**1. Introduction**

Drought phenomenon has been considered as one of the major natural hazards causing numerous economic, social, and environmental losses [1]. In the context of ongoing global warming, an increase in frequency and intensity of this kind of extreme event is expected [2–4], and therefore, its accurate characterization is of high relevance for both mitigation and adaptation. In broad terms, drought is usually characterized as the scarcity of precipitation over a prolonged time period, but it is difficult to define and identify due to the great number of variables involved as well as the variety of sectors affected. Moreover, the characterization of the spatiotemporal

patterns of drought is very complex since it is very variable in space and time, this being particularly true in transitional zones such as the Iberian Peninsula (IP) [5].

In recent years, drought indices have been commonly used to identify, analyze, and monitor the occurrence of droughts. As indicated in [6], drought indices are variables based on climate information (e.g., precipitation, evapotranspiration, soil moisture, or runoff). They are used to analyze the effects of drought, allowing the definition of different drought characteristics (i.e., the duration and severity of droughts as well as the spatial extent). However, their accuracy strongly depends on long-term consistent climate data, and unfortunately, spatially and temporally regular climate observations are rare. In this context, the regional climate models (RCMs) are valuable tools providing climate information at an adequate spatiotemporal resolution to characterize regional drought patterns. In fact, drought phenomena are spatially complex, so detailed spatial scales are required in the study of droughts [7]. Additionally, Abatzoglou et al. [8] investigated the sensibility of drought indicators to the spatial resolution and found that the indices computed with the highest resolution explained over 10% more variability than those from coarser datasets.

Among different drought indices developed in recent years, the Standardized Precipitation Evapotranspiration Index (SPEI) [9] was proposed as an alternative to the Standardized Precipitation Index (SPI) [10]. Contrariwise to the SPI, the SPEI takes into account the effect of temperature for detecting droughts, and therefore, it seems to be more accurate in the context of global warming [11]. In fact, the increased global temperature trend is expected to increase the atmospheric evaporative demand, so regions, where the precipitation is normal in a given period (or even higher than normal), could be considered to suffer droughts. For instance, the role of the temperature is clear for the event occurred during summer 2003, which had devastating effects in central Europe mainly because of the anomalous temperatures in this period [9]. Furthermore, taking into account temperature data, this drought indicator has shown better performance than others based solely on precipitation (e.g., the SPI) for detecting droughts during summer, when the related impacts may become stronger [11].

This work investigates spatiotemporal patterns of drought through the SPEI to understand the drought temporal behavior over the IP, identifying those periods especially accused. This is of major relevance because the IP is characterized by highly variable and scarce precipitation leading to recurrent drought occurrences [12]. For this purpose, a 35-year climate simulation has been completed using the Weather Research and Forecasting (WRF) model [13] with the purpose of obtaining high-resolution climate data.

The chapter is structured as follows: Section 2 is devoted to detail the WRF model configuration as well as the statistical method applied to characterize the spatiotemporal patterns of droughts over the study region. Section 3 presents the main results regarding the temporal evolution of drought for the 35-year period (1980–2014), the identification of drought characteristics as well as the analysis of the relationships of the SPEI with other drought-related variables, such as the surface (actual) evapotranspiration, the soil moisture content, and the runoff. Finally, Section 4 summarizes the main conclusions obtained in this study.

#### **2. Methodology**

#### **2.1 Climate input data**

Hydroclimate variables at high spatiotemporal resolution were obtained by using the WRF-ARW model version 3.6.1. This RCM was run to simulate current climate characteristics over the IP and Balearic Islands (**Figure 1a**). The WRF model

**131**

**Figure 1.**

*Understanding the Drought Phenomenon in the Iberian Peninsula*

was driven by the ECMWF ERA-Interim reanalysis data [14]. **Table 1** summarizes the main model setup used in this study. This configuration has been successfully used to represent the spatiotemporal patterns of droughts in the study region [15]. In [15], the authors analyzed the added value of using downscaled climate data to detect drought through the comparison with observational data and found that

*(a) WRF domains: the EURO-CORDEX region (d01) at 0.44° of spatial resolution and the IP region (d02)* 

*with a spatial resolution of 0.088°, and (b) the main river basins over the IP.*

WRF provided a benefit with respect to its driving data in this regard.

