3.1. Validation of Satellite Temperature and Rainfall Products
The validation results obtained from the scatter plots between stations and the CHIRPS rainfall dataset revealed correlation coefficient (r = 0.97), the lowest root mean squared error (RMSE = 16.22 mm/month), and an overall bias of 0.90 at α = 0.05 significance level (Figure 2a). This positive correlation indicates that when rainfall readings from ground stations increase, so do the corresponding estimates from the CHIRPS dataset. The CHIRPS dataset effectively captures rainfall patterns, as evidenced by a statistically significant relationship at α = 0.05 level. The study found a strong correlation between CHIRPS rainfall data and observed station data, particularly at lower rainfall values in Ethiopia, indicating the CHIRPS dataset’s exceptional reliability in capturing spatial and temporal rainfall variability across the investigated stations [47]. The CHIRPS dataset, with a low RMSE, accurately replicates recorded rainfall levels, indicating minimal discrepancies between observed and anticipated values, making it a valuable resource for hydrological and climatic research [48]. Although the CHIRPS dataset somewhat underestimates rainfall, this minor error is acceptable and illustrates the dataset’s excellent accuracy for practical applications [49]. These validation findings show that the CHIRPS dataset is strong and accurate at generating reliable rainfall predictions.
The scatter plots comparing station data with the ERA5 monthly mean maximum temperature dataset showed a correlation coefficient of r = 0.84, RMSE of 1.73 °C, and an overall bias of 0.99 (Figure 2b) at α = 0.05 significance level. The correlation coefficient for the monthly mean lowest temperature was r = 0.86, with an RMSE of 1.50 °C and an overall bias of 0.97 (Figure 2c). This positive correlation suggests that when temperature readings from ground stations rise, the estimates from the ERA5 dataset also increase correspondingly. The ERA5 dataset accurately measures temperature fluctuations, as demonstrated by a statistically significant association at α = 0.05 level. These findings support ERA5’s accuracy in depicting temperature patterns, making it an important tool for climate study and historical observation [50]. This strong correlation validates the reliability of CHIRPS and ERA5 datasets were preferred for this research due to their accuracy in representing climate variables across Ethiopia (Figure 2).
3.2. Spatial and Temporal Annual Analysis of the Extreme Temperature
The trend analysis of eight temperature indices across 103 stations from 1994 to 2023 reveals the following percentages, categorized by trend strength and statistical significance at the α = 0.05 level: significant positive trends account for 7.9%, non-significant positive trends comprise 43.2%, significant negative trends represent 9.2%, non-significant negative trends constitute 38.1%, and no trend was observed in 1.6% of the cases, as summarized (Table 2). The analysis of the maximum value of daily maximum temperature (TXx) shows a positive average trend over the study period. However, the annual time series shows that only 14.56% of the observed data had a statistically significant trend, with 13.59% positive and 0.97% negative trends. Furthermore, 83.50% of the stations showed statistically insignificant changes, with 45.63% of them positive and 37.86% negative. In examining spatial distributions of yearly time series for the daily maximum temperature maximum value (TXx), it was found that 59.22% of stations experienced an increase in maximum values. Notably, this included a significant trend of 13.59% in northeastern Ethiopia, whereas the central and north regions of the research area exhibited a non-significant trend of 45.63% (Figure 3a). During the study period, the highest annual mean maximum temperature (TXx) was recorded 35.5 °C in 2005, while the lowest was 32.3 °C in 1996. The annual TXx increased by 0.011 °C per year (Figure 4a), which aligns with findings from other studies [51]. This increase in TXx severe occurrences is mostly caused by oscillations in global sea surface temperatures, which influence meteorological conditions and shape extreme heat events [52].
In contrast, distinct regional patterns emerged in the geographical distribution of the minimum value of the daily maximum temperature (TXn), which tended to fall. Notably, TXn exhibited a non-significant decrease at 51.46% of stations in the northeastern and southeastern regions. Conversely, 47.57% of stations showed relatively minor increases, predominantly in the northern and southwestern areas (Figure 3b). This result is similar to that of a study in the Upper Blue Nile Basin, which identified an increasing trend in TXn, with 69.2% of stations reporting heightened intensity in temperature indices [53]. Because of the diverse local climatic conditions and varying geographical features across Ethiopia, temperature trends exhibit significant regional variations. During the study period, the annual TXn in Ethiopia increased at a rate of 0.01 °C per year. The lowest annual mean TXn was 20.3 °C in 2015, while the highest was 25.7 °C in 2002 (Figure 4b). At most stations, this slight upward trend is not statistically significant and suggests a gradual warming in the minimum values of maximum temperatures.
