Trend analysis of temperature and precipitation in Trarza region of Mauritania

Global climatic change has attracted the attention of many scientists in recent years. Studies of the effect of climate change are mostly monitored by using air temperature and precipitation trends rather than hydrologic variables. Many studies have provided evidence that temperature and precipitation increased in recent years. New et al. (2001) collected precipitation data from a number of countries to identify trends in precipitation and they concluded that daily precipitation had increased during the twentieth century. A study by Jones et al. (1999) on surface air temperature and its change over 150 years showed that the global temperature during the periods 1925–1944 and 1978–1997 rose by 0.37 °C and 0.32 °C, respectively. Lettenmaier et al. (1994) analyzed the hydro-climatological trends in the continental United States for 1948–1988 and they found increasing trends at 50% of the stations for temperature in March, as well as precipitation which increased from September to December at 25% of the stations. Chen et al. (2009) investigated the historical trends of meteorological drought in Taiwan by means of long-term precipitation records, and they detected positive trend in the daily precipitation time series for dry days all over Taiwan.

Mauritania receives very little rainfall at any time of year. The southernmost part of the country reaches the semi-arid region called the Sahel. Mauritania has a wet season between July and September, with up to 200 mm of rainfall per month. The mean temperatures are highest in the southern part of the country and lowest in the north for most of the year. Since the mid-1960s, long-term drought has killed most of the livestock and forced nomad populations to resettle, creating new villages or enlarging small cities in Mauritania (Arnaud 1999). Many other inhabitants moved towards the capital Nouakchott where the population increased from 6000 in 1962 to 760,000 in 2001 (Salama et al. 1991; Niang et al. 2008). However, there are few studies of trends of meteorological variables, such as precipitation and temperature, in Mauritania. MacSweeney et al. (2008) used a consistent approach for 52 developed countries to produce an analysis of climate data. Their results showed that the mean annual temperature in Mauritania has increased by 0.90 °C since 1960, with an average rate of 0.19 °C per decade. The mean annual rainfall over Mauritania has not changed with any consistent trend since 1960. Some unusually high rainfalls have occurred in very recent years (2000–2006), which were not part of a consistent trend analysis in the study of MacSweeney et al. (2008).

The historical hydro-climatic precipitation and temperature recorded data are generally used for planning and designing water resources projects. The changes in these meteorological variables in recent years have triggered research to find effective trend identification tests to investigate and analyze regime changes in the time series data. Many of these trend tests are classified into parametric and non-parametric methods. Parametric trend tests are more powerful than non-parametric ones (Shadmani et al. 2012). However, the parametric trend tests require data to be normally distributed. In contrast, the non-parametric trend tests do not assume any specific distribution.

The non-parametric Mann–Kendall test (MK test) and Spearman's rho test (SR test) are commonly used for detecting trend of hydro-meteorological variables (Gellens 2000; Garbrecht et al. 2004; Kahya & Kalaycı 2004; Shadmani et al. 2012). The reason for using these two nonparametric rank-based statistical tests is that they have an ability to identify the monotonic trends (upward and downward trends) in the time series data. These tests can deal with non-normal data, missing values, seasonality, censoring (detection limits), and serial dependence (Hirsch & Slack 1984). However, the power of these methods has not been well documented. Yue et al. (2002) investigated the power of these two tests and concluded that their power (decrease or increase) depends on different variables including slope of the trend, sample size, pre-assigned significance level, and the variation of the time series.

Şen (2012) provided a new trend analysis method, which is able to identify the trend in the time series especially in terms of low, medium, and high values of the data. This method is valid whatever the sample size, serial correlation structure of the time, and non-normal probability distribution functions (Şen 2012).

Many studies use the Theil–Sen approach which was described by Hirsch et al. (1982) for estimating absolute values for slope (Yue et al. 2002; Gallego et al. 2006; Chen & Grasby 2009; Shadmani et al. 2012; Some'e et al. 2012). Pettitt's test (Pettitt 1979) is used to detect change points in time series (Tomozeiu et al. 2000; Mu et al. 2007; Zhang et al. 2008; Gebremicael et al. 2013).

In this study, 44 years of meteorological records (collected at three stations) were used to investigate trends in precipitation and temperature time series in Trarza region in Mauritania. The MK test, the SR test, and the Şen trend test were used for trend identification in the time series and the Pettitt test for detecting the change point at the time series. The Theil–Sen approach was used to estimate the magnitude of the slope in the precipitation and temperature time series.

