The Synergic Effects of Climate Variability on Rainfall Distribution over Hare Catchment of Ethiopia

,


Introduction
Global warming due to climate change is one of today's most persistent problems.It has efects on the lives of people, property, the ecosystem health and services, land use/cover, watershed hydraulics, and large-scale water/groundwater resources [1][2][3].Most scholars suggested that the climate change efect was mostly manifested by an increase in global temperature.In developing countries such as Ethiopia, the efect of precipitation is high as compared to temperature [4].Te amount of rainfall is an indication of climate change since rainfall has a direct efect on the environment and is associated with the growth of the country's economy [5].Te usual form of precipitation in Ethiopia is rainfall.Te agricultural activities of a certain country also depend on the amount of rainfall, which is very important in this regard.
Te existence of a huge population in developing countries such as Ethiopia requires a detailed investigation of the variability of rainfall and its trend.In addition to this natural calamity, there will be, for instance, foods and droughts due to uneven distribution of rainfall.Hare catchment, the catchment found near Arba Minch city, was the susceptible region with the aforementioned problem.In the past, in 2001, 2007, 2013, and 2019, the surrounding areas of the catchment were extremely agonized by this problem, and in the year 2013, the highest food enormity of 60-120 cm was recorded with the resettlement of around 60 households from their residence.Similar situations have happened in diferent parts of the world.Excessive rainfall caused fooding, for instance, in Kendari, which is one of the cities in Indonesia, where most of the time foods happened, and historically, around 19 food events were documented from 2000 until 2017.Also, another city in Indonesia, Palu, sufered from similar circumstances, and in 2012, a huge magnitude of foods afected a lot of people [6][7][8].For best prediction and reliability purposes, most of the time, the annual rainfall data are preferred to the mean value [9][10][11][12][13][14][15].Te research focuses on inspecting the uneven distribution of yearly as well as monthly precipitation for a record period of 30 years in the Hare catchment of Ethiopia.Te data were utilized to explain which kinds of rainfall experienced signifcant variations as a consequence of the changing climate.Also, the authors in [16] provide the appropriate portion of the study's fndings.Tis paper aims to be used as a guide for comprehensive rainfall investigations and basic investigations of climate variation's impact on diverse forms of rainfall to come in subsequent rainfall and environmental change studies.Te present research applied a bias-corrected MME to nine GCMs [17].Tis helps to examine the erraticism of precipitation under historical and upcoming weather conditions.Te RCA4 RCM was selected while, in conjunction with the CORDEX project, and it downscaled an enormous amount of GCMs for Africa, indicating a desire to examine the efectiveness of the GCMs in downscaled scenarios compared to measured or actual data.Te research site is recognized as the cultivated endowed regions, plus extensive hydraulic structures or schemes for growing agricultural crops are presently ongoing.In addition, the region's growing population drives up the consumption of water.Terefore, it is essential to keep track of the region's water supply.A multi-RCM approach was employed in the current research for evaluating the impact of climate variability on hydrometeorology, including recently demonstrated emission pathway predictions that will enhance the nation's adaptation strategy.Te study encompassed the following specifc objectives in our study to accomplish this main aim: (i) efciency evaluation of the regional climate model; (ii) pattern identifcation for metrological and hydrological parameters; (iii) calibration and validation of the hydrological model; and (iv) sensitivity analysis of model parameters.In addition, by illustrating how to utilize RCMs in development attempts, this research is likely to be highly helpful to future scholars.Te anticipated rainfall and temperature outcomes from fve bias-corrected GCMs have been determined with two common conditions, namely, RCP4.5 and RCP8.5 [18,19].Te study's fndings are analyzed and compared to those of prior studies [20,21] in the same and nearby catchments.Te fndings might aid in the expansion of adequate adaptable strategies for proper water administration to make timely choices in response to the possible impact of changes in climate.

Time Series Metrological Data. Ethiopian National
Meteorological Agency (ENMA) is an essential source for metrological data designed for the Hare catchment from 1987 to 2021, or about 30 years of data.Tree climate data recording sites were situated nearby and surrounding the study area (Table 1).When comparing the availability of data for each station, one station has more available data than the others.Chencha and Dorze stations only ofer rainfall and temperature data (Table 1).But the Arba Minch gauging station (synoptic station) was utilized to provide additional meteorological data for other stations.Arba Minch's weather is much diferent from that of Chencha and Dorze.

Digital Elevation Model.
Te resolution of the DEM is determined by the size of each cell.On December 15, 2022, https://asf.alaska.edu/provided a 12.5 m × 12.5 m resolution DEM for the Hare catchment (Figure 2).Te obtained DEM grids were mosaicked with ArcGIS 10.1 software and utilized in the SWAT model to delineate catchments and for further analysis.

