Meteorological Drought Assessment in the Ziz Watershed (South East of Morocco)

In the face of climate change, drought monitoring and characterisation is indispensable in arid and semi-arid areas. This paper represents a metrological drought assessment in the Ziz watershed located in south-eastern Morocco, through the application of four meteorological indices: rainfall anomaly index (RAI), percentage of normal index (PNI), standardized precipitation index (SPI), z-score index (ZSI). Based on rainfall data series from four stations over a time scale of 37 years (from 1983), also taking into account the relationship between these indices. The analysis of the rainfall series indicates a variability in its spatio-temporal distribution. However, the calculated drought indices indicate that the Ziz basin has experienced drought periods with their different classifications between 1983–2019, noting that the years 1986 and 2001 were dry to extremely dry in all stations of the basin. Thus, the results highlight a strong correlation between the four indices (r ≥ 0.967).


INTRODUCTION
It is clear that the water scarcity is a concern in significant regions of the world. Indeed, this problem relies upon two crucial factors: hydro climatology and the socio-economic situation (El Moçayd et al., 2020; Wada et al., 2014). Consequently, water scarcity is reflected in 66% of the world's population suffering from it (Cuenca et al., in press;El Hafyani et al., 2020). Nevertheless, the Mediterranean regions are the most threatened by this environmental problem due to several factors, namely: The severe natural variation between years, furthermore, the seasonal water resources and the declined stream flows in expected contemplated in the forthcoming years. (El Hafyani et al., 2020; Martinez and Poole, 2004).
Nonetheless, these critical conditions of water scarcity are linked by drought. Hence, Stagna defined drought as the period and state of water supply resulting from the negative difference between water inflow and outflow in the hydrological system, thus it can cause imbalances and malfunctions of reversible or irreversible character (Stângă, 2009). In case where the precipitation deficit is the only considered, then is considered as a meteorological drought (Boken et al., 2005;Keyantash and Dracup, 2002). Arid and semi-arid areas are the most likely to be threatened by this type of drought due to their climatic characteristics which are reflected in the rainfall deficit. In addition, climate change increases the risks related to drought, due to a strong increase in temperature that contributes Meteorological Drought Assessment in the Ziz Watershed (South East of Morocco) to amplified evaporation, which leads to the consolidation of the intensity and duration of drought. Therefore, the occurrence and adverse effects of drought are likely to accelerate land degradation and desertification in the long term (OMM and GWP, 2014). Not only the environment, but also the impact of droughts could also affect the economic and social sectors (Layelmam, 2008), on account of the importance of water used for the production of goods and services (Stour and Agoumi, 2008).
Morocco by its geographical location is characterized by a high spatial variability of the climate (El Moçayd et al., 2020), which might vary from sub humid to arid on the country (Born et al. 2008), as well as an uneven distribution of precipitation in the space (El Moçayd et al., 2020). The characterization of climatic drought in Morocco throughout the last decades (Stour and Agoumi, 2008) has laid out a sturdy vigor of drought portrayed by a propensity reduction of precipitation, moreover, a notable warming of the climate equivalent to a noticeable jump of the evapotranspiration; all what have been referenced induces to an absolute crucial hydric deficit (Stour and Agoumi, 2008).
Within the similar environmental context, the demand for water among various user sectors is still increasing due to the demographic evolution, hence the rise of environmental problems and the severe impact of drought sequences (Jellali, 1997). In addition, climate change predictions envisaged that Morocco will experience dry weather in the future (El Moçayd et al., 2020; Giorgi and Lionello, 2008) and rainfall scarcity (El Moçayd et al., 2020; Patricola and Cook, 2010), as well as an expected change in their distribution and extreme events (Driouech et al., 2010).
Several previous studies on drought in Morocco have presented regional drought assessments through the application of meteorological indices.  (Daki et al., 2016). Another study carried out on the same basin and by applying SPI and other indices including rainfall deviation, deviation from average index, rainfall normal index and standardized rainfall index, the results showed dry spells in most of the 1980s and the first half of the 1990s, as well as in the early 2000s (Jouilil et al., 2013). In the south-east of Morocco precisely at the level of the Ziz catchment area the SPI showed that this catchment area experienced during the period 1982-2013 a drought severity (Mehdaoui et al., 2018).
To summarize, it is inevitable to implement a risk management policy intended to reduce the potential impacts of drought, yet first it is necessary to reinforce information on the characterization and analysis of drought sequences by the various useful materials excited for a given area over a given time interval. In this context, the present work is interested in characterizing and analyzing the sequences of meteorological drought in the Ziz catchment between 1983 and 2019. We will adopt the standardized precipitation index and other indices, namely: precipitation anomaly index (RAI), percentage of normal index (PNI), z-score index (ZSI), on a scale of 37 years in order to transmit precise information on the evolution of drought.

Study area
This study was carried out on the Ziz watershed, part of the large Ziz-Rhéris and Guir watershed (Errachdia region), located in the South East of Morocco. It covers an area of 13992 km², and reaches from the Rheris watershed to the West and the Guir watershed to the East (Fig. 1). Its main rivers come from the summits of the High Atlas, where they are supplied by springs emanating from the aquifers of the Jurassic limestone and by pluviometric contributions, including a snowy quantity. Consequently, they affirm in particular the impoundment of a large part of temporary tributaries (Hammada, 2007).

The adopted methodology
The meteorological indices were used to determine the intensity and duration of the drought, which allows us to define the current climatic characteristics of the study area. The first step required in completing this work was the collection of data from the Hydraulic Basin Agency of Guir Ziz and Rheris. These data are related to the monthly rainfall of the selected study area which is the watershed Ziz, precisely at Sidi Hamza, M'zizel, Erfoud and Taouz stations, during the period 1983-2019 (Fig. 1).
On the basis of these collected data, a series of calculations of meteorological indices (RAI, PNI, SPI, ZSI) was performed by the RDIT tool, which is a software that allows us to calculate eight meteorological drought indices from the precipitation data. Several indices are referred to in order to reinforce the information, and to study the compatibility of these indices through the results obtained.
Subsequently, two non-parametric tests, namely the Mann-Kendall test (Kendall, 1975;Mann, 1945) and the Sen slope (Sen, 1968), were applied to the calculated indices time series in order to determine the trend of the meteorological drought in the Ziz basin during the period 1983-2019. Figure 3 presents flowchart of the implemented methodology.

Rainfall anomaly index
The rainfall anomaly index was established by (Van Rooy, 1965) to identify positive and negative precipitation anomalies over a historical series, it can be calculated on a monthly or annual time scale. It is considered to be among the most efficient meteorological indices in terms of result, and the simplest in terms of its method of  calculation (Ndlovu and Demlie, 2020). The RAI is calculated using the following equations: where: P -the actual precipitation (mm); -the average precipitation of the historical series (mm); -respectively represent the average of the ten highest and lowest precipitation values of the historical series. The drought classifi cation based on the RAI is indicated in Table 2.

Percent of normal index
The percentage of normal index (Willeke et al., 1994) is used to describe the meteorological drought. Based on precipitation data, the PNI is calculated by a simple and rapid method, for various time scales. Indeed, it is the ratio of actual precipitation to normal precipitation multiplied by 100.
where: P i -the actual precipitation; P -the normal precipitation. Table 3 shows the different categories of drought according to PNI values.

Standardized precipitation index
The standardized precipitation index (McKee et al., 1993) is one of the most recognized meteorological indices for identifying and monitoring drought periods. Based on a long historical precipitation record, the SPI can be calculated for different time scales (1, 3, 6, 12, 24 and 48 months). The mathematical formula for SPI is as follows: where: P i -total precipitation of the period (i); P m -the mean precipitation of the period; σ -standard deviation.

Z-score index
Several researchers have admitted that the ZSI is as efficient as the SPI (Table 4), and that it can be calculated on several time scales (Jain et al., 2015). The ZSI does not require adjusting the precipitation data to the gamma distribution or the Pearson type III distribution. It is expressed by the following equation: where: P i -the precipitation of year or month (i) in (mm); P -the mean precipitation (mm); S -the standard deviation (mm).

Mann-Kendall (MK) trend test
The non-parametric Mann-Kendall test (Kendall, 1975;Mann, 1945) is used to detect the trend of a time series, particularly the series of meteorological and hydrological variables (Yagbasan et al., 2017). The Mann-Kendall S-test statistic is calculated using the following equation (Yagbasan et al., 2020): where: n is the number of data points, x i and x j represent the data in the timeseries i and j respectively.
The variance of the S-statistic is calculated as: where: P -the number of tied groups (a tied group refers a set of sample data and has the same value); t i -the number of ties of extent i.
If the sample size greater than 10, the Mann-Kendall Z value may be obtained by: According to the values of Z obtained, it can be deduced that there is a statistically significant trend if |Z| ≥ |Z| ≥ | 1−α/2 | in which case the null hypothesis is rejected and the alternative hypothesis is therefore accepted.