Trarza is a region in the southwest of Mauritania (Figure 1) with total area of 67,800 km². It is one of the most populated regions in Mauritania. The water resources in Trarza region support a range of uses from urban water supply to agriculture. The source of the groundwater is the Trarza aquifer covering about 40,000 km². This aquifer plays a very important role as the main ground source of water for Mauritania. Senegal River (located in the west of the region) is Mauritania's only permanent waterway and is used for irrigation, transport, drinking, and hydropower generation.

Long-term precipitation and temperature historical records (recorded monthly for 44 years) were used to investigate the temperature and precipitation trends for the period of 1970 to 2013 in Trarza region in Mauritania. The data used were collected from the National Office of Meteorology of Mauritania (Office National de la Météorologie). Three rainfall stations, namely Boutilimit (station 1), Nouakchott (station 2), and Rosso (station 3), as shown in Figure 1, were used in this study. One station (Nouakchott) is outside the catchment area; however, it was considered in this study since its location is very close to the study area. Table 1 summarizes the latitude, longitude, and altitude for each station.

The maximum annual rainfall reaches 311 mm in 2003 and 226 mm in 1995 in stations 1 and 2, respectively, while station 3, which is located in the Senegal River Valley region where the rainfall is higher than other regions, behaves differently with maximum rainfall of 477.8 mm in 2010. Table 2 summarize rainfall characteristics for each station.

Annual minimum temperature (T-min), maximum temperature (T-max), and average temperature (T-average) for 1970–2013 were used in this study. These data are summarized in Table 3. Station 3 showed the maximum temperature recorded with 39.21 °C in 1987 when it was the warmest year detected at stations 1 and 2 with average of 30.28 °C and 30.48 °C, respectively. The minimum temperature in this region was detected in 1971 at station 2 with 18.46 °C. Overall, 1974 was the coldest year in the region with the average temperature showing minimum values at stations 1, 2, and 3 of 27.46 °C, 25.06 °C, and 27.07 °C, respectively.

Trend analysis is considered as one of the most important issues in global climate change problems. The main purpose of the trend analysis in this study is to provide a view of change in meteorological (precipitation and temperature) variables. For this purpose, 44 years of meteorological records were used to investigate trends in the precipitation and temperature time series. The MK, SR, and Şen trend tests were used for trend identification in the time series. The Pettitt test was used to detect the change point in the time series (Pettitt 1979) and the Thiel–Sen approach was used to estimate the magnitude of the slope in the precipitation and temperature time series.

This method is based on dividing the time series of the data in to two equal parts, ranking them from the highest to the lowest and then plotting data points against each other in which the first sub-series is located on the x-axis, and the other sub-series is located on the y-axis on the Cartesian coordinate system. If the data points are clustered on the 1:1 (45°) straight line there is no trend; if the data points are in the triangular area below the 1:1 straight line there is a decreasing trend in the time series; and if the data points are in triangular area above the 1:1 straight line, there is increasing trend in the time series. The high, medium, and low values of the given data can be graphically evaluated in this approach (Şen 2012).

Maximum, minimum, and average annual temperature trends were investigated using different trend tests: the MK, SR, Şen, and Pettitt's tests. The MK, SR, and Şen tests were used for identifying trends while Pettitt's test was used for detecting the change point in the temperature time series. Almost all the test results show an increasing trend for the three stations (see Table 4 and Figures 2–4).

The MK and the SR tests show the same results for the three stations where they revealed significant trends in the minimum and average temperature in all the stations. The significant maximum temperature trends were observed at Nouakchott and Rosso stations, while no trend was observed in the maximum temperature records at Boutilimit, as shown in Table 4. After the trends had been identified by the MK and SR tests, the Theil–Sen approach was applied to estimate the magnitude of the slope (change per unit time). It showed that the annual average temperature increased at Boutilimit, Noakchott, and Rosso at a rate of 0.2, 0.3, and 0.3 °C per decade (Table 4). Abrupt changes in the annual temperature (maximum, minimum, and average) were detected by Pettitt's test for the results for all stations (Table 4). Significant abrupt change was detected in 1995.

The Şen trend test was used to investigate the trends especially in terms of the evaluation of low, medium, and high values of the temperature data. The results show significant increasing trends for the low values at all stations, compared to the high and average values which are closer to the 1:1 line (Figures 2–4). Maximum temperatures recorded at Boutilimit show different results: the medium values have almost no trend, the high values have decreasing trend, and low values have increasing trend.

