrano et al., 2010). In this study, the monthly SPEI was calculated to determine the climate change trends to investigate the response of vegetation to climate change using precipitation and potential evapotranspiration data.
To estimate potential evapotranspiration, Thornthwaite (1948) equation was used as:
(Equation 7)
Where; 
is the correction coefficient that is determined based on the month and the latitude of the studied area, 
is the average temperature of the month in °C, I is thermal index calculated for the whole year and α is a coefficient computed based on I.
Next, the monthly water deficit or surplus (Equation 9) is computed and after that, 
is calculated.
(Equation 8)
(Equation 9)
Where; i is the respective year, j is the respective month and k is time scale (k= 3, 6 and 12).
In order to standardize the series of water deficit or surplus observations to determine SPEI according to Vicente-Serrano et al. (2013), three-factor log-logistic distribution was used as the most appropriate adaptive distribution based on Equation 10:
(Equation 10)
Where; α, β, and γ are scale, shape, and boundary factors, respectively, for D values in the interval γ < D < ∞. Also, Equation 11 shows the probability distribution function of the observation series of water deficit or surplus according to the log-logistic distribution.
(Equation 11)
Eventually, based on F(x), the SPEI was computed as Equation 12:
(Equation 12)
Where; 
, 
, 
, 
, 
and 
. 
if P=1-F(x)
but if P > 0.5, P was replaced by 1-P, where; P represents precipitation.
SPEI has positive and negative values, the more negative value indicates the more drought intensiy and the positive values represent wet condition.
Standardized Precipitation Index
SPI was computed with monthly precipitation data for the period of 2000–2020 to assess the intensity of meteorological drought. To calculate the SPI, the precipitation data for each station should be fitted to the gamma probability distribution function. The probability density function of this distribution is given in Equation 13 (McKee et al., 1993):
(Equation 13) |
|
Where; x is the amount of precipitation, α is the shape parameter, β is the scale parameter, and Γ(α) represents the gamma function. It should be noted that x, α and β values must be greater than zero. α and β are calculated using Equations 14, 15 and 16:
