Assessment of the Impact of Climate Change on Floods Using Hydrological ‎Modeling

Climate change has changed the number and severity of flooding events worldwide [1], causing many problems for communities, especially in arid regions [2]. Jordan has desert climate [3]. Jordan's low greenhouse gas (GHG) emissions make it a negligible contributor to global climate change [4], but climate-related disasters like flash floods, landslides, rock falls, and droughts have killed 110 people, affected hundreds of thousands, and caused significant economic losses over the past 30 years [5]. Thus, successful disaster management and mitigation require accurate flood risk estimation based on flood-frequency analysis, which uses past rainfall and flood data to predict future flood magnitude and recurrence [6].

Current population trends show that global urbanisation has major direct and indirect effects on climate and weather. Under global climate change and variability, extreme events are more likely. Floods, unpredictability in rainfall patterns, and heat wave variability may be caused by climate change [7], which may affect dam design and operation. Thus, these changes affect hydropower and water supply. Thus, accurate identification of how climate change affects flood events is essential for drought risk management and flood control [8].

Hydrological modelling helps understand and predict climate change's effects on flood events by simulating diverse hydrological processes in changing climates [9]. Research objectives, data, and watershed attributes determine hydrological model selection [10]. The hydrologic engineering center's hydrologic modeling system (HEC-HMS) model has been widely used to model rainfall-runoff processes to assess how climate change affects flood behaviour [11]. Using HEC-HMS to assess flood risk in the Bagmati river basin in Nepal, high-emission scenarios increased peak flow, resulting in rainfall-runoff correlations and a flood discharge modelling tool [12]. Using the HEC-HMS model, the study [13] predicted climate-induced streamflow variability in the Blue Nile Basin. Another SWAT model study examined how climate change affects streamflow in the Mesoscale Rur watershed in western Germany, highlighting the risk of floods under higher precipitation levels [14].

HEC-HMS, designed by the U.S. Army Corps of Engineers, is widely used to simulate watershed-scale hydrological processes and evaluate flood dynamics [15]. HEC-HMS has been a versatile tool for climate change impact assessments, watershed management, and flood forecasting worldwide since its founding [16]. It can model precipitation-runoff relationships, channel flow, and flood hydrographs for many hydrological conditions and regions [17]. Users can evaluate storms and land-use changes using the model's basin modelling, meteorological modelling, control specifications, and simulation results analysis [11]. By calibrating and validating simulated outputs with real data, HEC-HMS improves flood risk assessment and water resource planning [18].

HEC-HMS has been used worldwide to model hydrological processes, flood risks, and runoff [19]. Al-Mukhtar and Al-Yaseen [20] simulated peak discharge and volumes for the Ngong River basin in Nairobi utilising HEC-HMS. The study [21] applied the model to evaluate flood management measures in the Audi-União District in Brazil. HEC-HMS also assessed how land-use changes affected the Kan watershed in Tahran, Iran [22].

Due to climate projection uncertainty, several models have been developed to guide hydrological events under climate change [23]. The coupled model intercomparison project (CMIP) has progressed through several phases to improve climate knowledge [24]. CMIP Phase 6 (CMIP6) presents an updated set of scenarios called shared socioeconomic pathways (SSPs) [25], which improve on Representative Concentration Pathways (RCPs) from CMIP5 [26]. By varying radiative forcing (2.6-8.5 W/m² by 2100), RCPs were used to study potential greenhouse gas emission pathways [27]. The more integrated SSP approach combines socioeconomic factors with RCP-style radiative forcing levels [28]. This allows for more advanced climate change research, especially on hydrological extremes like floods [29]. SSPs allow researchers to assess climate change's wider effects under different emission and adaptation scenarios by including economic development and policy reactions [30].

This research employed four SSPs-SSP126, 245, 370, and SSP585- to evaluate prospective flood hazards in the Wadi Al Mujib Basin. These scenarios were selected to examine a wide range of possible climate circumstances, from a low-emission, sustainability-oriented future (SSP126) to a high emission, worst case scenario (SSP585). Although previous research has investigated the effects of climate change on flooding using RCPs, there is a lack of studies that combine HEC-HMS hydrological modeling with CMIP6 SSPs in arid and semi-arid areas such as Wadi Al Mujib. This work addresses the gap by delivering more sophisticated prospective flood hazards, providing valuable insights for adaptive flood management and policy formulation in Jordan.