*DOI: http://dx.doi.org/10.5772/intechopen.85472*

### *Understanding the Drought Phenomenon in the Iberian Peninsula DOI: http://dx.doi.org/10.5772/intechopen.85472*

*Drought - Detection and Solutions*

ing high-resolution climate data.

patterns of drought is very complex since it is very variable in space and time, this being particularly true in transitional zones such as the Iberian Peninsula (IP) [5]. In recent years, drought indices have been commonly used to identify, analyze, and monitor the occurrence of droughts. As indicated in [6], drought indices are variables based on climate information (e.g., precipitation, evapotranspiration, soil moisture, or runoff). They are used to analyze the effects of drought, allowing the definition of different drought characteristics (i.e., the duration and severity of droughts as well as the spatial extent). However, their accuracy strongly depends on long-term consistent climate data, and unfortunately, spatially and temporally regular climate observations are rare. In this context, the regional climate models (RCMs) are valuable tools providing climate information at an adequate spatiotemporal resolution to characterize regional drought patterns. In fact, drought phenomena are spatially complex, so detailed spatial scales are required in the study of droughts [7]. Additionally, Abatzoglou et al. [8] investigated the sensibility of drought indicators to the spatial resolution and found that the indices computed with the highest resolu-

tion explained over 10% more variability than those from coarser datasets.

Among different drought indices developed in recent years, the Standardized Precipitation Evapotranspiration Index (SPEI) [9] was proposed as an alternative to the Standardized Precipitation Index (SPI) [10]. Contrariwise to the SPI, the SPEI takes into account the effect of temperature for detecting droughts, and therefore, it seems to be more accurate in the context of global warming [11]. In fact, the increased global temperature trend is expected to increase the atmospheric evaporative demand, so regions, where the precipitation is normal in a given period (or even higher than normal), could be considered to suffer droughts. For instance, the role of the temperature is clear for the event occurred during summer 2003, which had devastating effects in central Europe mainly because of the anomalous temperatures in this period [9]. Furthermore, taking into account temperature data, this drought indicator has shown better performance than others based solely on precipitation (e.g., the SPI) for detecting droughts during summer, when the related impacts may become stronger [11]. This work investigates spatiotemporal patterns of drought through the SPEI to understand the drought temporal behavior over the IP, identifying those periods especially accused. This is of major relevance because the IP is characterized by highly variable and scarce precipitation leading to recurrent drought occurrences [12]. For this purpose, a 35-year climate simulation has been completed using the Weather Research and Forecasting (WRF) model [13] with the purpose of obtain-

The chapter is structured as follows: Section 2 is devoted to detail the WRF model configuration as well as the statistical method applied to characterize the spatiotemporal patterns of droughts over the study region. Section 3 presents the main results regarding the temporal evolution of drought for the 35-year period (1980–2014), the identification of drought characteristics as well as the analysis of the relationships of the SPEI with other drought-related variables, such as the surface (actual) evapotranspiration, the soil moisture content, and the runoff. Finally,

Hydroclimate variables at high spatiotemporal resolution were obtained by using the WRF-ARW model version 3.6.1. This RCM was run to simulate current climate characteristics over the IP and Balearic Islands (**Figure 1a**). The WRF model

Section 4 summarizes the main conclusions obtained in this study.

**130**

**2. Methodology**

**2.1 Climate input data**

was driven by the ECMWF ERA-Interim reanalysis data [14]. **Table 1** summarizes the main model setup used in this study. This configuration has been successfully used to represent the spatiotemporal patterns of droughts in the study region [15]. In [15], the authors analyzed the added value of using downscaled climate data to detect drought through the comparison with observational data and found that WRF provided a benefit with respect to its driving data in this regard.

#### **Figure 1.**

*(a) WRF domains: the EURO-CORDEX region (d01) at 0.44° of spatial resolution and the IP region (d02) with a spatial resolution of 0.088°, and (b) the main river basins over the IP.*


#### **Table 1.**

*The main model configuration.*

In this study, the 3-hourly outputs in their native grid resolution have been used as climatic input data to analyze drought occurrences. In this way, the outputs of the WRF model at 10 km of spatial resolution (i.e., those from the inner domain d02, **Figure 1a**) were temporally aggregated at monthly scale, obtaining thus gridded longterm monthly climate data over land for the entire study region, for the period 1980– 2014. The monthly variables used here are the averaged maximum and minimum temperature (Tmax and Tmin, respectively), the accumulated precipitation (pr), the accumulated surface evapotranspiration (SFCEVP), the mean soil moisture contained in the upper 1 m (SM), and the runoff. The WRF model, through the coupled Noah LSM scheme, contemplates two different runoff components, the surface and subsurface runoff. Here, the runoff was understood as the sum of these two components.

