Statistical-based spatial analysis on urban water management under changing environments: a case study of Hawassa, Ethiopia

Hawassa characterizes a typical developing city in Ethiopia, owning to rapid urban growth and demographic trends. The combined effect of climate change and urban expansion is increasing the challenge to the environment and the services it provides. Relating changing environments with urban water management (UWM) is required to build resilience in the urban environment. This research analyzed local climate change and urban growth and linked it to UWM. The historical period 1990–2021 of daily rainfall, temperature variables, four satellite imageries, and DEM were analyzed. Changes in rainfall (annual and daily maximum) and temperature (maximum and minimum) trends are detected and projected to 2051 using a statistical-based model. With geospatial techniques sub-watersheds are delineated, and the urban cover change is quantified. The trend detection result implies an upward trend of annual and daily maximum rainfalls however a significance is insufficient (p > 0.05) to associate it with climate change during the study period. Maximum and minimum temperatures change indicate a positive and significant trend. The forecasting result suggests an increment of both temperatures (0.5 °C–1.5 °C) to the projected period compared to historical scenario. The land cover analysis results show the built-up area changed from 11.6 km2 (7.2%) to 42.5 km2 (26.5%) during the historical period, where the rate varies spatially. The surface runoff increased by 30.7% in the urban watersheds. With a growth rate of 8.9% built-up, the urban area will cover 73.6 km2 (45.9%) for the predicted period. The research finding justifies the potential to reorganize the relationship between the spatial effect of climate change and urban growth on UWM. Considering distinct characteristics of urban watershed, exposure to flooding risk, access to water demand and resilient to climate change have spatial variation. Thus, a local-specific planning approach will support effective UWM and climate adaptation for sustainable city development.


Introduction
Developing countries of the world are in continuous urbanization due to unprecedented rapid population growth, urban expansion, and socio-economic development.According to the United Nations UN (2018), the population living in urban areas will double in the middle of the century.With the current population agglomeration, the African continent will reach 90 percent urbanization during the same period.Ethiopia is a developing country in East Africa, where the rate of urbanization is the fastest and exceeds the average rate of 2.2 percent for developing nations (Central Statistical Authority of Ethiopia CSA 2007, UN-Habitat 2020).Hawassa City represents a typical urbanization process that takes place among the cities of Ethiopia.The city is experiencing rapid population growth and urban sprawl, where the number of urban population and urbanization will reach over one million and 79%, respectively by 2050, based on CSA's (2007) census projection.Like, many cities in low-income countries, Hawassa has also faced environmental sustainability challenges which are vulnerable to changing environments (Admasu 2015, Paes et al 2023).
Rapid urban growth patterns are continuing in upcoming decades, thus several aspects of socioeconomic, environmental, and water-related issues in urban areas are becoming more challenging.Intergovernmental Panel on Climate Change IPCC (2021) has reported that climate change is widespread and intensifying, and global temperature is rising with high variability in rainfall patterns over most regions.Growing evidence has recognized the impact of climate change at various spatiotemporal scales.Change in climate impact has exacerbated adverse effects on the environment and various development sectors, and climate adaptation has become a sustainable city agenda (Tapan and Luna 2017, Gu et al 2020, Abbass et al 2022).Nowadays, most climate impact studies are focused on large-scale global or regional dimensions, while a long-term climate change and associated risks specific to urban environment is not evidenced well (Ye et al 2021, Baack et al 2024).With the available regional climate models and down-scaling techniques, assessing the impact of climate change in urban areas may increase uncertainties about the changing environment.
Recent studies argue that the microclimate pattern of developing cities is unique in their climates, vulnerabilities, and exposures, which has significance for cities' climate adaptation strategies (Filho et al 2019, Wubaye et al 2023).Hence, the urban environment has spatial heterogeneous characteristics, and the impact of changing conditions and responses varies in time and location.Specifically, urban water management (UWM) aspects, such as ensuring access to adequate water service and flood risk management, are apprehensions of future change at various urban scales.However, the spatial impact of changing environments on UWM is not yet thoroughly associated with physical settlement and land use zoning systems.For instance, Bichai and Flamini (2018) indicated that UWM has often been done in hindsight and reflected after the adverse effects of water scarcity and flooding hazards happened.The conventional planning strategies are limited in responding to the emerging challenges of UWM.Lokidor et al (2023) also reviewed it often relied on past experiences by constructing short-term water management infrastructures to cope with upcoming changing conditions.
The management of urban water is a constraint in human development and life, which has a focus area of the sustainable development Goals (SDGs) and intimidates future city development (Oldekop et al 2016, UN-Habitat 2021).Thus, local-specific UWM can sustain to achieve the SDGs agenda, specifically SDG11-Sustainable Cities and Communities, and SDG6-Clean Water and Sanitation (United Nations UN 2016).The decision-making options for the adaptative management of urban water should be addressed at a local-specific scale.However, evidence-based knowledge on the spatial effect of change in local climate and urban sprawl patterns is inadequate and that has not been researched well on the most recent challenges of UWM.Understanding the local-specific influence of the explanatory variables may offer vital information for a future challenge of city development.Fallmann and Emeis (2020) also highlight a spatial planning approach has significant importance in maintaining environmental sustainability.The impacts are perhaps easily managed in a small urban watershed, while on a larger scale, the effect is more complex than the sub-watershed response (Dorning et al 2015).
Spatial analysis involves the identification of the dynamics and determinant variables and links with the challenge in specific areas, which leads to a new paradigm of specific-solutions to specific-problem approaches.Coupled with the impacts of a changing climate and rapid urban growth, many developing cities are experiencing environmental challenges.Recently, studies have discussed the requirement of fundamental shift to implement solutions to climate-resilient development and a sustainable urban environment (Cousins 2024).In cities across the world, nature-based solutions are gaining importance as promising solutions in local and global agendas, that enable spatial-specific urban sustainability and climate resilience development (Gulsrud et al 2018, Adams et al 2024).How spatial urban planning can help to achieve sustainability has been identified as a research gap in most developing cities.Spatial analysis on UWM enables us to plan better decision-making which can be closely linked to spatial urban planning strategies.This provides numerous opportunities, such as reducing the vulnerability of urban water systems, climate change adaptation and mitigation, and maintaining environmental sustainability.
A recent study by Özerol et al (2020) discusses the importance of considering cities as water-sensitive communities, and UWM is often part of the watershed and service provider.In recent years, several integrated planning approaches have been developed to the concept of natural-based techniques, suggesting urban water forms are valuable resources, and the urban watersheds as the management unit (Bichai andFlamini 2018, Marana et al 2019).The optimal selection of natural-based measures depends on local-specific characteristics, such as environmental sensitivity, socio-economic issues, and urban planning framework (Starzec et al 2020, Cong et al 2023).The combined effect of change in local climate and urban expansion patterns linked to spatial challenges of UWM are ongoing research agendas to deal with future uncertainties.Hence, integrated spatial planning processes are needed to quantify the current and emerging environmental challenges in response to changing circumstances.In this regard, detailed local-specific clarity on changing variables is required to plan adaptive UWM strategies ahead of time.
Considering a wider perspective of UWM challenges, spatial analysis of UWM is an important innovative approach for better decision-making to sustainable city development.However, characterizing the local-specific effect of changing conditions on UWM at various local-specific scales has not received attention and is less investigated thoroughly.Thus, obtaining evidence-based spatiotemporal information on urban water services and management is difficult in a consistent way.These have necessitated a paradigm shift to a holistic planning approach to integrate UWM and the built environment in the spatial urban planning process.This will determine the city's decision-making strategies to improve the urban services to an acceptable level for future sustainable city development.Therefore, the main objective of this research is to contribute quantitative knowledge on the trend of changing variables using statistical model-based geospatial techniques in urban watersheds.
The historical period 1990-2021 daily rainfall and temperature time series, four satellite imageries, and the Digital Elevation Model (DEM) of Hawassa are analyzed.Temporal changes in the climate variables are detected and projected to 2051 using statistical trend test techniques and projection models.The satellite images and DEM were analyzed with geospatial techniques, the study area was divided into sub-watersheds, and spatial land cover change was determined.The trend of changes in rainfall, temperature, and urban expansion are related to water demand and surface runoff change in the selected urban watersheds.This research also highlights the way local climate and urban land cover change are correlated.The research findings will help to devise local-specific adaptive strategies to address the challenge in UWM and climate adaptation, and better complement the spatial planning approach in the urban environment of Hawassa and developing cities.