Land
Use/Cover Data.In Ethiopia, rain-fed agriculture is the most widely employed traditional farming technique.Land use land cover can have a considerable impact on land surface sediment erosion.Vegetation cover can mitigate the efect of precipitation on soil erosion.Changes in land cover, such as the conversion of thick forest to agricultural land, have accelerated erosion and increased sediment output at catchment outlets [23].Te Hare catchment is mostly occupied by modestly farmed terrain, with some forestland in the higher reaches.Te lowest half of the catchment has heavily farmed land and shrub vegetation.Te study region also features a high concentration of riparian vegetation.Te study catchment is dominated by highly farmed, moderately cultivated, and shrub regions (Figure 3).

Soil Data.
Te soil type has a certain factor for the runof generation as well as the infltration capacity of the catchment [24].After overlaying clipped study area map to the Ethiopian soil map the classifed major soil type of the catchment are easily identifed and listed as follows: eutric nitisols dystric nitisols, orthic acrisos, eutric fuvisols, and dystric fuvisols (Figure 4).

Setup for SWAT Model
2.4.1.Delineation of Catchment.Arc Map interface voguish ArcGIS 10.1 was used to manage and interpret geographic data that were utilized as input for SWAT.Catchment delineation is the frst stage in starting a SWAT model catchment simulation.SWAT allows users to designate catchments and subcatchments utilizing DEM to perform sophisticated GIS tasks to assist users in splitting catchments 2 Advances in Meteorology into numerous hydrologically related subcatchments for use in SWAT catchment modeling [25].During the delineation of catchments and subcatchments, the Hare catchment was defned with an outfow point at the catchment's outlet.Te ArcGIS was enhanced by using a catchment delineation tool.Catchment delineation was performed by SWAT model.Te model created the stream network in place of the entire digital elevation model by utilizing the concepts of direction and accumulation of fow.Te smaller threshold area revealed greater drainage network information, as well as a signifcant number of subcatchment and HRU.Tis, however, necessitates additional processing time and a considerable amount of computer space.For this study, a threshold area of 540 ha was used, and the catchment outfow was manually inserted and selected before concluding the catchment delineation.Te model then defned a 187.14 km 2 catchment with 19 subcatchments (Figure 5).

Analysis of Hydrologic Response Unit (HRU).
HRUs are subcatchment regions that have a distinct land use/cover, soil, and slope combination.HRUs can be assigned to each subcatchment by assigning only one HRU based on the major spatial data combinations.A multiple HRU analysis option was employed for this investigation.SWAT land use datasets have four-letter codes established in the GIS interface (Figure 6).To connect the land use map to the SWAT database, the lookup table in the SWAT was prepared in a way that was consistent with the loading of the land use/cover map.By loading lookup table, the soil layer on the geographical map was connected and stored in the database.To incorporate this map into the model, a user-defned soil database which includes physical as well as chemical characteristics of each soil was created and amalgamated together with the combined lookup table.Te provided soil map's categories of soil have been encoded using a lookup table (Figure 7).
Furthermore, to land use and soil, HRUs have been categorized by slope classifcation.Te multiple slope approach tends to be desired when considering number slope categories for HRU defnitions.In the present investigation, a lot slope alternatives were selected and the slope class has   been divided into four classes with slopes that ranged from 0-3% to 3-6%, 6-12% to over 12% (Figure 8).Following redefning the spatial structure, entirely of the aforementioned corporeal characteristics was layered on top for HRU defnition.As stated by [26], the percentage of land use, soil and slope is 20%, 10%, and 10%, respectively, for the majority purposes of modeling.Small land use, soil, and slope classes inside a specifc subcatchment could be dominant over close to signifcantly larger physical characteristics throughout HRU identifcation under specifc threshold scales.Te HRU in the present research was established by a 10%, 5%, and 5%, respectively.In the end, 77 HRUs for 19 small catchments were established alongside the whole Hare catchment HRU map (Figure 9).

Weather Data Defnition
(1) Weather Generator.Te precipitation statistical analysis model (PCP STAT) produced by the weather generator was used for the statistical analysis of everyday rainfall records required for creating climate fles.Te dew point (dew02) is a confgurable value for the weather generator.As stated by [20], dew02 is used for calculating the daily mean metrological data.Angstrom-Prescott empirical equation ( 21) was applied for converting existing sunlight hour into solar radiation.Weather observatories were placed using latitude, longitude, and elevation measurements.