### **2.2 Defining drought occurrences over the Iberian Peninsula**

## *2.2.1 The Standardized Precipitation Evapotranspiration Index (SPEI)*

The SPEI has been widely used in recent years to characterize droughts, showing a good ability to detect, monitor, and analyze drought events [22–24]. This drought indicator is based on the SPI and differs in that it uses a simple "climatic water balance" instead of precipitation data in determining droughts. The temperature effect is indeed considered through the difference between the precipitation and the reference evapotranspiration (ET0), this being aggregated at a range of time scales (1–48 months). Then, the aggregated climatic water balance is fitted to a statistical distribution, and subsequently, its probability density function is transformed into a random normal variable Z with mean of 0 and variance of 1. This latter assumption enables the comparison across regions by determining the probability of occurrence of a given drought event (**Table 2**). Thus, Z is the corresponding SPEI and indicates the number of standard deviations from the climatological mean. Therefore, and similar to the SPI, the SPEI is statistically robust, simple to compute and easily interpretable. Even more, the option to compute it at different time aggregations makes this index suitable for assessing different drought types (i.e., meteorological, agricultural, or hydrological droughts) [9].

In this study, the SPEI was computed at two different time aggregations: the 3- and 12-month time scales (hereinafter SPEI-03 and SPEI-12, respectively) using the SPEI R-package [25]. These time scales were chosen with the purpose of characterizing

**133**

**Table 2.**

terms of absolute value of the SPEI.

*2.2.2 Temporal evolution of drought-related variables*

*Understanding the Drought Phenomenon in the Iberian Peninsula*

agricultural and hydrological droughts [6], respectively. Temporal series of the climatic water balance were obtained for the entire IP using the pr and the Tmax and Tmin from WRF at monthly scale. To approximate the ET0, the Hargreaves equation [26] was used. Studies such as [27] evidenced that the Hargreaves method is able to approximate the ET0 showing similar results to the Penman-Monteith equation [28] with the advantage that only temperature data are required to calculate it. The climatic water balance, here, is assumed to follow a log-logistic distribution [27]. This distribution has been used in many studies (e.g., [10, 27]) to fit the SPEI, mostly because it is a three-parameter distribution (i.e., negative values are permitted), which has been proved to better estimate the climatic water balance [27].

**SPEI value Category Probability (%)** SPEI ≥2 Extremely wet 2.3 2 ≥ SPEI >1.5 Severely wet 4.4 1.5 ≥ SPEI >1 Moderately wet 9.2 1 ≥ SPEI >0.5 Mildly wet 15.0 0.5 ≥ SPEI > −0.5 Nearly normal 38.2 −0.5 ≥ SPEI > −1.0 Mild drought 15.0 −1 ≥ SPEI > −1.5 Moderate drought 9.2 −1.5 ≥ SPEI > −2 Severe drought 4.4 SPEI < −2 Extreme drought 2.3

Then, the temporal series of SPEI at both 3- and 12-month time scales for each grid point were spatially averaged obtaining the SPEI evolution for each main river basin over the IP. Twelve main river basins were considered, as the result of aggregating smaller watersheds. They are North Atlantic (NA; composed by the Galician coast, the Western Cantabrian, and the Eastern Cantabrian watersheds), Miño-Sil (MS; the Miño-Sil, the Cávado, the Ave, and the Leça watersheds), Duero (DU), Ebro (EB), Northeastern Basins (NE), Portugal Basins (PB; the Vouga, Mondego, Lis, and Ribeiras do Oeste), Tajo (TJ), Southeastern Basins (SE), Guadiana (GU; Guadiana, Sado, Mira, and Ribeiras do Algarve), Guadalquivir (GQ; Guadalquivir, Tinto, Odiel, Piedras, Guadalete, and

In order to explore the severity of droughts, different drought characteristics have also been analyzed. For this, the established condition is that a dry event is occurring when at least two consecutive months are under the defined drought conditions. Therefore, a dry event was considered to begin when the SPEI falls below zero, and it ends when recovering positive values. Moreover, such a dry period is defined as drought if at least 1 month within the period reaches mild drought conditions (i.e., SPEI below −0.5 [29]). Then, the duration of droughts is understood as the number of months in each drought event. The intensity is the averaged value of the SPEI within the period expressed in absolute value. Additionally, the severity of the event (or minimum value reached) was also explored, which is expressed in

Time series for each river basin from other drought-related variables were also examined. In this regard, the SM and SFCEVP were examined because they directly

Barbate), Southern Basins (SB), and the Balearic Islands (BI) (**Figure 1b**).