Description of the study area
Hawassa is the largest urban center of the Lake Hawassa watershed, part of the Great Rift Valley basin of Ethiopia.The city is situated at 6°55′ to 703′ latitude North and 38°25′ to 38°34′ longitudes East, adjacent to the shoreline of Lake Hawassa (figure 1).It was founded in the 1960 s with a modern urban planning concept, and the historical development trend of the city varies with different periods.Since the inception of the federal government in the country in post-1991, the decentralization policy has designated Hawassa as a political and administrative center of southern Ethiopia (Admasu 2015).Hawassa is the capital city of Sidama Regional State of Ethiopia, and it is the center of the political, administrative, and cultural hub of the country.It has an area of 160.6 km 2 and an elevation range of 1680 and 1708 meters above sea level (masl) at the lake's surface and the urban center, respectively.Hawassa has a plain topography, and the entire urban drainage system of the city is towards the lake.
The city has moderate warm temperatures that range from 10 to 30 °C.It receives an average annual rainfall of 975 mm, which has a bimodal rainfall pattern, and maximum rainfall events have happened from July to December.The city comprises seven urban and one rural sub-cities, including 32 'kebeles' or the lowest administrative levels.Hawassa City Administration HCA (2018a) development plan of the city reveals the residential area is a major land use (>40%) which covers the central sub-cities and sideways to the lakefront.Other land uses include commercial, industrial, urban agricultural, and green open spaces.However, the development plan of Hawassa did not clearly show the relevant urban water infrastructures of the city.According to the Ethiopian Central Statistical Authority CSA (2007) earlier census report, the annual population growth rates for urban and rural areas are 4.08% and 2.08%, respectively.The population of Hawassa is estimated to be 454,698, of which 68% of residents are urban, based on the CSA (2007) census projection.
Over the past three decades, the city has undergone rapid urban expansion and population growth.The city's diversified biophysical nature to tourism attraction, socioeconomic character, various industrial developments, and reclassification of near rural areas into urban areas have accelerated rapid urbanization and urban population growth (Kinfu et al 2019).The rapid urban development trend of the city is continuing with positive prospects, exerting pressure on an essential aspect of UWM in the anticipated future.The rapid population growth, urban expansion, climate change, and insufficient management of urban water may cause a new challenge to the urban areas.As a developing city, Hawassa should address the current and emerging environmental challenges in response to changing circumstances.The environmental issue of the city has an impact on its capacity to handle the emerging challenges to sustainable city development.