Model Sensitivity, Calibration, and Validation
. Te model's appropriateness aimed at the intended purpose ought to be assessed by sensitivity analysis, standardization (calibration), and justifcation (validation) [22].Te calibration process involves altering the input variables and comparing anticipated results with actual results till the target function is achieved, and the calibration is carried out by using automatic calibration.Nevertheless, in the present investigation, an automated calibration approach was used for calibration from 1990 to 2001 and validation from 2002 to 2007, with a two-year warm-up phase from 1988 to 1989.Te calibrated model was evaluated in contradiction of a self-determining set of observed data in order to be used for measuring sediment production.Te capacity should be treated as sound in simulation stages of the evaluation [27].Model performance was evaluated by visual inspection of hydrographs value and with combination of objective functions.

Model Performance Evaluation.
A model's accuracy, consistency, and fexibility must all be considered.To evaluate the model's performance, a forecast efciency  Advances in Meteorology criterion is required.Assessing the efcacy of a hydrologic model necessitates subjective and/or objective judgments of the model's simulated behavior's proximity to data [23].In this study, the model's performance is evaluated by the following points.
(1) Nash-Sutclife Efciency (ENS).It helps to judge the ft concerning the outcome of the model and actual measured hydrograph shapes.Te efectiveness of the model is determined by ENS as described in the following equation: ENS can range between 1 and −∞ and operates best when it is one.Values ranging from 0.80 to 0.90 suggest that the model works very well, whereas values ranging from 0.90 to 1 designate that the model performs exceptionally well [24].
(2) Coefcient of Determination (R 2 ).R 2 refects the model approach to recreate the observed value through a given time period and for a given time step.R 2 values vary from 1.0 (best) to 0.0.
Te predisposition of anticipated threshold higher/ smaller than measured value is assessed by percent bias (PBIAS) [25].Te absolute value of PBIAS should be as low as possible for a well-performing model.Te PBIAS is provided by the following equation: (3) Te Ratio of Root Mean Square Error to Observation Standard Deviation (RSR).It serves as an error index indicator.RSR has a value between zero and one, with the lower value, closer to zero, suggesting better model representation and one indicating poor model performance.
where q si is the simulated discharge (m 3 /sec), q oi is the measured discharge (m 3 /sec), q s is the average simulated discharge (m 3 /sec), and q o is the average measured discharge (m 3 /sec).

Soil and Water Assessment Tool (SWAT).
SWAT model is a physically based, semidistributed, long-term simulation, deterministic, and originated from agricultural   models with spatially distributed parameters and operating on a daily time [26].SWAT model has been used worldwide and considered as adaptable environmental model that can be used to evaluate the biophysical impacts of intensifcation of interventions at the watershed scale, which supports more efective watershed management and the development of better informed policy [28,29].Te model has been widely applied for the simulation of runof, sediment yield, nitrogen, and phosphorus losses from watersheds in diferent geographical locations, with varying soils, land use, and management conditions over long periods of time.Several researchers for instances, the authors [30] were proved the applicability of SWAT model in the Ethiopian watersheds.
2.4.7.Description of Regional Climate Model.In this study, projected climate datasets were used.Datasets with a grid spacing of 0.44 °0.44 °(50 km × 50 km) were accessed from the CORDEX-Africa database at https://cordexesg.dmi.dk/esgf-web-fe/.Te reference period for the analysis was from 1986 to 2005, and the future midterm period was chosen between 2051 and 2080 to match the timeframes typically utilized in studies of climate change.Basic climatological information including precipitation, maximum and minimum temperatures, solar radiation, wind speed, and relative humidity are included in this climate model's output for both the reference and future periods.Te climate efect assessment was conducted using the extreme (RCP 8.5) and the intermediate (RCP 4.5) emission scenarios for the midterm era (the 2050s).In the present study, seven regional climate model datasets were employed.

Evaluation of CORDEX RCMs
. By plotting the actual and simulated data for the yearly cycle and interannual variability, the CORDEX rainfall simulations were evaluated using statistical performance indicators.Te analysis was carried out using a statistical technique for evaluating model performance that encompasses bias, root means square error (RMSE), Pearson correlation coefcient, and coefcient of variation (CV) and is discussed in the following sections [31].
where the R bar represents the mean value of rainfall in the analysis period; R is the average rainfall in the basin in a given year; rcm is a subscript for the regional climate model, while G refers to a subscript for rainfall values obtained from the rain gauge network.; σ is the standard deviation.Te RCMs used in this study are the Canadian Regional Climate Model CanRCM4, KNMI Regional Atmospheric Climate Model, Version (RACMO22T), SMHI Rossby Center Regional Atmospheric Model (RCA4), MPI regional model (REMO), CLMcom COSMO-CLM (CCLM4), and CLMcom COSMO-CLM (CCLM4).
2.4.9.Bias Correction.Bias correction for rainfall and temperature was made using the CMhyd tool with the nearest grid of RCM data.Researchers have utilized a variety of bias correction techniques to eliminate bias from climate model data [32].For this study, the distribution mapping   Advances in Meteorology technique was used to correct bias in the dynamically downscaled temperature and precipitation data.Teutschbein and Seibert (2012) went into greater detail about the techniques.On the other hand, the mean-based bias adjustment approach was used to correct a bias for RCM simulation of future relative humidity, wind speed, and solar radiation [33].Equation ( 6) provides the mean-based bias correction method used in this study.
where X adj � adjusted data; X rcm , future � RCM future data; and X obs,hist and X rcm,hist are the mean for observed and RCM data, respectively.