*DOI: http://dx.doi.org/10.5772/intechopen.85472*

*Drought categories and their probability of occurrence.*

*Understanding the Drought Phenomenon in the Iberian Peninsula DOI: http://dx.doi.org/10.5772/intechopen.85472*


**Table 2.**

*Drought - Detection and Solutions*

Spatial configuration

zone

**Table 1.**

Nudging and sponge

*See [15] for more details.*

*The main model configuration.*

**Parameters Description**

Temporal period Period 1980–2014 with 11 months of spin-up

Physics schemes Microphysics: WRF single-moment-3-class [16]

Vertical layer 41 vertical levels set the top of the atmosphere at 10 hPa

the planetary boundary (PBL)

Convection: Betts-Miller-Janjic [17, 18] PBL: convective asymmetric model version 2 [19]

Land surface model: Noah LSM [20]

In this study, the 3-hourly outputs in their native grid resolution have been used as climatic input data to analyze drought occurrences. In this way, the outputs of the WRF model at 10 km of spatial resolution (i.e., those from the inner domain d02, **Figure 1a**) were temporally aggregated at monthly scale, obtaining thus gridded longterm monthly climate data over land for the entire study region, for the period 1980– 2014. The monthly variables used here are the averaged maximum and minimum temperature (Tmax and Tmin, respectively), the accumulated precipitation (pr), the accumulated surface evapotranspiration (SFCEVP), the mean soil moisture contained in the upper 1 m (SM), and the runoff. The WRF model, through the coupled Noah LSM scheme, contemplates two different runoff components, the surface and subsurface runoff. Here, the runoff was understood as the sum of these two components.

Sponge zone: the 10 outer grid points of each domain

Radiation: community atmosphere model 3.0 [21]

Two "one-way" nested domains: (d01) EURO-CORDEX region at 0.44° (~50 km) of spatial resolution, and (d02) the IP at a spatial resolution of 0.088° (~10 km)

Spectral nudging for waves above 600 km, only over the coarser domain and above

The SPEI has been widely used in recent years to characterize droughts, showing a good ability to detect, monitor, and analyze drought events [22–24]. This drought indicator is based on the SPI and differs in that it uses a simple "climatic water balance" instead of precipitation data in determining droughts. The temperature effect is indeed considered through the difference between the precipitation and the reference evapotranspiration (ET0), this being aggregated at a range of time scales (1–48 months). Then, the aggregated climatic water balance is fitted to a statistical distribution, and subsequently, its probability density function is transformed into a random normal variable Z with mean of 0 and variance of 1. This latter assumption enables the comparison across regions by determining the probability of occurrence of a given drought event (**Table 2**). Thus, Z is the corresponding SPEI and indicates the number of standard deviations from the climatological mean. Therefore, and similar to the SPI, the SPEI is statistically robust, simple to compute and easily interpretable. Even more, the option to compute it at different time aggregations makes this index suitable for assessing different drought types (i.e.,

In this study, the SPEI was computed at two different time aggregations: the 3- and 12-month time scales (hereinafter SPEI-03 and SPEI-12, respectively) using the SPEI R-package [25]. These time scales were chosen with the purpose of characterizing

**2.2 Defining drought occurrences over the Iberian Peninsula**

meteorological, agricultural, or hydrological droughts) [9].

*2.2.1 The Standardized Precipitation Evapotranspiration Index (SPEI)*

**132**

*Drought categories and their probability of occurrence.*

agricultural and hydrological droughts [6], respectively. Temporal series of the climatic water balance were obtained for the entire IP using the pr and the Tmax and Tmin from WRF at monthly scale. To approximate the ET0, the Hargreaves equation [26] was used. Studies such as [27] evidenced that the Hargreaves method is able to approximate the ET0 showing similar results to the Penman-Monteith equation [28] with the advantage that only temperature data are required to calculate it. The climatic water balance, here, is assumed to follow a log-logistic distribution [27]. This distribution has been used in many studies (e.g., [10, 27]) to fit the SPEI, mostly because it is a three-parameter distribution (i.e., negative values are permitted), which has been proved to better estimate the climatic water balance [27].