Sen's slope estimator
The non-parametric method of Sen slope (1968) used to determine the magnitude of the trend in a time series (Sharma and Goyal, 2020). The slope of Sen is calculated using the following formula: where: N -the number of data; x j and x k -the data at time j and k respectively.
An upward trend is indicated by a positive value of Sen's slope estimator and a downward trend is indicated by a negative value of the time series (Sharma and Goyal, 2020).

RESULTS
In general, during the period 1983-2019 the rainfall regime in the Ziz watershed is characterised by a variability in its spatiotemporal distribution, considering the coefficient of variation as well as the coefficient of immoderation which express this variability in the four stations (Table  5). In addition, the average annual rainfall takes a maximum value of 20.8 mm in the M'Zizel station and a minimum value of 0.38 mm in the Taouz station. Furthermore, the Fisher asymmetry coefficient is positive in the four stations, this indirectly reflects an interannual irregularity of precipitation in the Ziz watershed (Table 5).
From the series of rainfall data, we can determine the rupture period, which means the point of change of the average of the data (Ozer et al., 2014). For this, several methods have been implemented to detect this rupture, we mention the pettitt test (1979), the Bayesian method of Lee and Heghinian (1977), the statistic of Buishand (1982Buishand ( , 1984  Indeed, the annual rainfall regime varies from North to South of the watershed, where the S. Hamza station records the highest values of precipitation, while the Taouz station records the lowest values (Fig. 4), which leads to the altitude factor where precipitation increases as a function of altitude (Fig. 5) R² = 0.93.  Note: coefficient of variation -standard deviation/average; coefficient of immoderation -maximum value/minimum value; coefficient of asymmetry (Fisher) -the centered moment of order three/ cube of the standard deviation.

Identi cation of meteorological drought in the Ziz watershed
In this section, we will analyze the results of calculation of meteorological drought indices (RAI, PNI, SPI, ZSI) for each station separately, in order to identify the intensity and duration of droughts over 37 years.

Sidi Hamza station
The analysis of meteorological drought indices RAI, PNI in Sidi Hamza station reveals very dry years during the period 1983-1988 except for the year 1985 which was slightly wet. In contrast, during the same period the SPI, ZSI indices indicate only one dry year which was 1986. From the results of all calculated indices, the period 1989-2008 was marked by the dominance of normal years and also very wet years. Furthermore, the period 2009-2019 saw years of extremely dry compared to normal and wet years (Fig. 6).

M'zizel station
About the M'Zizel station, the results of meteorological drought indices RAI, PNI show a succession of dry periods and wet periods. Moreover, the period 1987-2011 is generally characterized by the dominance of normal years expressed by the results of SPI, ZSI (Fig. 7).  Thus, from 2006 the results of all meteorological indices indicate a succession of wet years which are more dominant than normal years (Fig. 8).

Taouz station
The analysis of meteorological drought indices at the Taouz station indicates a dominance of dry years during the period 1983-1992. Nevertheless, during the same period, the year 1989 is considered extremely wet. We also notice a remarkable drought in the years 1999, 2001 and 2010 expressed by the different calculated indices (Fig. 9).

Trend analysis of meteorological drought
To identify the meteorological drought trend in the Ziz watershed, two non-parametric tests were used; the Mann-Kendall test (Kendall, 1975;Mann, 1945) and the Sen slope (1968).
These two tests were applied to the time series of the four indices for the four stations. The results showed non-significant positive trends of meteorological dryness at the 5% level of significance for the series of the four indices in all stations of the Ziz catchment. Indeed, this trend is evidenced by the two p-values of Kendall's tau and Sen's slope (Table 6). However, these results reflect the succession of dry and wet years in the whole catchment during the period 1983-2019 Frequency and spatial distribution of drought and moisture conditions from the SPI during the period (1983-2019) The SPI results of four stations were interpolated by the IDW method using ArcGIS software in order to analyse the spatiotemporal distribution of this index. Indeed, the SPI index varies from year to year throughout the watershed Ziz (Fig. 10), where we notice some years drier and others wetter, however the years 1986 and 2001 were marked by a drought that covered almost all of the watershed. Thus, the year 1999 was very dry for the southern part of the watershed, notably the stations Erfoud and Taouz. While, in 2012, 2013 and 2017 the drought was recorded just at the northern of the watershed (Sidi Hamza and M'Zizel). We also note that the wettest conditions prevailed in 1989, 2006 and 2008.

