1 Introduction

Recent years have witnessed severe droughts and increasing temperatures have had a devastating effect on terrestrial ecosystems (Mishra and Singh 2010; Dai 2013; Kaptué et al. 2015) and they have been one of the important factors causing vegetation degradation (Heydari Alamdarloo et al. 2018). The lack of moisture due to abnormal climatic patterns leads to the drought that occurs at a different time and spatial scales (Dogan et al. 2012; Nichol and Abbas 2015). This phenomenon is occurring in almost half of the world (Field et al. 2012; Bokal et al. 2014; Masud et al. 2015).

Drought is one of the most costly natural disasters that threaten food security and water supplies (Godfray et al. 2010; Zhang and Jia 2013; AghaKouchak et al. 2015). Drought is classified into four types: meteorological, hydrological, agricultural, and economic (Dracup et al. 1980; Orville 1990; Wu et al. 2016). Meteorological droughts cause a long-term average precipitation decline of a region, while the reduction of surface water and groundwater causes a hydrological drought. Agricultural drought occurs when sufficient water is not available for plant growth. The adverse effects of three previous droughts on the economy and society cause a socio-economic drought (Zhang and Jia 2013). Meteorological droughts are evaluated using indicators, such as the standardized precipitation index (SPI) (McKee et al. 1993) and SPEI, which are derived from meteorological station parameters. Considering that the vulnerability to drought varies in different regions, meteorological drought and agricultural drought are not the same, especially on a large scale (Bachmair et al. 2018). Therefore, by recognizing the characteristics of both meteorological and agricultural droughts, one can show different regional and local effects of drought on vegetation. Nowadays, drought has been widely evaluated by studying the health of vegetation using remote sensing (Silleos et al. 2006; Nemani et al. 2009). Remote sensing of time series of vegetation indices can provide continuous information on vegetation changes and is widely used to assess vegetation relationships and climatic parameters, such as temperature and precipitation (Zhang et al. 2010; Vicente-Serrano et al. 2013).

According to various remote sensing data, different indicators of drought have been developed and used. These indicators have specific characteristics and determine the severity of drought by different methods (Zhang et al. 2017).

Various studies have been done on the relationship between meteorological drought and agricultural drought indicators derived from remote sensing data. Qian et al. (2016) studied the drought in agricultural areas of China using the vegetation condition index (VCI) from 1982 to 2010 and its correlation with climate parameters. Results of their research showed that in addition to rainfall in different regions, weather factors had different effects on drought.

Gomes et al. (2017) used SPI and vegetation health index (VHI) to determine the drought period that happened during 2017 in the southeast part of Brazil. Results of their research indicated that VHI and SPI indices had a high consistency for determining the intensity of drought.

Zhao et al. (2018) investigated the relationship between the standardized precipitation-evapotranspiration index (SPEI) at various time scales and the normalized difference vegetation index (NDVI) on Loess plateau in China. Results of this study showed that although spatial and seasonal differences occurred at different times, the highest correlation was between quarterly SPEI and the summer season. Also, plants in arid and semi-arid areas were more closely correlated with the 3-month SPEI time series; on the other hand, plants in semi-humid areas were more closely correlated with the 12-month SPEI time scale. So far, little research has been done on the relationship between SPEI and VHI and in calculating both of these indicators, the temperature parameters are used. Investigating the relationship between these two indicators can indicate the effect of meteorological drought on vegetation affected by global warming.

Drought is one of the important natural disasters that affect numerous parts of Iran, and it has deleterious effects on agriculture, water resources, and ecosystems (Zarch et al. 2011). In the last half-century, there have been 27 droughts in Iran (Amirkhani and Chizari 2010). As Iran’s Meteorological Organization announced, today, 97% of parts of Iran face drought (Shahi 2019).

Also, the results of research on climate change indicate that the frequency and duration of drought in this area will increase (Dai 2013).

For instance, Zarei et al. (2016), Gohari et al. (2017), and Mafi-Gholami et al. (2019) showed that long-term drought would happen in the south part of Iran during the middle of 2040 to 2080. This phenomenon would lead to profound environmental effects in Iran (The OFDA/CRED International Disaster Database 2011).

It is, therefore, necessary to examine the effects of drought on environmental components, such as vegetation, to manage the risk of this phenomenon. This study aimed to investigate the relationship between meteorological and agricultural droughts in different climates of Iran during the growing season. However, there have been similar studies, in the present work, the SPEI and VHI maps of Iran were used as the meteorological drought index and the agricultural drought index, respectively. In the current research, the effect of the temperature was studied exclusively. Results of this study would be useful in assessing the temperature impact on agricultural and meteorological droughts.