Then, the temporal series of SPEI at both 3- and 12-month time scales for each grid point were spatially averaged obtaining the SPEI evolution for each main river basin over the IP. Twelve main river basins were considered, as the result of aggregating smaller watersheds. They are North Atlantic (NA; composed by the Galician coast, the Western Cantabrian, and the Eastern Cantabrian watersheds), Miño-Sil (MS; the Miño-Sil, the Cávado, the Ave, and the Leça watersheds), Duero (DU), Ebro (EB), Northeastern Basins (NE), Portugal Basins (PB; the Vouga, Mondego, Lis, and Ribeiras do Oeste), Tajo (TJ), Southeastern Basins (SE), Guadiana (GU; Guadiana, Sado, Mira, and Ribeiras do Algarve), Guadalquivir (GQ; Guadalquivir, Tinto, Odiel, Piedras, Guadalete, and Barbate), Southern Basins (SB), and the Balearic Islands (BI) (**Figure 1b**).

In order to explore the severity of droughts, different drought characteristics have also been analyzed. For this, the established condition is that a dry event is occurring when at least two consecutive months are under the defined drought conditions. Therefore, a dry event was considered to begin when the SPEI falls below zero, and it ends when recovering positive values. Moreover, such a dry period is defined as drought if at least 1 month within the period reaches mild drought conditions (i.e., SPEI below −0.5 [29]). Then, the duration of droughts is understood as the number of months in each drought event. The intensity is the averaged value of the SPEI within the period expressed in absolute value. Additionally, the severity of the event (or minimum value reached) was also explored, which is expressed in terms of absolute value of the SPEI.

#### *2.2.2 Temporal evolution of drought-related variables*

Time series for each river basin from other drought-related variables were also examined. In this regard, the SM and SFCEVP were examined because they directly affect the agricultural droughts. In fact, soil water availability is the major driver of the plant transpiration. To do this, the monthly time series of SM and SFCEVP were used to compute the standardized anomalies of such variables. These anomalies were computed with two purposes: (1) to remove seasonality and (2) to make the temporal evolutions comparable across regions. The SFCEVP was previously accumulated using 3-month time slices to also compare it with the SPEI-03, which is the time scale here used to characterize agricultural droughts.

On the other hand, to further investigate hydrological droughts, the temporal series of runoff, previously aggregated at annual scale, were used to compute the standardized anomalies series. This variable allows us to incorporate hydrological processes in drought characterization [6]. Hydrological droughts are developed slowly and persist longer than other forms of drought [30], so relationships between temporal series of runoff and the SPEI-12 must occur.

The relationship between such drought-related variables with the SPEI was investigated by computing temporal correlation coefficients for all river basins. Additionally, a t-test at the 95% confidence level was used to determine the significance of the correlation coefficients, previously considering the effect of the serial correlation by following the methodology proposed by Bretherton et al. [31].

### **3. Spatiotemporal patterns of droughts**

#### **3.1 Temporal evolution of the SPEI**

This study begins with the analysis of the temporal evolution of the SPEI-03 and the SPEI-12 for all river basins over the entire study period (**Figure 2**). In the interest of clarity, SPEI evolution is represented according to the different drought categories (see **Table 2**). Concerning the SPEI-03 evolution (**Figure 2a**), the results showed a high temporal variability, which is a characteristic of meteorological droughts. That is, the SPEI changes frequently between drought and wet conditions. Furthermore, these results evidence the complexity of drought phenomena and therefore its characterization since very different conditions may be found in the different watersheds in a given moment (i.e., while certain river basins are affected by drought conditions, others presented normal or wet conditions). Several drought events were recorded over the study period. For instance, moderate to extreme drought conditions appeared during the years 1985, 1994–1995, 2005, and 2012 across a large part of the IP. Contrariwise, the years 1984, 1996–1998, 2003–2004 2010, and 2013 showed wet conditions over a large part of the IP.

Looking at the longest time scale (**Figure 2b**), the results presented a less variable behavior. In fact, for 12 months, the SPEI is less sensitive to variations from a given month, and hence, the changes between wet and dry conditions are less frequent and these are also longer. Moreover, the results display that, in general, dry and wet conditions are allocated in the same periods than those from the 3-month time scale, with the most severe droughts happening during 1995, 1999, 2005, and 2012. For these years, many watersheds were affected by severe drought conditions (i.e., SPEI below −2). Conversely, wet periods also appeared over many river basins during the years 1984, 1997, 2010, and 2013. Note the extreme wet conditions occurring during 2001 in the PB and the MS river basins.

#### **3.2 Climatological characteristics of droughts**

In order to further explore drought phenomenon, the duration distributions of drought events were explored for all river basins. **Figure 3** displays the box

**135**

**Figure 2.**

*Islands (BI).*

NE, and PB river basins).