The three rain-gauge stations of Boutilimit, Nouakchott, and Rosso were used for analyzing the annual precipitation time series. The MK and SR tests showed the same result, both detecting significant increasing trends for Boutilimit and Rosso while no trend was seen for Nouakchott. Using the Theil–Sen method, the annual precipitation was found to increase at a rate of (+)3.35 and (+)2.93 mm per year at Boutilimit and Rosso stations, respectively (Table 5). The change points in Boutilimit and Rosso stations were found to be 1988 and 1993, respectively (Table 5).

Unlike the MK and the SR tests, the Şen trend test shows significant increasing trends for all stations for the annual precipitation time series, with the same trend magnitude (slope) (Figure 5).

Trends in temperature and precipitation for the period 1970–2013 were analyzed for Trarza region using different types of trend detection methods. The results showed that there were positive trends in precipitation and temperature during the period studied. Although all stations have slight increasing precipitation and temperature trends, they differ in terms of trend direction changes (slope and jumps).

The annual regional temperature time series analysis indicates that the changes in temperature over 1970–2013 reflect warming of the region as a whole. The annual average temperature increased at a rate of 0.2, 0.3, and 0.3 °C per decade at Boutilimit, Nouakchott, and Rosso stations, respectively. This increase of average temperature is affected by the minimum more than the maximum temperatures as seen in Table 4. The maximum values increased at the rate of 0.2 and 0.3 °C per decade for stations of Nouakchott and Rosso, respectively, and no trend in the maximum temperature was detected in station of Boutilimit. While the minimum temperatures have significant positive trends detected at all stations, the magnitudes of the increasing trend for the minimum temperatures for stations Boutilimit, Nouakchott, and Rosso were found to be 0.2, 0.4, and 0.3 °C per decade, respectively. The Şen trend test showed that the low values of temperatures (maximum, minimum, and average) increased more than the medium and high values during the study period of the (1970–2013). The abrupt change in the annual temperatures (maximum, minimum, and average) was detected in 1995 for all stations and this means that the region had significant jump in the trend direction in that year. The magnitudes in two periods: 1970–1995 and 1995–2013 were calculated and the results showed the rise of the average temperature in the first period (1970–1995) for Nouakchott and Rosso; in contrast at Boutilimit the increase of the temperature was detected in the second period (1995–2013) (see Table 5).

Analyzing annual precipitation time series by MK and SR tests showed that two stations (Boutilimit and Rosso) have significant positive trends while no trend was detected for Nouakchott station. However, the Şen test showed different results by detecting a positive trend for station Nouakchott, but in terms of the medium values of precipitation during 1970–2013. The results of the Theil–Sen test showed that the most significant positive trend was detected at Rosso station with value of 3.35 mm per year while the precipitation for Boutilimit station increased at 2.93 mm per year. Abrupt changes in annual precipitation were detected in 1988 for Boutilimit and in 1993 for Rosso.

Based on historical precipitation and temperature data recorded for 1970–2013 from three stations in Trarza region in Mauritania, this study investigated the performances of five trend tests for the precipitation and temperature time series. The following conclusions were drawn:

• 1.

  The annual average temperature time series analysis indicates that the changes in temperature over 1970–2013 reflect warming for the region as a whole.

• 2.

  The annual temperature is increasing about 0.3 °C per decade.

• 3.

  Significant increasing trends for the low values of temperature in all the stations are occurring, compared to the high and the medium values.

• 4.

  The abrupt change in temperature was detected in 1995 for the whole region.

• 5.

  The precipitation time series has positive trend detected for two stations (at the rate of 3.35 mm per year in Boutilimit and 2.93 mm per year in Rosso) while no trend is detected in the other (Nouakchott).

• 6.

  Significant increasing trend is revealed for all the values (high, medium, and low), with almost the same trend magnitude (slope).

As a result of hotter climate and erratic rainfall, much of Mauritania has become progressively more unsuitable for agriculture and villages. Irregular rain has ruined crops, and the country is experiencing both droughts and floods making arable land unsuitable for cultivation. More than 5,600 people countrywide have been affected by the floods, with over 2,000 people having to evacuate their homes in 2013, according to the International Federation of Red Cross and Red Crescent Societies (IFRC). However, Mauritania has been experiencing drought since the late 1960s; Trarza region experienced almost 8 drought years during 1970–2013 (Yacoub & Tayfur 2016). Therefore, the government should work on suitable adaptation strategies for agricultural system with the existing climate conditions. This study would be a guide and help in that direction.