2 Materials and methods

2.1 Study area

With an area of 1,648,195 km2, Iran is located between 25° and 40° north latitude and between 44° and 63° east longitude (Fig. 1). There are different climates in Iran, which are primarily because of the Zagros and Alborz mountains (Heydari Alamdarloo et al. 2018). The proximity of Iran to the low-pressure region causes the northern hemisphere’s desert strip to cross through the central part of the country and creates inclement natural conditions. In general, topography and environmental factors play a principal role in the distribution and enrichment of plants and biodiversity in Iran. For instance, the highest varieties of vegetation are located in the north of the Alborz Mountains to the Caspian Sea coast, and the least varieties of plants are located in the south of Iran, which is a flat plain. In this study, the effect of meteorological drought on agricultural drought in different climates, based on the climatic classification map (Fig. 1), was determined. This map is based on Domarten climate classification prepared by Iran Forests, Range and Watershed Management Organization (http://www.frw.org.ir/00/En/default.aspx). However, to achieve the best result, some classes were merged.

Fig. 1
figure 1

Climatic classification and digital elevation model maps of Iran and location of meteorological stations

Full size image

2.2 Research methodology

2.2.1 Calculation of VHI index

VHI exhibits the effects of humidity and heat on vegetation which in turn are impacted by meteorological drought (Tran et al. 2017), and remote sensing can be effectively used to monitor responses of vegetation (Xu et al. 2018). To calculate the VHI index, LST (land surface temperature) was obtained from MOD11A2 with a spatial resolution of 1 km and a time resolution of 8 days; and the NDVI derived from the MOD13A2 product with a resolution of 1 km and a time resolution 16 days of the MODIS sensor between 2000 and 2017. It should be noted that Iran is located in five tiling h21, 22, 23 v5 and h22, 23 v6 modis (MODIS). To investigate the linkage between meteorological drought and agricultural drought, NDVI and LST were used for the whole of Iran in four periods of 16 days of the growing season, including March, April, May, and June, in each year, actually when the most natural vegetation coverages were at the peak of vitality. First, the process of mosaicking five tile images of Iran was carried out by importing the products of MOD13A2 and MOD11A2 into TerrSet software. Then, the basic time synchronization of LST and NDVI was performed by interpreting 28-day LST images and turning them into 16-day LST for each year (Heydari Alamdarloo et al. 2018).

VHI was calculated from the combination of VCI and TCI indices by Eq. (1):

$$ \mathrm{VHI}=\upalpha \times \mathrm{VCI}+\left(1-\upalpha \right)\times \mathrm{TCI} $$
(1)

where α specifies the weight of TCI and VCI indices in the VHI. Due to the lack of clarity of the needs of plants during the growth cycle, the coefficient was equal to 0.5 (Cong et al. 2017). The VCI was computed by Eq. 2 (Kogan 1990):

$$ VCI=\frac{{\mathrm{NDVI}}_i-{\mathrm{NDVI}}_{\mathrm{min}}}{{\mathrm{NDVI}}_{\mathrm{max}}-{\mathrm{NDVI}}_{\mathrm{min}}}\times 100 $$
(2)

In this equation, NDVIi is the NDVI value of a specific pixel in the year (i), and NDVImax and NDVImin are the highest and the least NDVI values of that pixel during the study period of 2000–2017. This index decreases the effect of noise arising from seasonal changes (Heydari Alamdarloo et al. 2018). TCI calculated using the following formula (Eq. 3) provides an opportunity to test the response of plants to thermal stress (Kogan 1995):

$$ \mathrm{TCI}=\frac{{\mathrm{LST}}_i-{\mathrm{LST}}_{\mathrm{min}}}{{\mathrm{LST}}_{\mathrm{max}}-{\mathrm{LST}}_{\mathrm{min}}}\times 100 $$
(3)

Here, LSTi is the value of a particular pixel in the year (i) and LSTmax and LSTmin are, respectively, the highest and least values of that pixel during the study period.