*Understanding the Drought Phenomenon in the Iberian Peninsula*

plots of the duration of droughts for both the SPEI-03 and the SPEI-12 and for all river basins. This representation shows the scores of duration (in months) as well as their distributional characteristics. At 3 months (**Figure 3a**), the median was between 3 and 4 months, and the mean was around 4 months. Therefore, the mean was slightly higher than the median in all river basins, showing that the distributions are skewed to the right. The 97.5th percentile (limits of the upper whiskers) ranges from about 7.5 to 14.5 months, with the shortest and the longest durations occurring for the NE and the GQ river basins, respectively. On the other hand, the spread of the distribution is a signal of high variability in drought events, and it is marked by the size of the box (i.e., the interquartile range). In this regard, southern river basins (i.e., the TJ, SE, GU, GQ, and SB river basins) showed larger interquartile range than the northern ones (i.e., the NA, MS, DU,

*Temporal evolution of the SPEI in all river basins between 1980 and 2014 at (a) 3- and (b) 12-month time scales. The color code is established according to the drought categories. Nomenclature: North Atlantic (NA), Miño-Sil (MS), Duero (DU), Ebro (EB), Northeastern Basins (NE), Portugal Basins (PB), Tajo (TJ), Southeastern Basins (SE), Guadiana (GU), Guadalquivir (GQ ), Southern Basins (SB), and the Balearic* 

For the SPEI-12 (**Figure 3b**), longer durations appeared for the period 1980– 2014 in general, showing also more spread in their distributions. The median was between 5 and 14 months, reaching the maximum values again over the southern river basins. The mean duration of droughts was around 12 months, indicating that the mean was higher than the median in many cases. The 97.5th percentile was between 12 and 53 months with the highest one over the BI river basin. The major variability in terms of interquartile range seems to appear again over the BI river

*DOI: http://dx.doi.org/10.5772/intechopen.85472*

### *Understanding the Drought Phenomenon in the Iberian Peninsula DOI: http://dx.doi.org/10.5772/intechopen.85472*

#### **Figure 2.**

*Drought - Detection and Solutions*

affect the agricultural droughts. In fact, soil water availability is the major driver of the plant transpiration. To do this, the monthly time series of SM and SFCEVP were used to compute the standardized anomalies of such variables. These anomalies were computed with two purposes: (1) to remove seasonality and (2) to make the temporal evolutions comparable across regions. The SFCEVP was previously accumulated using 3-month time slices to also compare it with the SPEI-03, which is

On the other hand, to further investigate hydrological droughts, the temporal series of runoff, previously aggregated at annual scale, were used to compute the standardized anomalies series. This variable allows us to incorporate hydrological processes in drought characterization [6]. Hydrological droughts are developed slowly and persist longer than other forms of drought [30], so relationships between

The relationship between such drought-related variables with the SPEI was investigated by computing temporal correlation coefficients for all river basins. Additionally, a t-test at the 95% confidence level was used to determine the significance of the correlation coefficients, previously considering the effect of the serial correlation by following the methodology proposed by Bretherton et al. [31].

This study begins with the analysis of the temporal evolution of the SPEI-03 and the SPEI-12 for all river basins over the entire study period (**Figure 2**). In the interest of clarity, SPEI evolution is represented according to the different drought categories (see **Table 2**). Concerning the SPEI-03 evolution (**Figure 2a**), the results showed a high temporal variability, which is a characteristic of meteorological droughts. That is, the SPEI changes frequently between drought and wet conditions. Furthermore, these results evidence the complexity of drought phenomena and therefore its characterization since very different conditions may be found in the different watersheds in a given moment (i.e., while certain river basins are affected by drought conditions, others presented normal or wet conditions). Several drought events were recorded over the study period. For instance, moderate to extreme drought conditions appeared during the years 1985, 1994–1995, 2005, and 2012 across a large part of the IP. Contrariwise, the years 1984, 1996–1998, 2003–2004

Looking at the longest time scale (**Figure 2b**), the results presented a less variable behavior. In fact, for 12 months, the SPEI is less sensitive to variations from a given month, and hence, the changes between wet and dry conditions are less frequent and these are also longer. Moreover, the results display that, in general, dry and wet conditions are allocated in the same periods than those from the 3-month time scale, with the most severe droughts happening during 1995, 1999, 2005, and 2012. For these years, many watersheds were affected by severe drought conditions (i.e., SPEI below −2). Conversely, wet periods also appeared over many river basins during the years 1984, 1997, 2010, and 2013. Note the extreme wet conditions occur-

In order to further explore drought phenomenon, the duration distributions of drought events were explored for all river basins. **Figure 3** displays the box

2010, and 2013 showed wet conditions over a large part of the IP.

ring during 2001 in the PB and the MS river basins.