Data acquisition and processing
This research has carried out a comprehensive study on changing environments and challenges to UWM in urban areas of Hawassa.The research collected and processed various data, including station-based climate data, satellite imageries, Digital Elevation Model (DEM) with 30 m resolution of ASTER satellite image, soil type, and supplementary data of the study area.The trend of urban climate is analyzed using the historical 30+ years Changes in daily rainfall, temperature variables, and urban expansion patterns are reflected in water demand and changes in surface runoff at delineated urban watersheds of the study area.The trend of change in climate variables is detected using a statistical model.Spatial land cover change is assessed using freely available Landsat and Sentinel-2 satellite datasets with geospatial techniques.The change in surface runoff depends on watershed characteristics and rainfall depth, which is estimated using the Soil Conservation Service-Curve Number (SCS-CN) hydrological model.Specifically, this study has used various research methods and is described briefly in the following sections.

Trend analysis of change in urban climate
A long-time series of daily rainfall and temperature data are usually used to detect the effect of climate change in urban hydrological studies.The historical period 1990-2021 of daily rainfall and temperature data are used for station-based trend detection.World Meteorological Organization, WMO (2017) suggests a 30-year early period reference data of climate variables are required to reflect change in climate using statistical and climate models.The meteorological station of Hawassa location is 38°28′ longitude East and 7°06′ latitude North with an elevation of 1694 masl, which describes the urban climate characteristic of the city.The consistency and missed values of the station-based climate data were checked and assembled using Easy Fit 5.5 and XLSTAT statistical software.Further, the daily (annual and maximum) rainfall, and temperature (maximum and minimum) variables are considered as representative of urban climate change detection and forecasting for the future period up to 2051.Hence, statistical model-based analyses and values are involved and described as follows.Statistical trend detection techniques, linear regression/parametric and non-parametric Mann -Kendall (MK) test at 95% (α = 5%) confidence level, are widely used in meteorological time series trend detection.Parametric trend tests are involved in normally distributed climate data yet are more susceptible to outliers (Mudelsee 2019).On the other hand, the non-parametric MK and Sen's slope trend test is not required to meet the normality of observed data and is less affected by the outliers, as their statistics are based on the sign of differences rather than values of the random variables (WMO 2018).Both parametric and non-parametric trend detection methods are recommended to reduce uncertainty in trend analysis.The trends of change in urban climate data are detected using linear regression and MK trend test techniques and then projected up to 2051 by a statistical forecasting model.The two-sided students' t-test is used to evaluate the statistical significance of change and variability of climate variables.
With hypothesis test procedures, statistically significant signals of urban climate changes are detected at α = 5%.The null hypothesis (H0): there is no statistically significant monotonic trend when the computed p-value > 0.05.The alternative hypothesis (H1): change in climate time series has a significant trend when the p-value is < 0.05.In descriptive statistics and variability analysis, the variance (σ 2 ) and standard deviation (σ) detect variability of climate data from the mean (μ) value of the observed period.σ 2 measures the average squared difference between climate data and μ values, which is used to avoid bias when the observed value is less than the mean.Also, coefficient variation (CV) value indicates the variability of climate elements in time series.The CV value is computed as the percentage of σ divided by μ value.Dessu et al (2020) discuss the value of CV < 20 shows low variability, medium (20 < CV < 30), and high (CV > 30).
The normal probability distribution with the normal curve is a widely used method to analyze the normality of distribution data (Aditya et al 2021).The normality of the climate data distribution is analyzed by kurtosis, skewness, mean and median values, first and third quartile values, and visualized with a box plot.The kurtosis values indicate how the data tails off from the central position of the normal distribution curve.The skewness measures symmetry of the data distribution around the mean or median values.The box plot and whisker illustrate the shape and skewness of the distribution data clearly through their descriptive statistical and quartile values.Whisker, the lines extended from the box indicate outliers from quantiles.The spatial-specific climatic impact study in urban environments becomes a primary step for effective adaptation strategies to changing environments.
The trend of change in rainfall (annual and daily maximum), and temperature (maximum and minimum) climate variables are detected, and then the possible effect of local climate change on surface runoff generation and urban water demand in the study area are discussed.The Intensity (Depth)-Duration-Frequency (I(D) DF) curve is widely used for the design of surface runoff control structures and hydrological studies in urban watersheds.The I(D)DF curve quantifies the design rainfall considering different rainfall durations.The curve is often developed by using a historical local maximum rainfall value of various durations upon the assumption of the hydrological time series.The I(D)DF curve is developed at the national level through the standard procedure by responsible governmental agencies.Ethiopian road authority ERA (2013) standard for the urban drainage design guideline suggests, that the daily maximum rainfall data are formatted to hourly and sub-hourly design rainfall using a regional storm intensity equation (1).
Where: P t is rainfall depth for time t; P 24 is daily rainfall depth in hour; b = 0.3 and n = 0.78 to 1.09 are coefficients, and n = 0.935 is considered.The DMRF data is considered normally distributed and representative of the study area.Change in daily maximum rainfall effect is integrated into the I(D)DF curve with a return period of 2-to 100-years.The rainfall depth of 5 min, 10 min, 30 min, 1 h, 2 h, 6 h, 12 h and 24 h duration for corresponding return periods was determined.The Gumbel extreme value distribution equation is adopted, which is widely applied for modeling extreme events of normally distributed observed rainfall time series (Carlier andEl-Khattabi 2016, Awofadeju et al 2018).Statistically significant changes in temperature variables are considered to identify the long-term climatic effect on urban water demand forecasting.The climatic impact factor is defined by using equations (2) and (3).
Where: ΔUWD is change in domestic water demand due to change in temperature (m 3 ), D t is the climatic elasticity of domestic water demand (%), and ΔT is the change in temperature (°C) with in the projection period.