Stream Flow Calibration and Validation.
Te goal of the calibration procedure is to see if the simulated and observed fow values coincide by changing the sensitive model fow parameters within the specifed range.For subsequent iterations in the calibration periods, the six more powerful (governing) fow characteristics are used as described in Table 2.
Te validation procedure was also followed without altering the model fow parameters that had been altered during the calibration step.Te model's performance was examined throughout the validation period (from January 1, 2002 to December 31, 2007).
According to Figure 10, the maximum model output occurs in August 1998 for calibration and September 2004 for validation.Furthermore, the hydrographs (Figure 10) revealed that the model somewhat overstated fow in most years while underestimating fow in others.Tis means that the uncertainty analysis revealed that around 27% and 26% of the calibration and validation data, respectively, were questionable.6 depicts the monthly patterns in maximum and lowest temperatures.Accordingly, February has the greatest maximum temperature (28.8 °C) and July has the lowest (23.0 °C), while March has the highest minimum temperature (13.1 °C) and November has the lowest (10.4 °C) (Figure 11).

Baseline Hydroclimatic Variables. Figure
As seen in Figure 12, the comparison of actual precipitation to GCM reference values.Similarly, in Figure 7, the highest signifcant variation between the MIROC5 baseline and the July values was 14 mm/month.Te IPSL-CM5A-MR had the most substantial divergence in rainfall among all GCMs.
3.2.2.Temperature.Changes in the worldwide diurnal temperature range, DTR (T max T min ), are an important indication of climate change [34].We analyzed the DTR in the Hare catchment using the averages of the fve GCMs for the baseline and future eras to study the signifcance of the variability of temperature as presented in Figure 13.Under all scenarios, the DTR will be smaller in the Kiremt and late Belg periods than in the baseline period, with just a slight diference in the beginning of the Belg season.

Rainfall. Figure 14 depicts percentage variations in
expected rainfall on a seasonal and yearly basis for the near and distant future timeframes.Te forecasts suggest that the Bega season will be longer, while the Belg and Kiremt seasons would be shorter.Te estimated yearly rainfall loss varies from 6.5% (MPI-ESM-LR-RCP4.5) in the near future to 38.3% (CanEMS2-RCP8.5) in the distant future.CanESM2-RCP8.5 seasonal estimates indicate the greatest substantial declines in the Belg (55.2%) and Kiremt (51.5%) seasons.MPI-ESM-LR-RCP4.5 produced the lowest seasonal decreases of 8.5% in the Belg season and 11.2% in the Kiremt season.

Evapotranspiration.
Te current study's fndings show that both emission scenarios enhance evapotranspiration.Te spike is linked to the higher-than-expected temperature rise (Figure 14). Figure 15 depicts a steady increase in average annual and seasonal evapotranspiration rates across all scenarios.Te anticipated annual average rise for RCP4.5 and RCP8.5 is 6.3-14.8%and 8.9-16.8%,respectively.
Figure 16 shows a strong connection between the monthly baseline stream fows predicted with various GCMs and the actual stream fow, with a coefcient value of roughly 0.89.Tere is, however, a diference between the observed and baseline periods.Te highest diference is 7.8 mm/ month achieved with the MIROC5 model in November, followed by 7.3 mm/month acquired with the IPSL-CM5A-MR in March.For the baseline period, the simulated stream fows with CanESM2 and IPSL-CM5A-MR are higher than the observed values, but with CSIRO, they are lower, except for January.
(1) Impact of Climate Change on Drought Characteristics.We examined the likelihood of drought incidence and mean drought index (SPI, RDI, and SDI) in the study area using an  Advances in Meteorology ensemble average of fve GCMs for nearby and future periods under, RCP4.5 and RCP8.5 (Figure 17).Te chance of drought incidence from January to December is calculated by dividing the number of years with a drought index value of 1 by the total number of years in the period [34].Under RCP8.5, it reached its maximum (0.19) for the Belg (from February to May) in the distant future era (Figure 17(a)).Tis increased likelihood is mostly due to a considerable drop in predicted rainfall.Another cause might be the higher unpredictability of rainfall in Belgium.Te SPI and RDI data demonstrated a similar temporal pattern of drought incidence (Figure 17(c)).Tis suggests that drought susceptibility is also strongly connected to the sorts of land use circumstances that may lead to available water shortages [35].