The highest value of the daily minimum temperature (TNx) found that only 18.45% of the series had a statistically significant trend at α < 0.05, with 17.48% positive and 0.97% negative. In addition, 79.61% of the stations showed a statistically insignificant trend, with 60.19% of these trends positive and 19.42% negative (Table 2). The distribution of the highest daily minimum temperature (TNx) revealed that 77.67% of stations had a positive trend, mainly in the northeast and southeast areas. Notably, 17.48% of stations, including Abala, Aisha, Atsebi, Bure, Debrebirhan, Dire Dawa, Dollooddo, Elidar, Filtu, Harawa, Harghelle, Jijiga, Kelafo, Lmi, Moyale, Omorate Sheble, and Wegletena, showed a significant upward trend (Table S2). This result is consistent with the findings of the study Trends in Rainfall and Temperature Extremes in Ethiopia and Agro-Ecological Zone Levels of Analysis, which found an increase in TNx at 75.7% of the stations examined [54]. In addition, 60.19% of stations in northeastern and southern Ethiopia showed a statistically non-significant increase trend (Figure 3c). This indicates that the warmest nighttime temperatures are rising, indicating an increase in minimum temperatures at night, while both day and night temperatures are rising, with nighttime temperatures increasing at a faster rate [31]. During the study period, the annual TNx in Ethiopia increased at a rate of 0.007 °C per year. The lowest annual mean TNx was 18.5 °C in 1994, 1996, and 1997, while the highest was 20.5 °C in 1998 (Figure 4c). The distribution of TNn shows that 50.49% of stations had a positive, increasing trend. Only 2.91% of stations, including Moyale, Bure, and Aware, showed a substantial rise, while the remaining 47.57%, primarily in central and northeastern Ethiopia, showed nonsignificant increasing trend. In contrast, 46.60% of stations showed a decrease in TNn, with 2.91% significantly decreasing and 43.69% showing a statistically insignificant decline (Figure 3d). This positive TNn value suggested that the lowest temperatures were rising over time, pointing to an overall warming trend [53]. The minimum value of daily minimum temperature (TNn) trended lower by 0.003 °C every year. The lowest annual mean TNn was 9.9 °C in 2017; while the highest was 12.9 °C in 2002 (Figure 4d).
The spatial distribution of cold days (TX10p) revealed positive trends in 63.11% of stations, especially in central, northwest, and southern Ethiopia. Notably, 4.85% of stations, including Arbaminch, Hagermariyam, Jeba, Kibish, and Sherkole, showed a clear increase trend. Additionally, 58.25% of sites in northern, central, and southern Ethiopia showed nonsignificant rise. Meanwhile, 5.83% of stations, including Semera, Jimma, Elidar, Bure, Aisha, and Abala, had a significant downward trend (Figure 3e). These findings indicate that the frequency of extremely hot days per year is increasing, implying that hot extremes are growing faster than cold nights are decreasing [55]. The number of extreme temperature indices, such as TX10p, has decreased by 0.111 days per year. Notably, the lowest annual mean value was 1.09 days in 2004 and the highest was 34.2 days in 1996 (Figure 4e). On the other hand, 90.29% of stations showed a decline in TN10p prevalence across the study period, suggesting a significant change in temperature trends. TN10p significantly decreased at 50.49% of these locations, particularly in central and northeastern Ethiopia. This indicates a significant decrease in the frequency of days with minimum temperatures below the 10th percentile in these areas. A less severe but still discernible decline in TN10p was seen in 39.81% of stations, mostly in the southeast, where non-significant negative trends were recorded (Figure 3f). The overall prevalence of TN10p decreased, indicating a drop in the number of days with minimum temperatures below the 10th percentile, as well as a decrease in severely cold nights, which is consistent with [56]. Quantitatively, the research showed that over the study period, TN10p decreased at an average annual rate of 0.496 days per year. A notable decrease in exceptionally cold nights in recent years is indicated by the lowest annual average of TN10p, which was recorded at 2.6 days in 2023. To demonstrate the amount of change during the research period, the maximum annual average of TN10p was 32.8 days in 1999 (Figure 4f). This finding is consistent with previous studies that indicate a statistically significant negative trend in cold nights (TN10p), with declines ranging from 2.9 to 4.4 days per year [57]. The annual distribution of extreme warm nights (TN90p) found that 88.35% of stations experienced a rise, with warm nights being more common in the southeast. Of them, 