2.2.2 Calculation of SPEI index

SPEI is an appropriate indicator for monitoring the severity of droughts in the conditions of rising global temperature (Vicente-Serrano et al. 2010). To calculate SPEI, monthly precipitation data and temperature of 68 weather stations (Fig. 1) were used for the period 1987 to 2016. The index was calculated using two parameters of precipitation (P) and potential evapotranspiration (PET). The Thornthwaite (1948) equation was used to estimate the potential evapotranspiration:

$$ \mathrm{PET}=16{N}_m{\left(\frac{10{T}_m}{I}\right)}^a $$
(4)

where Nm is a correction coefficient determined because of the desired moon and latitude of the area. Tm is the average monthly air temperature (°C), I is the thermal index calculated for the whole year, and (a) was calculated based on I. Then, using Eq. (5), the shortage or excess of water was estimated monthly. In the next step,\( {X}_{i,j}^k \) was calculated from Eq. (6). It should be noted that in these equations (i) is the desired year, (j) is the desired month, and (k) is the time scale. In this research, SPEI was calculated on a 12-month time scale (k = 12).

$$ {D}_i={P}_i-{\mathrm{PET}}_i $$
(5)
$$ {X}_{i,j}^k=\left\{\begin{array}{c}{\sum}_{l=13-k+j}^{12}{D}_{i-1,l}+{\sum}_{l=1}^j{D}_{i,j}\kern5em \mathrm{if}\kern1.25em j<k\\ {}{\sum}_{l=j-k+1}^j{D}_{i,j}\kern11.75em \mathrm{if}\kern1.25em j\ge k\end{array}\right. $$
(6)

Vicente-Serrano et al. (2010) examined the compatible distributions with the D-series of observations and concluded that the three-factor log-logistic distribution was the most proper distribution and was a suitable choice for standardizing the D-series observations to determine the SPEI. The probability density function or the three parameters log-logistic distribution is:

$$ f(x)=\frac{\beta }{\alpha }{\left(\frac{x-\gamma }{\alpha}\right)}^{\beta -1}{\left[1+\left(\frac{x-\gamma }{\alpha}\right)\right]}^{-2} $$
(7)

In Eq. (7), α, β, and γ are, respectively, the scale, shape, and boundary factors for the values of D in the interval γ< D < ∞. These factors can be estimated using different methods based on observations. Among them, the torques method is simpler (Ahmad et al. 1988). When torques are calculated, the log-logistic distribution parameters using the proposed relationships (Singh et al. 1993) were estimated. Equation (8) shows the probability distribution function of the D-series based on the log-logistic distribution:

$$ F(x)={\left[1+{\left(\frac{x-\gamma }{\alpha}\right)}^{\beta}\right]}^{-1} $$
(8)

Finally, based on the function, SPEI was calculated using Eq. (9).

$$ \mathrm{SPEI}=W-\frac{C_0+{C}_1W+{C}_2{W}^2}{1+{d}_1w+{d}_2{w}^2+{d}_3{w}^3} $$
(9)

In Eq. (9), C0 = 2.515517, C1 = 0.802853, C2 = 0.010328, d1 = 1.432788, d2 = 0.189269, and d3 = 0.001308. Also, in this Eq., W=\( \sqrt{-2\ln P} \), if P = 1−F (x), ≤ 0.5 but if P > 0.5, P was replaced by 1-P and the sign of the resultant SPEI was reversed. All 12-month SPEI calculations were performed and coded in MATLAB software for 68 meteorological stations. The SPEI has positive and negative values, where the more negative SPEI represents a higher intensity of drought. After calculating the 12-month SPEI for each 68 meteorological stations, using the inverse distance weighted (IDW) method, drought severity maps according to the VHI (pixel size and pixel position, such as VHI) were constructed for March, April, May, and June of each year.

2.2.3 Statistical analysis

Using the Pearson correlation (Lee Rodgers and Nicewander 1988), the relationship between the 12-month SPEI and the VHI maps in ARC GIS 10.5 software was calculated.

The amount of correlation coefficient (r) varies between 1 and − 1, where + 1, zero, and − 1 mean a complete positive correlation, no correlation, and a complete negative correlation, respectively. To determine the amount of correlation, the absolute value of correlate action coefficient was used, and then, according to Table 1, the classification of the correlation was calculated in ArcGIS 10.5 software.

Table 1 Classification of the Pearson coefficient based on the Evans division (1996)
Full size table

The present classification in Table 1 is based on the r absolute value and indicates the correlation strength.

