Identifying drought- and flood-prone areas based on significant changes in daily precipitation over Iran

Abstract

Variations in frequency and intensity of extreme events have substantial impact on water resources and environment, which in turn are reflected on agriculture, society, and economy. We assessed spatiotemporal changes in pattern of daily precipitation to identify drought- and flood-prone areas of Iran. To do this, we generated gridded daily precipitation for the period of 1962–2013 over Iran using measured daily precipitation and the Kriging approach. We applied 11 precipitation indices that were stated by the Expert Team on Climate Change Detection and Indices (ETCCDI) to identify significant changes in frequency and intensity of extreme precipitation events. According to significant changes of these 11 precipitation indices, drought- and flood-prone areas of Iran were, then, detected. We observed significant changes in pattern of daily precipitation over more than half of the country. 40% of the country, which were located in the elevated regions of Iran, particularly along Zagros Mountain, was identified as flood-prone areas. As a result, in these regions, there is a need for flood risk management based on changes in stormwater events such as runoff generated from rain on snow and snowmelt events. In addition, we detected drought-prone areas in large portion of the northwest of Iran and in the low elevated regions of the country that have semiarid or arid climate. This suggests that it is necessary to prepare a long-term drought plan to mitigate impacts of severe drought events.

Appendix

1.1 Validation of the gridded precipitation product

We calculated the coefficient of correlation (R) and the bias between the observed precipitation at several weather stations and interpolated precipitation at a grid cell that the station of interest is located (Eqs. 10, 11). This method was used to test the reliability of a gridded precipitation product in prior works (e.g., Feidas 2010; Kidd et al. 2012; Darand et al. 2017).

$$R = \frac{{\sum\nolimits_{i = 1}^{n} {(o_{i} - \overline{o} )(p_{i} - \overline{p} )} }}{{\sqrt {\sum\nolimits_{i = 1}^{n} {(o_{i} - \overline{o} )^{2} } \sqrt {\sum\nolimits_{i = 1}^{n} {(p_{i} - \overline{p} )^{2} } } } }}$$

(10)

$${\text{Bias}} = \frac{{\sum\nolimits_{i = 1}^{n} {(p_{i} - o_{i} )} }}{n}$$

(11)

where o is the observed precipitation from stations, $\bar{o}$ is the mean value of the observed value, pi is the gridded precipitation product, $\bar{p}$ is the mean value of the product value, i is the index of the station number and n is the number of stations (49 stations). We observed R values greater than 0.9 at all stations and biases ranged from − 9.2–4.4 mm per month (Fig. 6).

Fig. 6

Spatial distribution of the coefficient of correlation (a) and the bias (mm month⁻¹) (b), which were calculated between observed precipitation at weather stations and interpolated precipitation of a grid cell that the station of interest is located

1.2 Homogeneity of precipitation data

We used the cumulative deviation test to identify homogeneity and breakpoint in the precipitation time series. In the cumulative deviation test (Buishand 1982), the departure from homogeneity is tested using the statistic Q and R, which are defined as:

$$Q = \mathop {\hbox{max} }\limits_{0 \le k \le n} \left| {{\text{S}}_{\text{k}}^{ * *} } \right|$$

(12)

and

$$R = \mathop {\hbox{max} }\limits_{0 \le k \le n} ( {\text{S}}_{\text{k}}^{ * *} ) - \mathop { \hbox{min} }\limits_{0 \le k \le n} ( {\text{S}}_{\text{k}}^{ * *} )$$

(13)

We divided the cumulative deviations from the mean, ${\text{S}}_{\text{k}}^{ *}$, by the sample standard deviation, D_x, to obtain the rescaled adjusted partial sums, ${\text{S}}_{\text{k}}^{ * *}$:

$$S_{k}^{**} = \frac{{S_{k}^{*} }}{{D_{x} }},\quad k = 1,2, \ldots ,n$$

(14)

where

$$D_{x} = \sum\limits_{i = 1}^{n} {\frac{{(x_{i} - \overline{x} )^{2} }}{n}}$$

(15)

$${\text{S}}_{ 0}^{ *} {\text{ = 0, S}}_{\text{n}}^{ *} {\,= 0}$$

$${\text{S}}_{\text{k}}^{ *} = \sum\limits_{i = 1}^{k} {(x_{i} - \overline{x} )} ,\quad k = 1,2, \ldots ,n - 1$$

(16)

$$\overline{\text{x}} = \frac{1}{n}\sum\limits_{i = 1}^{n} {x_{i} }$$

(17)

High values of Q and R indicate departure from homogeneity. Critical values of $Q/\sqrt n$ and $R/\sqrt n$ obtained by Buishand (1982), are given in Table 3.

Table 3 Critical values of $Q/\sqrt n$ and ${\text{R}}/\sqrt n$ for cumulative deviation test.

Figure 7 indicates the results of the test for all grid points. Over 35% of the country had significant temporal inhomogeneity at 95% confidence level. We observed breakpoints over 38% of the country during period 1994–2004. Only 4.8% of the study area experienced the breakpoints during 2004–2013.

Fig. 7

Temporal homogeneity of precipitation data by cumulative deviation test (CDT). Colors refer to observed temporal (year) breakpoints. Black dots show significant temporal inhomogeneity at 95% confidence level