66.02% indicated insignificant increases. In contrast, 22.33% of stations, particularly in the Somalia and Afar areas, showed significant upward trends (Figure 3g; Table S2). The TN90p index, which counts the number of nights with minimum temperatures over the 90th percentile, reflects this tendency. It was shown to increase by about 0.232 days every year, peaking at 26.5 days in 2019 (Figure 4g) [58,59]. The increased TN90p frequency suggests more frequent heat extremes and rising nighttime temperatures, as well as warmer nights. These patterns are in line with global climate change observations [60]. The annual distribution of DTR (Diurnal Temperature Range) was decreasing in 85.44% of stations. Of them, 72.82% showed insignificant negative trends in most locations, but 12.62% indicated significant negative trends, mainly in the Oromia and Somalia regions (Figure 3h; Table 2. The findings show that the lowest temperatures are rising faster than the maximum temperatures, resulting in a smaller diurnal temperature range. Previous research has demonstrated that increased cloud cover contributes to this trend by retaining heat at night, resulting in warmer evenings, but also reflecting sunlight during the day, resulting in cooler daytime temperatures [61,62]. Over the period of the study, the Diurnal Temperature Range (DTR) decreased by 0.01 °C per year, with significant drops seen in 1994–1999, 2006, 2010, 2013–2014, and 2016–2023 (Figure 4h). These results are consistent with findings from the study ‘The Long-term Trend in the Diurnal Temperature Range and Its Association with Total Cloud Cover and Rainfall’, which reported that between 1951 and 2014, the global land average diurnal temperature range (DTR) decreased by 0.054 °C per decade [63].
3.3. Spatial and Temporal Annual Analysis of the Extreme Precipitation
The results of annual trend analysis of eight precipitation indices at 103 stations from 1994 to 2023 revealed the following percentages, grouped by trend strength and statistical significance at α = 0.05 level; significant positive trend 6.8%, non-significant positive trend 43.1%, significant negative trend 5.8%, non-significant negative trend 42.6% and no trend 1.7% as summarized (Table 3). The spatial distribution of CDD revealed that 61.17% of stations had a negative, decreasing trend. Notably, 9.71% of stations, including Ambamariam, Assaita, Bati, Cheifra, Combolcha, Geladi, Majete, Mille, and Shahura, showed statistically significant downward trends at the α < 0.05 significance level (Table S3). Meanwhile, the remaining 51.46%, mostly found in northern, central, and southern Ethiopia, exhibited nonsignificant negative trends. In contrast, 36.89% of stations exhibited positive trends, with 32.04% showing insignificant positive trends. Notably, the Shebele and Ziway stations displayed statistically significant positive trends over the study period at a level of α < 0.05 (Figure 5a; Table S3). The analysis of continuous dry days (CDD) showed an insignificant increasing trend of 0.152 days annually. During the study period, the lowest annual mean for CDD was 13 days in 1998, and the highest was 78 days in 2019 (Figure 6a). Conversely, the number of continuous wet days (CWD) has been increasing at a rate of 1.115 days per year. The lowest annual mean for CWD was 53 days in 2009, and the highest was 169 days in 2019 (Figure 6b). This increasing trend in wet days indicates that the frequency and duration of sustained rainfall events are likely to rise [64,65]. Consecutive wet days (CWD) have exhibited notable trends in Ethiopia, with 73.79% of monitoring stations reporting increases. Among them, 16.50% displayed statistically significant positive trends at p < 0.05, primarily in the southeastern area. However, 57.28% showed non-significant positive trends, notably in central and northwest Ethiopia. In contrast, 25.24% of stations showed non-significant negative trends (p < 0.05) (Figure 5b). The increase in CWD combined with a decrease in consecutive dry days (CDD) suggests a probable increase in the number of rainy days across most climatic stations. This tendency indicated more continuous rainfall periods, which might improve water supply and boost agriculture. This pattern shows that dry spells are becoming shorter, while periods of continuous rainfall are becoming more common. This finding aligns with previous studies, which found that consecutive dry days (CDD) decreased while consecutive wet days (CWD) increased [46,66]. The factors contributing to the decrease in consecutive dry days (CDD) and the increase in consecutive wet days (CWD) are atmospheric circulation patterns, the seasonal migration of the Intertropical Convergence Zone (ITCZ), and complex topography [67].