Analysis of spatial urban expansion pattern
Urban areas of the city are defined based on administrative boundaries, socio-economic activities, satellite imagery, and urban water services.Thematic land cover maps for 1991, 2006, 2016, and 2021 were prepared using geospatial techniques, like a recent study by Abreham et al (2023).A supervised -maximum likelihood image classification algorithm was used for urban land cover change detection, and further manually corrected and validated to reality on the ground truth.Four major land cover classes were identified and defined as built-up/urban area, vegetation area, agricultural area, and waterbody.The classified images were validated and verified through 50 random sample reference points integrated with Google Earth imagery for each land cover class.
The overall accuracy and Kappa coefficient of the classified land cover class was more than 85%, which suggests the accuracy requirement for the classified images (Abreham et al 2023).Also, the city is delineated into sub-watersheds using the DEM image and geospatial procedure.The spatial land cover change and urban expansion were determined in the delineated into urban watersheds of the city.Based on the historical change in built-up, urban land cover is projected up to 2051.The percentage of land cover change between each period was determined using equation (4).
Where: P i is percentage of the land cover change; LC i and LC f is the Land cover of the initial and final years, respectively.The linear regression result (figure 2) implies a temporal change in DARF at a period (t) that can be expressed by the equation, DARF t = 4.4248 t + 904.17 with a coefficient of determination (R 2 ) of 0.0691 and coefficient (β) value of 4.4248 mm/year.The DMRF is related by equation, DMRF t = 0.1672t + 47.741, R 2 = 0.0239 with β = 0.0239 mm/year.However, the trend in both ARF and DMRF computed p > 0.05.Characterizing local rainfall patterns is substantial for enhancing the resilience of runoff control infrastructures to changing climate.The design rainfall for the usual return period of 2, 5, 10, 25, 50, and 100 years were obtained.The daily maximum rainfall data was formatted to rainfall depth of 5 min, 10 min, 30 min, 1 h, 2 h, 6 h, 12 h and 24 h duration for corresponding return periods.Figure 4 shows the design rainfall depth generated for each duration and frequency/return period for Hawassa.This research suggests an insight into the potential of change in local climate and its impact on the urban water sector.Thus, the first attempt to introduce evidence-based knowledge for the spatial challenge of UWM to changing local climate.The complex processes of rainfall patterns and natural spatial variation influence the signal of climate change at local and regional scales.The review of recent studies suggests a consensus that daily and short-duration rainfall events are expected to increase in response to climate change (Vassilios and Tsihrintzis 2022).With single station-based rain gauge   observed data and similar landscape conditions, the effect of topography may be invisible in the spatial variation of urban environments.

Results and discussions
In the case of Hawassa, statistically significant trend of local rainfall time series is not observed, which may not prevent mitigation and adaptation measures against climate change impact.Martel et al (2021) argue that observed short-duration rainfall records in urban environments remain scarce to link the climate model results.The design concept of the I(D)DF curve is based on the probability distribution of daily maximum rainfall, expecting a future that is the same as historical observations.To account for local climatic impact the curves should be computed using the latest observational records.Thus, the adaptation of UWM under changing conditions can be enhanced.Urban areas are characterized by spatial variation, and a single meteorological record may also limit the spatial representation of rainfall patterns (Kidd et al 2017).Additional local-specific rain gauge observations are needed to characterize the climate of an urban environment.Public participation and knowledge have recently been recognized for dealing with local climatic-related challenges in the urban environment.Alemseged et al (2019) highlight citizen science has gained popularity as a promising cost-effective approach, which helps to narrow gaps in conventional ways to deal with challenges in the urban environment of developing cities.

Trend of change in urban temperature time series
Temporal trends and variability of station-based temperature data are detected to represent the urban climate pattern of the city.Figure 5 shows the trend of maximum (T_ max) and minimum (T_ min) temperature time series for the historical    the model accuracy.Normality analysis results (table 1) reveal that the kurtosis values of 0.12 & −0.8 for T_ max and T_ min, respectively, and both temperature variables tails are relatively normal to the distribution curve.The negative kurtosis value of the T_ min probability function is slightly flatter, and a small positive value of T_ max indicates an oval probability function than the normal distribution.The skewness values for T_ max and T_ min is 0.78 & 0.44, respectively which reveals relatively symmetrical normality.The negative kurtosis value of the T_ min indicates the probability function is slightly flatter, and a small positive value of T_ max is a probability function of an oval shape than the normal distribution.The positive value suggests a symmetrical distribution form to the right of the mean value.The box plot (figure 3 The linear regression trend results (figure 5) signify a temporal change of T_ max and T_ min with increasing and significant trend at α = 5%.T_ max over time (t) is expressed as T_ max = 0.0196 t + 27.242, R 2 = 0.1808 and β = 0.0196°C.The variability of T_ max is indicated by a small change in gradient value of 1.96% per year.The T _max is depicted as a tendency of continuous rise each year, and it will increase by 0.20 °C within ten years.Also, T_ min is defined by the linear equation: T_ min = 0.0665 t + 12.202, R 2 = 0.7145, and β = 0.0665 °C/year.The MK trend test statistics result (table 3) reveals an upward and significant (p < 0.05) trend of T_ max and T_ min, with positive values of Kendall's tau (0.250 & 0. 654) and Sen's slope (0.017 & 0.066), respectively.The S value of 115 & 304 for T_ max and T_ min implies a tendency of rising temperature to the anticipated period.The MK test also agrees with the linear trends detection result of the temperature variables and computed p < 0.05.Thus, a signal of change in urban climate is prevalent due to an increase in temperature during the study period.
The research finding suggests an upward trend with the gradient of 0.020 & 0.067 °C yr −1 for T_ max and T_ min, respectively.If we assume the present scenario and limited adaptation effort to climatic impact in future periods, then the T_ max will reach 27.3, 27.5 & 27.7 °C, respectively, for the projected 2030, 2040, and 2050.The predicted change in temperature variables increment will range from 0.5 °C-1.5 °C by 2050 as compared with historical station-based records, which characterize the urban climate of the city.Chaka and Oda (2019) revealed a comparable result of an increase in the surface temperature of Hawassa.Recent studies have discussed that the temperature change is projected to increase, yet with no clear trend in rainfall change in major cities in Ethiopia (Dessu et al 2020).Globally, it is predicted by 2050 rapid urban land expansion will increase the average daytime and nighttime air temperature by 0.5 °C-0.7 °C, up to nearly 3 °C change will occur in most developing countries of Asia and Africa continents (Huang et al 2019).
In general, statistically significant signal is observed in increasing trend of temperature variables.Hawassa is changing rapidly in the face of the urban environment, which may increase surface temperature.For instance, higher impervious surfaces in urban areas compared to rural areas contribute to an increase in urban temperature effect (Vujovic et al 2021).The form and effect of change in temperature vary in time and space, which depends on natural features and build-up characteristics within the urban environment.