3.1 Model setup

HEC-HMS Version 4.12 was employed to simulate rainfall-runoff dynamics in the Wadi Al Mujib Basin. The model consists of a basin model that depicts watershed characteristics, a meteorological model for climatic inputs, and control specifications for establishing simulation parameters. The model development began with the import of a 30-meter Digital Elevation Model (DEM) to delineate sub-basins and extract hydrologically relevant features using the Sink Fill method.

Thirty-one sub-basins and fourteen reaches were identified (Figure 3). Curve Number (CN), land use, soil cover, and antecedent moisture conditions were derived by using spatially distributed LCLU and HSG data. The CN method (Figure 3) was employed for rainfall-runoff transformation to ensure the model captured spatial and temporal variations in infiltration and runoff generation. Precipitation input was based on historical rainfall data (1980-2023) stations. However, areal precipitation was computed by assigning representative weights to each gauge by using the Thiessen polygon.

ping_mu_jie_tu_2025-05-08_171629.png

Figure 3. Curve number and impervious percentage for each subbasin, reaches, discharge and rainfall stations

Muskingum method, for flow routing, was selected due to its suitability for channel storage representation in the basin. Whilst Clark Unit Hydrograph method was used for runoff transformation, and for reflect the watershed response dynamics concentration time (T_c) and storage coefficient (R) were incorporated.

T_c and R in relation to precipitation and watershed properties are:

${{T}_{c}}=2.2\times {{(\frac{L*{{L}_{c}}}{\sqrt{Slop{{e}_{{{10}^{-85}}}}}})}^{0.3}}$                   (1)

$\frac{R}{{{T}_{c}}+R}=0.65$  or$\text{ }\!\!~\!\!\text{  }\!\!~\!\!\text{ }R=\frac{13}{7}{{T}_{c}}$                      (2)

These parameters were determined based on watershed characteristics, where L denotes the longest flow path mile, $L_c~​$ is the centroidal flow path in mile and slope 10^-85 is average slope of the flow path represented by 10 to 85 percent of the longest flow path (ft/mi) [31, 32].

3.2 Model calibration and validation

March 1991 and January 1998 floods were among the most notable in the seven-year streamflow statistics at Wadi Al Mujib Basin station CD0041. Several factors led to the selection of these events for HEC-HMS model calibration and validation. First, this period's minimal land use and land cover changes-maintained watershed characteristics. The short time between these events reduced the likelihood of significant changes in other watershed attributes like soil properties or basin slope. The March 1991 event was used for model calibration. Parameters influencing hydrological processes, such as the CN in the SCS method, time of concentration (T_c), and storage coefficient (R) in the Clark Unit Hydrograph method, were adjusted to ensure the simulated discharge accurately corresponded with the observed data regarding of peak discharge, hydrograph shape, and timing of the peak. Calibration trials employed the Peak-Weighted RMS Error objective function with the Univariate Gradient method to minimize discrepancies systematically. The Nash-Sutcliffe efficiency coefficient (E) was used to evaluate the model's performance quantitatively is shown in:

$E=1-\frac{\mathop{\sum }_{i=1}^{n}{{\left( {{Q}_{o,i}}-{{Q}_{s,i}} \right)}^{2}}}{\mathop{\sum }_{i=1}^{n}{{\left( {{Q}_{o,i}}-\overline{{{Q}_{o}}} \right)}^{2}}}$                    (3)

where, ${{Q}_{o,i}}$ denotes observed discharge at time step i, ${{Q}_{s,i}}$ is the simulated discharge at time step i,$~\overline{{{Q}_{o}}}$ is mean of the observed discharges and n is the total number of time steps.

Nash-Sutcliffe efficiency coefficient ranges from -∞ to 1, when it equal is one, that means simulated and observed peak discharge perfectly match, but when its value is zero this implies model predictions are as mean of the observed data. Whereas negative values indicate that the observed mean is better than model [33].

5.1 LCLU and HSG

LCLU classifies the scene as desolate. The Wadi Al Mujib Basin was 94% bare ground, 3% built-up, and 2% agricultural. This indicates little farming. The remaining 1% includes trees, water, and land uses. This distribution shows the basin's arid and semi-arid climate, which limits vegetation and agriculture. Group D soils with high runoff potential and low infiltration cer most of the basin, according to the HSG study. The basin's hydrological response and flooding susceptibility are affected by sparse land cover, heavy rainfall, and surface roff, wch generate and strengthen floods.