The spatial distribution patterns of annual total precipitation (PRCPTOT) indicated a positive trend in 72.82% of stations, predominantly in the southeast Ethiopia. Among them, 64.08% showed negligible positive trends in the southeast and northwest, whereas 8.74% showed significant upward trends, especially in western Ethiopia (Figure 5c). The positive trend in annual total precipitation (PRCPTOT) indicates a rise in annual rainfall. This upward trend in annual total precipitation indicates rise in overall rainfall, leading to more frequent and intense rainfall events. Over the study period, PRCPTOT increased at a rate of 1.697 mm per year, with the lowest annual average recorded at 763 mm in 2009 and the highest at 1050.3 mm in 2019 (Figure 6c). The increasing trend implies a significant rise in yearly precipitation levels, resulting in more frequent and heavy rainfall events. These findings are congruent with [68], the research title assessment of precipitation extremes and their association with NDVI, monsoon, and oceanic indices. Extreme like annual total precipitation (PRCPTOT) are influenced by factors such as global warming and teleconnection systems, including the El Niño-Southern Oscillation (ENSO) and the positive Indian Ocean Dipole (IOD) [54,69,70].
The annual distribution of RX1 days indicated that 55.34% of stations decreased, mostly in central and northwest Ethiopia. Within this, 48.54% exhibited small decline patterns, especially in central and northern Ethiopia, whereas 6.80% (Aware, Awasharba, Gewane, Gonder, Mahoni, Nazeret, and Quara) showed significant decrease trends at p < 0.05 level (Figure 5d; Table S3). Similarly, the RX5day analysis observed that 56.31% of stations noticed decreasing trends, predominantly in central and northwest Ethiopia. Of them, 48.54% exhibited negligible reduction trends, mostly in northern, northwest, and central Ethiopia, whereas 11.65% indicated statistically significant decreases (Figure 5e; Table 3). The annual time series data revealed that maximum 1-day precipitation (RX1day) increased at an average rate of 0.011 mm. Over the study period, the lowest annual average RX1day was 9.9 mm in 2015, and the highest was 21.5 mm in 2020 (Figure 6d). This increased trend in RX1day indicates that more severe rainfall events occur in a single day, potentially leading to concerns such as flash flooding and urban drainage issues. In contrast, maximum 5-day precipitation (RX5day) declined at an annual rate of 0.011 mm. The lowest annual average RX5day was 37.8 mm in 2015, and the highest was 67.8 mm in 2001. This decrease indicates fewer persistent heavy rainfall episodes lasting five days or more (Figure 6e). Warmer temperatures might increase the atmosphere’s ability to retain moisture, leading to stronger, shorter precipitation episodes (RX1day) and fewer extended heavy rainfall occurrences (RX5day). These findings are similar to those reported in earlier investigations [71,72]. The annual distribution of R10 shows that 54.37% of monitoring stations are experiencing an increase, with 2.91% showing a statistically significant rise at the p < 0.05 level (notably at Debretabor, Neghele, and Sheble), while 51.46% show a non-significant negative trend, most notably in western and southeastern Ethiopia (Figure 5f; Table S3). The time series analysis for R10mm, which represents the number of days with rainfall exceeding 10 mm, found a statistically significant decrease of 0.126 days annually. Throughout the study period, the lowest annual average R10mm was recorded in 2015, when there were no days with rainfall greater than 10 mm. In contrast, the highest annual average R10mm was 30 days in 1998 (Figure 6f). This negative trend indicates a decrease in the frequency of moderate to heavy rainfall events over Ethiopia [73]. The yearly distribution of very wet days (R95p) indicates a downward trend in 66.02% of stations, mainly in central and northwest Ethiopia. Within this category, 56.31% of stations showed a non-significant negative trend, mostly in northern and southern Ethiopia, whereas 9.71% showed a significant negative trend in central Ethiopia (Figure 5g). Moreover, the annual mean precipitation for days with rainfall above the 95th percentile (R95th) decreased by an average of 1.441 mm each year. Throughout the study period, the lowest annual mean R95th was 29.3 mm in 2005 and the highest was 239.9 mm in 1998 (Figure 6g). This indicates a decrease in both the frequency and intensity of extreme rainfall occurrences that exceed the 95th percentile. Similarly, on extremely wet days (R99p), the geographical patterns were similar to those of R95p, with 52.43% of stations indicating a declining trend. Among these, 49.51% showed an insignificant decreasing tendency, whereas 2.91% indicated a significant decreasing trend (Figure 5h; Table 3). The annual mean precipitation on days with rainfall levels above the 99th percentile (R99th) declined at an average rate of 0.9 mm each year. The lowest annual mean total precipitation for R99th was zero mm in both 2004 and 2005, while the highest was 108 mm in 2001 and 2006 (Figure 6h). This decline shows that days with unusually heavy rainfall have become less common over time. These results show that the frequency of very wet days (R95p) and extremely wet days (R99p) is declining over time, indicating a decrease in heavy rainfall events. This trend is consistent with findings from other studies [74,75].