Spatial land cover dynamics and urban expansion
Land cover change analysis result for the historical period 1991-2021 signifies Hawass has experienced a rapid urban expansion to rural areas of the city.As a result, an increase in built-up/urban area and vegetation land cover, while a decrease in agricultural land cover is observed.The urban sprawl pattern is typically a transition from agricultural and vegetation areas to built-up, which varies spatially.Hawassa is progressing in green city development, and vegetation land cover of the urban area has improved remarkably than the rural area.Table 4 shows the overall built-up area is changed from 11.6 km 2 (7.2%) to 42.5 km 2 (26.5%), vegetation area increased from 19.2 km 2 (12.0%) to 36.9 km 2 (23.0%), whereas agricultural area decreased from 128.0 km 2 (79.7%) to 78.9 km 2 (49.1%), and water body increased by smaller proportion during the study period.
The urban growth extent of the city varies with the historical urban development process over three periods with different intervals.The driver of rapid urban expansion pattern is associated with the physical setting, political decisions, administrative boundary condition, demographic change, and historical development plan of the city (Degife et al 2019, Abreham et al 2023).With an average growth rate of 8.9% per annual of built-up area, future urban land cover for 2036 & 2051 is predicted to be 58.1 km 2 (36.2%) and 73.6 km 2 (45.9%), respectively.The drainage line of the city is towards Lake Hawassa (figure 1), and most of the urban center and shoreline of the lake are covered with built-up areas.Thus, the spatial effect of urban expansion will increase surface imperviousness and runoff volume and affect the quality of the lake environment.Figure 6 illustrates the spatial extent of the built-up area over the last 30 years in the delineated sub-watersheds.Spatial urban land cover variation is an emerging challenge of UWM to maintain environmental quality, which draws attention to complement the local-specific planning approach for a better adaptation strategy.Several studies have evidenced an increase in surface imperviousness directly related to an increase in urban runoff depth.Wang et al (2015) indicate a threshold value of imperviousness ranging between 3% to 20%, and the hydrological response of a watershed can be classified as an urban area.Change in land cover of each subwatershed is determined, and the sub-watershed of 10% and above increase in built-up are considered as urban watersheds.Table 5 summarizes the selected urban watersheds and change in imperviousness from 1991 to 2021.Nowadays, several hydrological models are developed to simulate surface runoff from a given rainfall depth at various watershed scales.Hence, the urban watershed has heterogeneous characteristics, and selecting a suitable surface runoff estimation method is essential for effective stormwater management under changing conditions.
The SCS-CN hydrological model is identified as a reliable method to simulate surface runoff from a given rainfall depth in urban watersheds (Hu et al 2020, Verma et al 2021).The runoff depth estimation is based on the Curve Number (CN) parameter, which accounts for the effect of soil type, antecedent soil moisture condition, and land cover variation of the watershed on rainfall-runoff event.For instance, Soulis (2021) suggests that the SCS-CN model can effectively simulate direct runoff depth at a small watershed of less than 30 km 2 and various return periods of rainfall depth.Based on the Natural Resources Soil Conservation Service, NRSCS (1986) guideline, the hydrological soil group, average antecedent soil moisture condition, and land cover class of the selected urban watershed are considered.The SCS-CN empirical method is used to estimate direct surface  runoff depth, and the effect of spatial land cover variation on the runoff generation at the urban watersheds is evaluated.