En though Wadi al Mujib doesn't have a lot of developed land, human activities like building cities, improving infrastructure, and changing the way farms work are changing the way water flows through the basin. Meanwhile, the expansion of impervious surfaces like roads will likely reduce infiltration while increasing runoff, therefore aggravating flood risks. Additionally, agricultural practices, although limited in the study area, may contribute to soil degradation and reduced water retention capacity.

5.2 Projection and observed precipitation

Cibrated precipitation data from 1900 to 2014 was used as the control scenario for comparison. In the calibration phase, lear scaling and bias correction reduced differences between simulated and observed precipitation data, improving the baseline model's accuracy. The calibrated model projected 2015-2100 precipitation trends under several SSPs.

In Figure 5, the Wadi Al Mujib Basin's precipitation frequencies for different return intervals and durations show clear patterns. The 100-year return period precipitation amounts for 1, 2, 3, 4, and 5 days are 53.15 mm, 90.88 mm, 112.91 mm, 121.94 mm, and 127.90 mm. The fifth day of 100-year precipitation is 15.96 mm (13%), while the fifth day of 50-year precipitation is 7.48 mm (7%). These findings show that precipitation intensities increase with duration, but over longer periods, the difference between durations becomes irrelevant. Data shows that most of the 5-day precipitation occurs in the first few days due to their higher quantities, but they are not significantly different from shorter periods.

image008.png

Figure 5. Observed precipitation in Al Mujib Basin

Figures 6-9 show that precipitation intensities rise with longer return times and durations for the selected SSPs. In the SSP126 scenario Figure 6, 100-year precipitation values for 1, 2, 3, 4, and 5 days are 73.72, 108.78, 125.34, 135.29, and 141.78 mm. The 5-day 100-year precipitation is 4.8% higher than the 4-day. In Figure 7, SSP 245 precipitation amounts to 86.13, 122.8, 142.77, 152.35, and 160.26 mm. The 100-year precipitation values for 1, 2, 3, 4, and 5 days in the SSP370 scenario Figure 8 are 78.10, 110.47, 136.03, 152.17, and 162.11 mm. The 5-day precipitation is 6.5% higher than the 4-day. The SSP585 scenario (Figure 9) predicts 90.74, 133.20, 156.68, 165.27, and 177.04 mm of 100-year precipitation, 7.1% more than the 4-day event. As emission scenarios increase from SSP126 to SSP585, precipitation rises continuously, indicating that climate change is affecting expected precipitation patterns.

image009.png

Figure 6. Precipitation in Al Mujib Basin with SSP126

image010.png

Figure 7. Precipitation in Al Mujib Basin with SSP245

image011.png

Figure 8. Precipitation in Al Mujib Basin with SS370

image012.png

Figure 9. Precipitation in Al Mujib Basin with SSP 585

5.3 Hydrologic model setup, calibration and validation

5.3.1 Hydrologic model setup

Figure 10 reveals differences between observed and simulated discharges, with the simulated values typically exhibiting greater peaks in some years. for instance, the simulated discharges are much higher than the observed discharges for the years 1984, 1990, and 1992.

image013.png

Figure 10. Observed and simulation annual peak discharge

The highest discrepancy is seen at 23 March 1991with value 411.6 CMS and with same time the maximum observed discharge along observed period is 288 CMS.

5.3.2 Hydrologic model calibration and validation

The hydrological model was calibrated by means of simulated discharge comparison comparatively to observed discharge data at the watershed outflow for the flood event in March 1991, Table 2 and Figure 11 shows that although the simulated peak discharge was 325.7 cubic meters per second (CMS), the measured peak discharge for this event was 287.6 CMS, therefore producing a 13.24% difference. For both observed and simulated data, the peak discharge fell on March 23, 1991, at 01:00. With a 0.619 Nash-Sutcliffe efficiency for this calibration, the simulated and observed discharge values fit somewhat moderately.