**3.2 Climatological characteristics of droughts**

the time scale here used to characterize agricultural droughts.

temporal series of runoff and the SPEI-12 must occur.

**3. Spatiotemporal patterns of droughts**

**3.1 Temporal evolution of the SPEI**

**134**

*Temporal evolution of the SPEI in all river basins between 1980 and 2014 at (a) 3- and (b) 12-month time scales. The color code is established according to the drought categories. Nomenclature: North Atlantic (NA), Miño-Sil (MS), Duero (DU), Ebro (EB), Northeastern Basins (NE), Portugal Basins (PB), Tajo (TJ), Southeastern Basins (SE), Guadiana (GU), Guadalquivir (GQ ), Southern Basins (SB), and the Balearic Islands (BI).*

plots of the duration of droughts for both the SPEI-03 and the SPEI-12 and for all river basins. This representation shows the scores of duration (in months) as well as their distributional characteristics. At 3 months (**Figure 3a**), the median was between 3 and 4 months, and the mean was around 4 months. Therefore, the mean was slightly higher than the median in all river basins, showing that the distributions are skewed to the right. The 97.5th percentile (limits of the upper whiskers) ranges from about 7.5 to 14.5 months, with the shortest and the longest durations occurring for the NE and the GQ river basins, respectively. On the other hand, the spread of the distribution is a signal of high variability in drought events, and it is marked by the size of the box (i.e., the interquartile range). In this regard, southern river basins (i.e., the TJ, SE, GU, GQ, and SB river basins) showed larger interquartile range than the northern ones (i.e., the NA, MS, DU, NE, and PB river basins).

For the SPEI-12 (**Figure 3b**), longer durations appeared for the period 1980– 2014 in general, showing also more spread in their distributions. The median was between 5 and 14 months, reaching the maximum values again over the southern river basins. The mean duration of droughts was around 12 months, indicating that the mean was higher than the median in many cases. The 97.5th percentile was between 12 and 53 months with the highest one over the BI river basin. The major variability in terms of interquartile range seems to appear again over the BI river

#### **Figure 3.**

*Box plots of duration of drought events for all river basins, (a) for the SPEI-03 and (b) for the SPEI-12. The lower and upper parts of the boxes represent the 25th and the 75th percentiles, respectively; the line in the middle of each box is the median and the upper and lower whiskers indicate the 2.5th and 97.5th percentiles, respectively. The mean is displayed with black dots.*

basin, which presented a large difference between the median and the third quartile (a difference of about 27 months). This latter feature reveals that most of the events for this period were short, but a longer event also occurred. The same behavior also appears in the SE river basins, but here the distribution is more homogenous, showing shorter events.

The box plots for the intensity of the drought events were also examined. In fact, the effects of drought phenomenon are stronger for longer events, but they are also the results of strong intensities. Hence, the distribution of the intensity has to be analyzed. In general, both time scales presented distributions with a low spread, particularly for those events from the SPEI-12. For the 3-month time scale (**Figure 4a**), the median was around 1 for all river basins, indicating that the events were, on average, moderate. As for duration, the mean was slightly higher than the median, reaching values of around 1.1. Concerning the extreme values, the 97.5th percentile was around 1.5 in general and reached a maximum above 1.7 (severe drought) over the EB and the PB river basins. Slightly lower medians are presented for events computed at 12 months (**Figure 4b**), where median intensities rise to values between 0.86 and 1.17 (i.e., from mild to moderate droughts). The mean values were around 1, and these were slightly higher than the median values in most of them. Some basins as MS, EB, or PB presented the opposite behavior. Concerning the extreme values in terms of intensity, the 97.5th percentile was also slightly lower than for the previous time scale (values around 1.45), reaching values above 1.5 over the GQ , the SB, and the PB river basins.

In addition, to examine trends of drought occurrences along the study period, the duration, intensity, and severity of drought events were analyzed by computing such parameters in two different periods: 1980–1999 and 2000–2014. Then, the median changes between both periods in the three parameters were used to develop a categorical classification.