Spatial distribution of drought and moisture conditions from RAI, PNI, SPI and ZSI indices
The four meteorological drought indices indicate drought in the majority of the watershed    (Fig. 11). In 1986, the drought covered the whole basin except for the Erfoud station which was characterised by normal conditions. However, in 2001 the RAI, PNI and SPI show a more intense drought in the southern stations compared to the northern stations. On the other hand, the ZSI in its distribution indicates only one category of drought (moderate drought) in the whole basin except for the M'Zizel station which was marked by a normal climate also expressed by the SPI. In contrast, the wettest conditions prevailed in 1989, 2006 and 2008. All four indices indicate very wet to extremely wet conditions at all four stations in 1989 (Fig. 12). The stations Erfoud and Taouz showed respectively in 2006 and 2008 the least intense conditions compared to the other stations (Fig. 12).

Correlation of drought indices
The Pearson correlation (Pearson, 1896) is used to determine the strength and direction of the linear relation between two variables following the normal distribution. Otherwise, the relation between the two variables can be estimated by the Kendall (1938) or Spearman (1904 a, b) test. In this study we used the Pearson correlation since the values of the calculated indices follow a normal distribution.
The Pearson correlation matrix (Table 7) shows a positive and strong correlation between . Cont. Spatial and temporal distribution of RAI, PNI, SPI, ZSI in the basin during the driest years (1983,2001) all the indices calculated in the four stations (r ≥ 0.967), thus the correlation coeffi cient between the two indices ZSI and PNI is still equal to 0.999, which ensures a strong relation between them.

DI SCUSSION
Pr ecipitation, as a major factor in the identifi cation of meteorological drought, requires analysis on a spatial and temporal scale to monitor its variability. On an overview of the whole of Morocco, the spatio-temporal distribution of TRMM precipitation varies according to season, proximity to the sea and altitude (Ezzine et al., 2014), However, this study has highlighted a spatio-temporal variability of precipitation at the level of the Ziz watershed, the spatial variability translated by the altitude factor, precipitation belonging to the eastern part of the central high Atlas increases from south to north. It can be noted that in the arid and semi-arid regions the areas characterized by low altitude are more vulnerable to drought because of their rainfall character.
Although the drought indices are based on precipitation, it is evident that the spatiotemporal variability is not limited to precipitation but is also refl ected in the results of the drought indices obtained over the whole basin.   On the other hand, a similar study by (Diani et al., 2019) found that the annual precipitation in the high Ziz basin had no significant trend. In the same context the results of present work also indicate an insignificant trend of meteorological drought over the whole study area, indicated by the four indices. The correlation between these indices is strong precisely between ZSI and PNI (r= 0.999). Indeed, this result corroborates with other studies, (Katipoğlu et al., 2020) and (Di̇ki̇ci̇ and Aksel, 2021) showing high correlation for metrological drought indices of the same time periods. Thus (Ekwezuo and Madu, 2020; Salehnia et al., 2017;Keyantash and Dracup, 2002), who conducted studies based on meteorological drought indices showed a high correlation between them.

CONCLUSION
Within the context of drought monitoring in arid and semi-arid areas, the present work was conducted to determine the drought characteristics in the Ziz watershed located in south-eastern Morocco using four meteorological drought indices.
The results showed a biennial irregularity of precipitation that varies according to the altitude factor, which is reflected by a rainfall trend increasing from the South to the North of the watershed. The drought indices applied in this study showed a strong correlation between them (r ≥ 0.967). Therefore, the assessment of drought by these metrological indices indicated that the Ziz River basin experienced alternating dry and wet periods. Thus, in most years the distribution of drought was not homogeneous in the basin. According to this result, adaptive water management strategies to climate change should be further strengthened, in order to reduce the impact of drought periods and ensure the sustainability of water resources throughout Morocco, especially in the arid and semi-arid environments, the most sensitive to climate change and the adverse effects of drought.