image014.png

Figure 11. Calibrated peak discharge over time by HEC-HMS

Table 2. Summery result for March 1991 flood event

On the other hand, for model validation purposes, was tested with the flood events in November, December, and January of 1998. Table 3 summarizes the results of this validation process. As illustrated in Figure 12, the observed peak discharge for the January 12, 1998, event was 201.5 CMS, while the simulated peak was 232.7 CMS, showing a difference of 15.48%. At the same time the peak discharge occurred at 23:00 on January 13, 1998, according to the simulation, which was slightly later than the observed peak. Nevertheless, the Nash-Sutcliffe efficiency for this validation was 0.819, suggesting a good agreement between the simulated and observed data.

image015.png

Figure 12. Validation summery result of Al Mujib Subbasin

Table 3. Validation summery result for November, December, and January 1998 flood event

Though there are variations in both peak discharge values and timing, the calibration and validation findings show generally that the model can fairly match the observed discharge patterns. These findings highlight the need of optimizing model parameters for correct forecasts in next flood occurrences.

5.4 Shifts in flood streamflow due to climate change

Precipitation data for four different SSP scenarios (SSP126, SSP245, SSP370, and SSP585) was utilized in the calibrated hydrological model to evaluate the effects of climate change on flood occurrences in the study area. Thus, flood flows were estimated under the historical observations and the four SSP scenarios as presented in Figure 13 for return periods of 2, 5, 10, 25, 50, and 100 years.

image016.png

Figure 13. Observed and projections peak discharge with various return periods

Floods vary across return periods, showing a steady rise. The SSP126 scenario estimates 487 CMS for the 100-year flood, 22.0% higher than 399.4 CMS. The 50-year flood in the SSP126 scenario is 412 CMS, up 14.9% from 358.8 CMS. However, the 10-year flood is 327 CMS, 24.5% higher than 262.7 CMS. Compared to 153.1 CMS and 219 CMS, the two-year and five-year floods increased 9.7% and 21.9%, respectively. In the SSP245 scenario, flood discharge increases significantly from observed to expected values for all return periods. The 100-year flood discharge under SSP 245 is 572 CMS, up 43.2% from 399.4 CMS. The 50-year flood under SSP 245 spikes 37.3% to 492.6 CMS, matching the observed 358.8 CMS. The SSP 245 scenario projects 42% increases for the 2-year flood and 72% for the 100-year flood, rising flood risks above past records.

For the SSP370 scenario, throughout all return periods the rise in flood stream flows from the observed data to the future forecasts is noteworthy. While the 5-year flood rose by 35.7% (297 CMS) the 2-year flood discharge rose by 45.0% (511 CMS compared to 399.4 CMS). The 10-year flood saw an increase of 32.9% (349 CMS compared to 262.7 CMS), and the 25-year flood increased by 32.5% (421 CMS compared to 317.8 CMS). The 50-year flood saw a 21.2% increase (434.9 CMS compared to 358.8 CMS), and the 100-year flood increased by 28.0% (511 CMS compared to 399.4 CMS).

These rises point to a notable increase in flood risk. The 50-year flood discharge for the SSP585 scenario is far higher than the 25-year flood. While the 25-year flood is 531 CMS, suggesting a roughly 6.3% rise, the 50-year flood discharge is 564.4 CMS. The SSP585 scenario projects a discharge of 614 CMS when compared to the measured 100-year flood of 399.4 CMS, 54% more. Furthermore, the 100-year flood under SSP585 surpasses the 10-year flood (361 CMS) by 70%, so stressing the significant rise in flood hazards under high emission scenarios.

Overall, the SSP 585 scenario under all return periods exhibits the highest flood stream flows compared to the recorded values. In addition, the peak discharge ranges for the 2-year flood are from 24% to 54% and for the 100-year flood compared to the observed values. With the SSP 245 and SSP 370 scenarios showing more moderate flood stream flows, the SSP 585 scenario, especially at the 100-year return period showcases the biggest rise.