**137**

**Figure 5.**

**Figure 4.**

*respectively. The mean is displayed with black dots.*

*Understanding the Drought Phenomenon in the Iberian Peninsula*

**Figure 5** displays the trends for the main river basins and for both time scales. At 3 months (left map), different behaviors were shown. The DU river basin has a positive trend in the three parameters, and the EB presented the opposite behavior. Intermediate trends appeared in the rest of the IP, showing watersheds where a positive trend occurred only in the median severity (i.e., the MS and the SE river basins). In other watersheds, the positive trend only appeared in the median intensity (i.e., the NE, the BI, and GQ river basins), but there were also watersheds with an increase in two of the characteristics analyzed. However, for the 12-month time scale (right

*Box plots of intensity of drought events for all river basins, (a) for the SPEI-03 and (b) for the SPEI-12. The lower and upper parts of the boxes represent the 25th and the 75th percentiles, respectively; the line in the middle of each box is the median and the upper and lower whiskers indicate the 2.5th and 97.5th percentiles,* 

*Categorical classification of droughts for the SPEI-03 (left) and the SPEI-12 (right) based on the changes in* 

*severity (ΔS), intensity (ΔI), and duration (ΔD) between the periods 1980–1999 and 2000–2014.*

*DOI: http://dx.doi.org/10.5772/intechopen.85472*

*Understanding the Drought Phenomenon in the Iberian Peninsula DOI: http://dx.doi.org/10.5772/intechopen.85472*

**Figure 5** displays the trends for the main river basins and for both time scales. At 3 months (left map), different behaviors were shown. The DU river basin has a positive trend in the three parameters, and the EB presented the opposite behavior. Intermediate trends appeared in the rest of the IP, showing watersheds where a positive trend occurred only in the median severity (i.e., the MS and the SE river basins). In other watersheds, the positive trend only appeared in the median intensity (i.e., the NE, the BI, and GQ river basins), but there were also watersheds with an increase in two of the characteristics analyzed. However, for the 12-month time scale (right

#### **Figure 4.**

*Drought - Detection and Solutions*

basin, which presented a large difference between the median and the third quartile (a difference of about 27 months). This latter feature reveals that most of the events for this period were short, but a longer event also occurred. The same behavior also appears in the SE river basins, but here the distribution is more homogenous, show-

*Box plots of duration of drought events for all river basins, (a) for the SPEI-03 and (b) for the SPEI-12. The lower and upper parts of the boxes represent the 25th and the 75th percentiles, respectively; the line in the middle of each box is the median and the upper and lower whiskers indicate the 2.5th and 97.5th percentiles,* 

The box plots for the intensity of the drought events were also examined. In fact, the effects of drought phenomenon are stronger for longer events, but they are also the results of strong intensities. Hence, the distribution of the intensity has to be analyzed. In general, both time scales presented distributions with a low spread, particularly for those events from the SPEI-12. For the 3-month time scale (**Figure 4a**), the median was around 1 for all river basins, indicating that the events were, on average, moderate. As for duration, the mean was slightly higher than the median, reaching values of around 1.1. Concerning the extreme values, the 97.5th percentile was around 1.5 in general and reached a maximum above 1.7 (severe drought) over the EB and the PB river basins. Slightly lower medians are presented for events computed at 12 months (**Figure 4b**), where median intensities rise to values between 0.86 and 1.17 (i.e., from mild to moderate droughts). The mean values were around 1, and these were slightly higher than the median values in most of them. Some basins as MS, EB, or PB presented the opposite behavior. Concerning the extreme values in terms of intensity, the 97.5th percentile was also slightly lower than for the previous time scale (values around 1.45), reaching values above 1.5 over the GQ , the SB, and the PB river

In addition, to examine trends of drought occurrences along the study period, the duration, intensity, and severity of drought events were analyzed by computing such parameters in two different periods: 1980–1999 and 2000–2014. Then, the median changes between both periods in the three parameters were used to develop

**136**

basins.

a categorical classification.

ing shorter events.

*respectively. The mean is displayed with black dots.*

**Figure 3.**

*Box plots of intensity of drought events for all river basins, (a) for the SPEI-03 and (b) for the SPEI-12. The lower and upper parts of the boxes represent the 25th and the 75th percentiles, respectively; the line in the middle of each box is the median and the upper and lower whiskers indicate the 2.5th and 97.5th percentiles, respectively. The mean is displayed with black dots.*

#### **Figure 5.**

*Categorical classification of droughts for the SPEI-03 (left) and the SPEI-12 (right) based on the changes in severity (ΔS), intensity (ΔI), and duration (ΔD) between the periods 1980–1999 and 2000–2014.*

map), the positive trends are less frequent, showing an increased trend in more than one parameter only for the PB, the DU, and the EB river basins. In these three basins, the results showed an increase in both severity and duration of drought events.