Using Eq. (10), the slope of the regression line was calculated to evaluate the relationship between the 12-month SPEI and VHI by linear regression equation (Chatfield 2016) to assess the effect of the 12-month SPEI fluctuations on VHI. The 12-month SPEI is an independent variable, and VHI is a dependent variable.

$$ \kern1.25em \mathrm{slop}=\frac{17\left({\sum}_{i=2000}^{2016}{\mathrm{SPEI}}_i\ {\mathrm{VHI}}_i-{\sum}_{i=2000}^{2016}{\mathrm{SPEI}}_i{\sum}_{i=2000}^{2016}{\mathrm{VHI}}_i\right)}{17{\sum}_{i=2000}^{2016}{\mathrm{SPEI}}_i^2-{\left({\sum}_{i=2000}^{2016}{\mathrm{SPEI}}_i\right)}^2} $$
(10)

The slope of the regression line between the 12-month SPEI and the VHI was calculated for all four periods. For each month, the 12-month SPEI and VHI were used. In general, the negative slope indicates an inverse relationship, the positive slope indicates a direct link, and the value of slope shows the degree of dependence of the variables. The slope of the regression line between the 12-month SPEI and the VHI was expected to be positive; with reducing 2-month SPEI drought severity increased and VHI diminished.

Natural breaks was used in ArcGIS 10.5 software to classify the slope effect (seven classes). Natural breaks was used for all four periods, and the average threshold of each class was considered the basis of main classes.

3 Results

The results of 12-month SPEI and VHI correlation analysis for different periods are presented in Fig. 2. In March, the central regions, southeast and the Persian Gulf had more correlation. In this month, the northern areas of the Zagros Mountains had the least correlation. In April, compared with March, correlation declined in the southeast and the Persian Gulf. However, it had increased at the height of the Alborz and the southern margin, and the south of the Zagros area. In May, the correlation coefficient was decreasing in most regions of Iran, but in the areas of Alborz and southwest of Iran, the correlation coefficient was increasing somewhat. In June, in the southwest and the northeast of the center of Iran, a sharp increase in correlation coefficient occurred. In general, there was no strong correlation between 12-month SPEI and VHI, but in some areas, the intensity of correlation was remarkable.

Fig. 2
figure 2

Classification of SPEI and VHI correlation during the studied period: a March, b April, c May, and d June

Full size image

Figure 3 shows the average correlation coefficient of 12-month SPEI and VHI in different climates in the four study periods. In April, the average correlation coefficient increased in dry, semi-dry, and semi-humid climates and decreased in hyper-dry and wet climates. In May, the average correlation coefficient generally reduced in all climates. However, in June, the average correlation coefficient increased in all climates, except in the wet climate, where the average correlation coefficient did not change. In general, the highest average correlation coefficient for hyper-dry, dry, semi-arid, semi-wet, and humid climates occurred in June, April, and March, respectively, and the lowest average correlation coefficient was in May in all weathers.

Fig. 3
figure 3

The average correlation coefficient of 12-month SPEI and VHI in different climates

Full size image

Figure 4 shows the 12-month SPEI slope map on the VHI for all four periods of the study. The slope map of March (Fig. 4a) shows the highest positive slope line that the southern hillside of Zagros, Persian Gulf margin, southeast, and northeastern part of Iran have. In other words, increasing or decreasing of the 12-month SPEI will have a more significant effect on the amount of VHI. Also, the areas around Zagros, Alborz, east, and northeast have the most negative slope coefficient. In April, the Persian Gulf coast and the Gulf of Oman, parts of the center and the northeastern part, parts of the southeast, and east of Iran have the highest slope coefficient of positive effect. But, the areas around the Zagros and Alborz mountains, the southern margin of the Alborz, and the Caspian Sea coast have the most slope coefficient of adverse impact. In May, the highest slope coefficient of the positive impact is related to the southern regions of Iran, while the northern parts have the most slope coefficient of adverse effect. In general, the slope of the effect reduced in May compared with previous periods. In June, the slope coefficient of effect increased compared with the last month. The margins of the Zagros Mountains and the northeast regions were the most positive, while the southern areas of Alborz and eastern Iran had the most negative slope coefficient.

Fig. 4
figure 4

The slope of the regression line between SPEI and VHI during the study period: a March, b April, c May, and d June

Full size image

Figure 5 shows the graph of the effect of average slope coefficient of regression of 12-month SPEI on VHI in different climates in different months. In March, the average slope coefficient of the regression line was positive in all climates except in wet climates, indicating that in super dry, dry, semi-arid, and semi-humid climates, on average, in most areas, drought affected plant health. This effect was more severe (Slop March, arid = 3.6) in the dry climate than in other climates. In April, the average slope coefficient of the regression line (Slop April, arid = 2) was positive only in the dry climate. However, in May, the average slope coefficient of the regression line was negative in all climates, although it was close to zero in the dry climate. Also in May, it was positive in all months, and in dry climates, this amount was more than in other climates (Slop June, arid = 4.4), which showed a more significant effect of 12-month SPEI on VHI. In general, in all months, the average slope coefficient of the regression line was more pronounced in the dry climate and less in wet climates than in other climates.