5.5 Relation between precipitation scenarios and peak discharge changes

The precipitation and flood streamflow data for return periods of 2, 5, 10, 25, 50, and 100 years are investigated and shown in Figure 14 to help one grasp how flood characteristics vary over several SSP scenarios. Precisely for the SSP scenarios, precipitation values often surpass the matching observed values during all return periods. With a precipitation during the 2-year return period of 24.08 mm in the SSP126 scenario, the precipitation shows a 29% rise over the recorded figure of 18.63 mm. Likewise, under the SSP245 scenario, from 53.15 mm to 86.13 mm the 100-year precipitation rises by 61%. With values of 57.53 mm against the recorded 47.45 mm, the precipitation rises by 52% in the SSP370 scenario at the 50-year return period.

image017.png

Figure 14. Observed and projections maximum precipitation with various re-turn periods

Peak discharge values rise most in the SSP585 scenario. At SSP585, the 100-year flood discharge is 614 CMS, 54% higher than 399.4 CMS. This increase exceeds the SSP126 100-year flood's 31% increase to 487 CMS. In particular, the high-emission SSP585 scenario shows that flood stream flows are more responsive to SSP scenarios than precipitation, with flood discharges increasing more than precipitation. These findings suggest that greenhouse gas concentrations will rapidly increase flood hazards, affecting flood control.

5.6 Study limitations and uncertainties

Despite of this study assessment climate change impacts on floods in the Wadi Al Mujib Basin, but must be acknowledged several limitations and uncertainties:

1). Uncertainties in Downscaling Climate Data.

The projections of precipitation in this study rely on downscaled CMIP6 climate models under various SSP scenarios. While to improve model accuracy, bias correction and linear scaling techniques were applied inherent uncertainties in downscaling remain. These uncertainties arise from spatial resolution limitations, and potential biases in future climate projections, which could influence in pattern of precipitation like intensity and frequency.

2). Precision of HSG Data.

Hydrological soil groups affect infiltration rates and runoff. Soil datasets may display spatial inaccuracies resulting from inadequate data or low resolution. This may lead to inaccurate estimations of infiltration rates and, consequently, variations in anticipated peak discharge.

3). Assumptions in the HEC-HMS Model.

HEC-HMS model is employed in hydrological simulations based on various assumptions, that may include some principal constraints:

· The model assumes uniform rainfall distribution over each subbasin, which may not reflect variability in storm events.

· For runoff estimation, SCS-CN method does not fully capture the dynamic changes in soil moisture, particularly in arid and semi-arid area.

· The routing method applied in the model (Muskingum method) may introduce uncertainties in peak discharge predictions, particularly for extreme events.

4). Calibration and Validation Constraints.

The model was calibrated and validated based on historical flood events; nevertheless, the availability of extensive high-resolution discharge data is constrained, so the moderate Nash-Sutcliffe Efficiency (NSE) values indicate differences between actual and simulated values, either arising from data inconsistencies, measurement errors, or neglected hydrological processes such as groundwater interactions.

The model was calibrated and validated based on historical flood events; nevertheless, the availability of extensive high-resolution discharge data is constrained, so the moderate NSE values indicate differences between actual and simulated values, either arising from data inconsistencies, measurement errors, or neglected hydrological processes.

5). Impact on Results.

These limitations introduce some degree of uncertainty in the projected flood risks under different SSP scenarios. Uncertainties in downscaling may affect precipitation intensity, while soil data limitations could influence runoff estimations. Additionally, HEC-HMS model assumptions may result in slight deviations in peak discharge predictions. Future studies should explore higher-resolution climate models and improved soil datasets.

In hydrological modeling, particularly when utilizing climate projections, uncertainty is an inherent challenge. Furthermore, the sources of uncertainty, like model parameter sensitivity and diversity in hydrological responses, require more investigation. For more comprehensive confidence levels for peak discharge values, future studies ought to integrate uncertainty quantification methodologies, such as Monte Carlo simulations or sensitivity analysis, and to assess a range of possible outcomes while improving the resilience of flood estimation and the confidence levels linked to anticipated peak discharge values.

Based on findings and conclusions of this study several recommendations can be proposed:

1. Develop and implement robust flood management plans that account precipitation variability includes upgrading existing infrastructure to be withstand higher flood volumes and intensities, this involves improving water storage capacity and promoting water conservation practices especially to mitigate evaporation.
2. Increase awareness among local communities about the risks associated with intensified precipitation events. Implement early warning systems for floods and promote community-based disaster preparedness initiatives.
3. Advocate for and implement policies that support sustainable development practices and effective climate change adaptation strategies, and this includes incentivizing using climate smart agricultural practices, promoting reforestation efforts, and fostering international cooperation on transboundary water management and evaluation the effect of climate as national scale.
4. Develop spatial and temporal distribution models for precipitation, evaporation, and streamflow across Jordan, incorporating climate change scenarios to assess influence of climate change on the water budget include ground water, and flood risks.