Correlation between Local climate and urban land cover change
Statistically significant signals of change in local temperature and rapid sprawl urban expansion are observed during the study period.Temperature change may be related to an increase in surface imperviousness of the urban area.For instance, Gebreyesus et al (2022) discuss that rapid urban expansion patterns are a driver of microclimate change, which eventually increases surface temperature.The correlation analysis result indicates a correlation (r) value of 0.47 between maximum temperature (T _ max) and urban land cover (ULC).The r value greater than 0.5 often implies a good correlation between two variables.The linear regression result (figure 7) also shows the effect of ULC change on T _ max.The relationship of change in T_ max and ULC is defined by the linear equation: T_ max = 0.0186 ULC + 27.14, with a coefficient (β) value of 0.0186, R 2 = 0.2215 and p = 0.006 at 95% confidence level.
A numerical understanding of how urban expansion will affect the surface temperature of the urban environment will help to better decision-making strategies under uncertain future conditions.Recent studies have indicated a good correlation between the increase in built-up areas and surface temperature that varies at spatial scales.Hence, the radiative and hydraulic properties of paved surfaces differ significantly from agricultural and vegetation areas.A study by Xu et al (2019) also revealed a positive relationship between a change in surface imperviousness and an increase in urban temperature, because of the modification of local  energy balance and change in materials' thermal properties.The finding generally suggests the significant importance of a spatial adaptation strategy to climatic impact ahead of time to increase the resilience of the urban environment.The effectiveness of adaptation strategies to microclimatic impact, particular emphasis should be given to the local-specific situation of the urban area.The urban environment is undergoing a rapid transformation of demographic trends and socio-economic development activities, which will accelerate the rate of microclimate change and its impact on future changing conditions.
3.5.Effect of changing environments on urban water management Spatial variation of climate and urban growth patterns were considered to assess their effect on UWM aspects, which include changes in urban water demand and surface runoff depth of urban watersheds under changing conditions.Changes in temperature and rainfall variables are contemplated to identify the long-term climatic impact on UWM.The forecasted result of local climate change suggests a statistically significant temperature increase (0.5 °C-1.5 °C) is expected by 2050 compared to the historical scenario.Temperature change probably increases annual urban water demand.The rationale is that water demand is higher during hot temperatures, as more water is required for domestic water use, personal hygiene, and gardening.Historical data on urban water demand change is not available in consistent way, hence quantitative knowledge relating local climatic impact to changes in domestic demand is scarce in most developing cities.
The impact of climate change on domestic water demand can be forecasted by assuming hypothetical scenarios.Reviewing the literature is also considered more appropriately, and the magnitude of climate effect can be linked with change in domestic water demand with a plausible range of impact factors.Wang et al (2018) and Dimkić (2020) signify an increase in temperature per 1 °C results in a range of 0.85-5.3%increase in domestic water consumption in various cities worldwide.The effect of temperature change is represented by a climate impact factor and an increase in annual water demand for each degree change in T _max is estimated.According to urban water supply design criteria of Ethiopia, towns are characterized by widely varying climatic conditions and the variation in water consumption reflected by climatic factors which are also vary.
Generally, to account for the climatic effect on the domestic water demand, the towns are grouped based on their similarity on average annual rainfall (ARF) characteristic.Group A towns with ARF of less than or equal to 900 mm a climatic factor of 1.1, group B towns with ARF of 900 to 1200 mm the factor of 1.0, and group C towns with the ARF of greater than or equal to 1200 mm the factor of 0.9 are considered.Hence, Hawassa city has average annual rainfall of 975 mm, the factor of 1% increase in domestic water demand is used to adjust the climatic effect.With the evidence of scientific studies, a 1.5% increase in domestic water consumption per 1 °C increase in T _ max is suggested for the study area to projected periods.The change in temperature and its effect on domestic water demand in the city will be noticeable during the anticipated period.
Changes in rainfall characteristics directly influence surface runoff generation at different return periods.A series of rainfall depths of return period 2-to 50-years (figure 4) are selected to represent the effect of change in rainfall on surface runoff depth in the urban area.Table 6 summarizes the change in runoff depth of return periods 2-, 5-, 10-, 20-, and 50-years at 24-hour rainfall duration and 2021 land cover condition of the selected urban watersheds.It is observed that an increase in the return period from 2-to 50-years results in an increase in surface runoff depth in the study area.The probability that runoff will exceed in any year for return periods 2-, 5-, 10-, 25-and 50-years, respectively, are 50, 20, 10, 4, and 2 percent.Thus, the safety design standards are a threshold capacity of the system against flooding risks, which are usually based on a probability of exceedance of a certain runoff level and economic design consideration.For a more realistic assessment of the existing urban drainage system, its design capacity is limited to a return period of 25-years under the year 2011 urban land cover condition of Hawassa (HCA 2018b).The capacity and coverage of the system also vary spatially with the historical growth pattern of the city and the progressive construction of drainage structures.A study by Zhou et al (2017) indicates the designed runoff frequency for minor drainage systems is often 2 to 5 years in developing countries.The threshold capacity of the system is a safety design that prevents flooding risks resulting from climate change with a probability of exceedance of a certain runoff level of consideration.However, the urban drainage infrastructures are scheduled for renewal in the short-term reconstruction to cope with upcoming changing conditions.This implies that the climate change effect was not considered to predict rainfall intensity or depth, which is necessary for the design of a drainage system.
The urban drainage systems are designed to serve a maximum rainfall value based on historical meteorological data.The climate change effect should be integrated to update Intensity (Depth)-Duration -Frequency curves.Further, continuous research on local-scale meteorological representation is needed, which improves insight into the physical process of short-duration rainfall patterns in urban watersheds.Essentially, there is a lack of detailed knowledge on changing variables at the urban scale, and spatial planning strategies often rely on past experiences of dealing with similar risks.Fuldauer et al (2022) and Moure et al (2023) suggest that no-regret strategies are beneficial when dealing with climate uncertainty, as the cost of adaptations is relatively low compared to potential climate risks.Climate impact adaptation strategies should be incorporated into the design of stormwater infrastructures to cope with future changing conditions.
On the other hand, spatial land cover change analysis result indicates that an increasing trend of surface runoff is associated with an increase in impervious areas.Rapid sprawl urban expansion has increased in builtup areas and gives rise to surface imperviousness, increasing surface runoff.Like changes in local rainfall characteristics, spatial urban land cover variation affects surface runoff magnitude.Figure 8 illustrates SCS-CN surface runoff depth for the selected urban watersheds with >10% ILC change (table 5) from 1991 to 2021 at a 25-year return period and 24-hour rainfall depth.
Owing to an evidence-based approach, the effect of rapid urban growth has been observed during the last 30 years.The increasing trend of surface runoff depth is associated with a change in surface imperviousness of the study areas.The surface runoff depth increased by 30.7% (509 to 665.2 mm) from 1991 to 2021 in the selected urban watersheds.The change in runoff volume is a product of the runoff depth and area of the selected urban watersheds.An increase in ILC signifies an increase in the runoff depth at urban watersheds, while agricultural and vegetation areas (low CN) generate less SCS-CN runoff depth and volume than urban/built-up areas.Studies by Miller et al (2014) and Luo et al (2022) have noticed that an increase in the percentage of impervious areas significantly increases surface runoff depth in urban areas.
This research mainly focused on quantitative aspects of UWM, particularly on spatial analysis of urban water demand and stormwater management in a changing environment associated with changes in local climate, urban watershed characteristics, and urban sprawl patterns.The existing planning approach traditionally often relied on constructing short-term water management infrastructures to cope with upcoming changing conditions.With the heterogeneous characteristic of urban watersheds, spatially explicit analysis of UWM is vital to better decision-making strategies when linked to spatial urban planning.The local-specific analysis of a unique characteristic of urban watersheds provides an opportunity for public involvement in the effective water management of developing cities (Kassay et al 2024).For instance, the neighborhood and household levels expose opportunities for the beneficial use of urban water forms using natural-based techniques.Hence, naturebased solutions are gaining importance to address urban sustainability, adaptation to changing conditions, and efforts to increase resilience means to promote a range of social, environmental, and economic benefits.
The research finding provides an imperative insight to assess the spatial impact of changing conditions, which is an integral part of identifying adaptation strategies to local-specific urban watershed characteristics.The spatial land cover variation has a significant effect on UWM for a small urban watershed, while at a large scale, the effect is more complex than the sub-watershed response.The trend of climate change is obvious, yet the magnitude of future climatic effects may be uncertain.Further research is required to quantify climate change impact on a local scale to better spatial planning and decision-making strategies in urban watersheds.Spatial analysis on UWM challenge in changing environments leads to a new paradigm of specific-solution to specific-problem approach.Also, a detailed local-specific assessment provides quantitative clarity to integrate adaptive water management with spatial urban planning strategy in an uncertain future.Thus, the concern of spatial analysis of the UWM challenge is to increase resilience in the urban environment under changing conditions.