Fig. 5
figure 5

Average slope coefficient of the regression line of the 12-month SPEI effect on VHI in different climates

Full size image

4 Discussion

In this research, the association between meteorological and agricultural droughts was studied in the four periods of late March, April, May, and June from 2000 to 2017. In these periods, most of the natural vegetation cover is at the peak of vitality in Iran. The correlation results of the 12-month SPEI and VHI showed that the effect of meteorological drought on vegetation varied with time and place. This result was consistent with the results of Ma’rufah et al. (2017). Also, the results showed that the correlation between 12 -month SPEI and VHI increased from March to June when the air got warmer; this increase was occurring in dry and hyper-dry climates. In other words, the temperature increase had a direct relationship with the 12-month SPEI correlation intensity and VHI. The reason is that as the temperature increased, the water requirement of the plants grew. These results are consistent with the results of Zhongbao et al. (2008) and Jiang et al. (2017).

As a result, in the drought period and water scarcity, plants tend to be tense with higher temperatures and this tension is directly related to the severity of drought. Zheng et al. (2018) showed that the sizeable increase in temperature would likely lead to a negative impact on vegetation during the growing season, as it would also increase the water demand for vegetation growth that matches with our results. It can also be said that agricultural drought is directly related to meteorological drought (Samadi Neghab 2007; Araste et al. 2017). Dry and hyper-dry areas are more sensitive than other areas due to the lack of moisture.

Also, in the four periods, the highest correlation was observed in wet and dry climatic classes. Probably, the reason is that in arid areas due to the presence of deserts and the semi-arid and semi-humid regions because of high mountains, there were uncovered areas or areas with low vegetation, which reduced the mean correlation coefficient between the 12-month SPEI on VHI.

Results of the regression line slope of 12-month SPEI on the VHI also indicated that the heights had a slope of adverse effect, which was probably due to the low temperature of these areas. As we moved from March to June, with the warming of the air, the negativity of the slope line decreased. Also, from March to June, the average slope of the regression line increased. The highest average slope of the regression line was related to the dry climate, and the lowest was related to the wet climate. In other words, changing the severity of drought in arid areas had a more significant effect on vegetation health than had other climatological classes. As a result, plants were more susceptible to drought in desert areas. In general, it can be said that in June, the dry climate class had the highest correlation and effect on the regression line slope.

In general, in elevated west and northwest, there was a negative correlation between vegetation cover and drought intensity during all four periods of the study. It was because plants in this area have a different phonological cycle in comparison with other regions and they are more resistant against cold weather. Xu et al. (2018) and Sun et al. (2016) also had achieved similar results. The variation of correlation strength in these areas is under the influence of slope and direction. These two factors influence the intensity and duration of receiving sunshine and consequently caused to change the temperature.

In general, it should be noted that the role of temperature was also undeniable in the increasing correlation and impact on the regression line slope, but the function of temperature varied in different topographical and climatic conditions. Pei et al. (2018) concluded that the relationship between temperature and vegetation cover was influenced by climate and topography.

5 Conclusion

Results of this study illustrate that climate and topography have an essential role in how drought can affect vegetation cover. The most positive correlation between drought intensity and vegetation cover was for dry and wet climates. Also, results show that the slope effect for a drought on vegetation most of the time is negative in the deserts of Iran. However, in general, the slope effect has a positive correlation with temperature, and it increases when the temperature goes up.

Generally, results showed that the meteorological drought had different effects on agricultural drought in time and space. This effect has a direct relation with temperature. It should be mentioned that only 12-month SPEI was considered for meteorological drought in this research and the use of this index at different time scales may have different results. Also, the VHI index was assessed in months that vegetation had the best condition. So it is possible that vegetation had a different reaction to the meteorological drought in other months. Another point is that NDVI and LST indices are the basis for assessing the VHI index. Most researches show that NDVI is not an appropriate index for evaluating vegetation in arid and semi-arid regions, so it is better to use other indices for estimating vegetation condition in these regions.

The results of assessing drought indices at different time scales and vegetation indices in a different time and space scales can be useful in monitoring the temperature impact on the relationship between meteorological and agricultural droughts.