Advances in Meteorology
Te probable function of the hydrological drought from the three indexes shows a value of 0.07 for the base line scenario and 0.14 for the predicted one.Figure 17 describes almost similar pattern for the three drought indexes.Te output of this study also comparable with the scholars who put their own suggestion in upper blue Nile and Bilate catchment showed that the possibility of existence of metrological drought 0.22 and 0.16 in Belg and Kiremet, respectively [36,37].Also, the domain of existence of drought is manifested by the variability of climate especially rainfall over the area.
Figure 18 depicts the average mean monthly precipitation change of all individual RCMs over the research region for the midterm 2050s period.Climate model projections suggest that, with the exception of the CanRCM4 model, there will be a probable constant decrease in future rainfall quantity for virtually all months.Te CanRCM4 model predicts a rise in rainfall quantity from January to September.Te model projection for the base period is correct based on the data collection as shown in Figure 19.
Figure 19 depicts the expected midterm change in mean seasonal and annual rainfall for the Hare catchment related to the reference period.With rare exceptions, the anticipated changes in the basin's mean annual rainfall have showed   Advances in Meteorology the selected RACMO22T model, show increasing precipitation, while the remaining three models show decreasing precipitation.Te expected mean seasonal rainfall variations for Bega and Belg seasons are from −71.5% to 67% and from −69% to −8.5%, respectively.Te predicted variations in average periodical rainfall for the Kiremt season range from −35% to 65%.Rainfall in the three seasons (Bega, Kiremt, and Belg) will likely vary by −18.1%, −0.023%, and −26.4%, on average and its approachable in value with the study [38].16

Advances in Meteorology
With a few exceptions, rainfall simulations from most RCMs under RCP 8.5 indicate a steady drop intended for entire periods.For the Bega and Belg seasons, the estimated decreases in mean seasonal rainfall changes range from −68.6% to 87.4% for RCP 4.5 and from −59.6% to 15.5% for RCP 8.5, respectively, and similar study is conducted by [32].Climate models predict a range of −12.1%-1.33% in mean seasonal rainfall for the Kiremt season.Te average quantity Advances in Meteorology of average periodic precipitation will likely fall by a certain amount for Belg, Kiremt, and Bega seasons as 28.2%, 12.0%, and 22.6%, respectively.

Conclusion
Except for the CanRCM4 model, almost all RCMs indicated a considerable decline in average annual rainfall and stream fow.Te CanRCM4 model simulation of rainfall in the Hare catchment difers dramatically from all other models.Tis might be ascribed to its unfortunate presentation in the study area, which could be linked to structural discrepancies in RCMs and process parameterization.For both climate emission scenarios, the mean yearly rainfall would be decreased from 16.7% to 10.2%.Tis study's expected outcomes are reliable through further scientifc researches in the catchment.Te study is crucial for rain-fed agriculture, management of reservoir storage and further water-related activities.Troughout general, maximum temperature, lowest temperature, and potential evapotranspiration exhibit reasonably consistent trends throughout the project region, whereas predicted rainfall shows signifcantly less consistency and volatility.Moreover, three drought indices SPI, SDI, and RDI are also used to examine how climate change afects drought aspects.Te results show that future droughts will be more severe and protracted than the baseline era (under both emission scenarios).Residents of the Hare catchment depend greatly on the provision of runof that found in natural waterways and periodic streams, that is negatively impacted by absence of rainfall brought on by climate change.Signifcant courtesy must be given to adopting new and problem solving mechanisms which is safe to climate alteration to preserve viable agricultural production and food for the future generation.

Figure 9 :
Figure 9: Full hydrologic response unit (HRU) map of the Hare catchment.

Figure 10 :
Figure 10: For calibration and validation, a monthly observed and simulated hydrograph.

Figure 19 :
Figure 19: Projected change in monthly rainfall of the Hare river basin for the medium-term future (2051-2080) compared to the reference period (1986-2015) under the RCP4.5 and RCP8.5 scenarios.

Table 1 :
Summary of availability of metrological data.
No Recorded stations Latitude

Table 2 :
Summary of calibrated model fow parameters.