3.4. Variability of Rainfall
The Empirical Orthogonal Functions (EOF) examined spatial and temporal trends of seasonal precipitation from 1994 to 2023. This study identified significant spatial trends and regional variations in precipitation anomalies (Figure 7). The first three modes of Empirical Orthogonal Functions (EOF) account for 59.78% of the variation in JJAS season (Kiremt) precipitation in Ethiopia. The investigation revealed a decline in the seasonal rainfall pattern across Ethiopia. Specifically, the three principal EOF modes accounted for 35.84%, 15.83%, and 8.11% of the variance during JJAS. Most regions in Ethiopia experienced considerable seasonal rainfall variability, with the first mode (EOF1) accounting for 35.84% of the variance, mostly as a positive anomaly. Areas with higher positive loadings showed significant increases in rainfall variability, with these locations observing higher shifts in seasonal rainfall (Figure 7a). This conclusion is congruent with [76], who found that the first empirical orthogonal function of reported rainfall in Ethiopia accounted for 50.6% of total variability The positive anomalies of EOF1 correspond with the decreased pattern shown in PC1. The PC1 time series showed differences in seasonal rainfall during the JJAS season in 1994, 1995, 2001, 2008, 2014, 2015, 2022, and 2023 (Figure 7b). These findings are consistent with earlier research, which has shown that spatial patterns for monthly and seasonal precipitation vary between stations and areas due to differences in climatic dynamics [77,78]. EOF2 revealed negative anomalies over northern Ethiopia, accounting for 15.83% of the overall variance (Figure 7c). This indicated more consistent rainfall amounts throughout time, with fewer extreme and significant variations. The second mode of EOF2 closely follows the general seasonal precipitation pattern, with negative and low loading areas showing less precipitation variability [79]. The third mode of JJAS EOF3 shows a decrease in precipitation (negative anomaly) with an 8.11% variation throughout almost all of Ethiopia (Figure 7e). Rainfall variability in both tropical and extratropical regions is linked to global atmospheric, oceanic conditions and pressure systems interact with the highlands to affect rainfall [80,81].
During the FMAM season, three principal forms of EOF accounted for 45.31%, 15.04%, and 7.44% of the variance. The first mode (EOF1), which accounted for 45.31% of the variation, was predominantly positive and demonstrated significant seasonal rainfall variability throughout the entire country of Ethiopia (Figure 8a). The PC1 time series anomalies throughout the FMAM season show a multi-seasonal decline from 1998 to 2009, followed by an increase from 2010 to 2020 (Figure 8b). This study suggests that FMAM season rainfall in these locations fluctuates dramatically year after year, with considerable precipitation in some years and dry conditions in others, which is consistent with results by [82]. In the southern and southeastern parts of Ethiopia, the second mode (EOF2) accounted for 15.04% of the variance, indicating a negative anomaly and a general drying trend (Figure 8c). The time series anomalies for PC2 throughout the FMAM season exhibit significant seasonal variations, particularly in 1995, 2004, and 2016 (Figure 8d). In the western and southeastern areas, the third mode (EOF3), which accounts for 7.44% of the variance, demonstrated fluctuation with a modest upward anomalous trend (Figure 8e). The principal component analysis for FMAM rainfall (PC3) showed variations in the years 1997, 2001, 2006, 2009, 2017, 2019, and 2023 (Figure 8f). These findings align with a previous study, which proposed different forms of climatic variability as possible drivers of inter-annual and intra-seasonal variability in East African rainfall [80].
Source link
Endris Ali Mohammed www.mdpi.com