Conclusions and recommendations
This research analyzed spatial challenges of the UWM under changing environments in the urban environment of Hawassa City.The spatial variation of change in local climate and urban expansion pattern and its effect on urban water demand and surface runoff generation during the historical and future scenarios are considered.A statistically significant signal of change in temperature is detected in the historical period, and a noticeable change is observed in the projected scenario.Climatic elasticity in domestic water demand of 1.5% per 1 °C change in temperature is suggested for the anticipated period.Change in climate effect is integrated into the updating of Intensity (Depth)-Duration-Frequency curves with a return period of 2-to 50-years.Because of the rapid urban expansion pattern, surface imperviousness has increased by 30.9 km 2 from 1991 to 2021, where the coverage varies spatially.With an average growth rate of 8.9% in the built-up, the city's urban area will cover 73.6 km 2 (45.9%) for the predicted period.
The SCS-CN empirical method has estimated the effect of change in rainfall depth and spatial land cover dynamics on direct surface runoff generation in small urban watersheds.When the return period increases from 2-to 50-years, surface runoff depth also increases significantly.The surface drainage system conveys runoff safely from urban areas.The system is often designed from maximum rainfall value using historical meteorological data without considering changing conditions.This induces uncertainty in coping with the adverse effects of climate change.The drainage capacity is constrained to a 25-year return period and the 2011 land cover of the city.The rapid urban growth has changed previous areas to impervious areas, and the surface runoff depth increased by 30.7% in the urban watersheds during the historical period.The increasing trend of surface runoff depth is also associated with the change in surface imperviousness of the study areas.Overall, the research results provide insight into the potential of spatial change in local climate and urban expansion pattern and their impact on urban water management aspects.
The urban areas have spatial variation, and a single meteorological station record may limit the spatial representation of the urban climate.A detailed assessment of climate impact at a local scale requires a high spatiotemporal resolution model and an innovative approach to represent the urban climate.In this study, the urban land cover projection method provides insight into the effect of change in urban land cover on UWM.Future studies should include urban growth simulation modeling techniques.An individual and combined effect of change in local climate and land cover dynamics increasingly cause spatial-specific challenges of UWM in urban areas.Integrating spatial planning with UWM will reduce the vulnerability of urban areas.This is precedent to insight on a new paradigm to specific -solutions to specific-problems approaches that determine mitigation and adoption strategies to changing conditions.This research will recommend that recognizing the possible spatial effect of changing environments is a vital concern in reducing the challenges of UWM under future uncertainties.These can help to plan innovative mitigation and adaptation strategies to maintain the sustainability of the urban environment in developing cities.

(
1990-2021) daily rainfall and temperature time series obtained from the National Meteorology Agency (NMA) of Ethiopia.Considering the historical urban growth pattern of the city, four cloud-free satellite imageries of Landsat 30 m for 1991& 2006 and Sentinel-2 10 m resolution for 2016 & 2021 are acquired from the US Geological Survey (USGS official website).Worldwide Reference System (WRS) path 168 and 55row, and the Universal Transverse Mercator (UTM) zone of WGS_1984_UTM_Zone_37N cover the whole study area.The DEM is processed to delineate the city into sub-watersheds.Google Earth image and GPS ground truth data are collected for image classification accuracy assessment.The major soil groups in the study area are identified based on Food and Agricultural Organization (FAO) soil classification system, and the soil map is obtained from Ministry of Water, Irrigation and Energy of Ethiopia.

Figure 1 .
Figure 1.Map of the study area (Hawassa City).
3.1.Trend of change in rainfall time seriesThe temporal trends of daily annual rainfall (DARF) and daily maximum rainfall (DMRF) for the historical period and future projection (2022-2051) are illustrated in figure 2. The DARF varies between 672 to 1277 mm with a mean value of 975 mm, and the DMRF ranges from 33.7 to 71.5 mm with a mean value of

Figure 2 .
Figure 2. The daily annual rainfall (a), and daily maximum rainfall (b) data for the historical 1990-2021 and projection up to 2051.
. The lower negative values indicate that DARF and DMRF probability functions are slightly flatter, while the positive kurtosis value indicates the function is oval than the normal distribution.The skewness values for DARF and DMRF, respectively, −0.01 & 0.66 implies relatively symmetrical normality.The negative value suggests a symmetrical distribution form to the left, while the positive value is to the right of the mean value.The box plot (figure 3) illustrates the mean (μ) and median values (974.97 & 980.1) of DARF and (50.42 & 48.3) for DMRF are relatively closer to each other.Hence, descriptive statistical values indicate that both rainfall time series are consistent and normally distributed.
Further, MK trend test results for DARF and DMRF indicate a positive value of Kendall's tau (0.213 & 0.073) and Sen's slope (5.580 & 0.125), respectively, as shown in table 2. The trend of both DARF and DMRF show p-values of 0.096 & 0.575, respectively at α = 5%.The MK trend test result also agreed with the linear model, and a statistically insignificant (P > 0.05) trend is observed in both rainfall time series.The research result reveals an upward trend of DARF and DMRF but statistically insufficient evidence to conclusively link to local climate change during the study period.Mulugeta et al (2017) and Belay et al (2021) also reported comparable results in the change and variability of rainfall in the Lake Hawassa watershed including the city.Several studies have discussed that change and variability of rainfall patterns in regions are associated with various factors, such as global atmospheric circulation, geographical location, watershed characteristics, and land use change, which differ from location to location.Wubaye et al (2023) discuss spatial variations of topographical characteristics and land cover change in urban areas influence daily maximum rainfall events.

Figure 5 .
Figure 5. Maximum temperature(a), and Minimum temperature (b) data of the years 1990 to 2021 and projection period 2022-2051.
) shows the mean and median values(27.55& 27.51) and (13.27 & 13.16) for T_ max and T_ min, respectively, which are closer to each other, and indicate consistency and normality of the distribution dataset.

Figure 6 .
Figure 6.Spatial extent and urban expansion pattern of Hawassa between 1991 to 2021.

Figure 7 .
Figure 7.The relation between change in maximum temperature and urban land cover in Hawassa.

Figure 8 .
Figure 8.The SCS-CN surface runoff depth for the selected urban watersheds with >10% ILC change, 25-years return period and 24hour rainfall depth 1991 to 2021.
Accuracy statistical values of forecasting for DARF and DMRF, respectively, were Mean Absolute Square Error (MASE, 1.08 & 0.57), Mean Absolute Percentage Error (MAPE, 0.15 & 0.16), Mean Absolute Error (MAE, 144.7 &7.82), and Mean Square Error (MSE,173.59& 9.65).These values are small enough to indicate an acceptable level of accuracy in the forecasting result.Normality analysis result (table 1) shows the kurtosis values of −0.57& −0.26, respectively, for DARF and DMRF, both rainfall tails are relatively as appropriate to the normal distribution curve 50.42 mm for the historical period.The descriptive statistics and variability value for the historical period in table 1 summarize, that the standard deviation (σ) statistic with low values (153.0 and 9.83) for DARF and MDRF, respectively implies less variability.The values of CV < 20, also show low variability of both rainfall time series in the study period.

Table 1 .
Descriptive statistics, variability, and normality analysis of urban climate time series.DARF = average daily annual rainfall, DMRF = daily maximum rainfall, T_ max = maximum temperature, and T_ min = minimum temperature.

Table 2 .
Mann-Kendall trend test/Two-tailed test results of DARF and DMRF time series.
and projection periods (2022-2051) of Hawassa station.T_ max varies between 26.97 to 28.55 °C, and T_ min ranges from 12.31 to 14.81 °C.The mean values are 27.56 & 13.27 °C, respectively, for T _max and T _min to the historical period (table 1).The smaller statistical values of σ 2 (0.18 & 0.51) and CV < 20 for T_ max and T _min, respectively, indicate a consistency of temperature time series and are close to their mean and average of the distribution data are reliable.

Table 3 .
Mann-Kendall trend/Two-tailed test results of T_ max and T_ min time series.

Table 4 .
Land cover class coverage and change between 1991 to 2021 of Hawassa.

Table 5 .
Land cover change of the selected urban watershed of Hawassa from 1991 to 2021.

Table 6 .
Return period 2-to 50-years and surface runoff depth at 24-hour rainfall and 2021 land cover.