Assessment of land use and land cover changes on hydrological responses in the Wadi Soummam watershed, Algeria using the HEC–HMS model

Floods are among the most common natural disaster phenomena on Earth (Abdelkebir et al. 2021). Floods are caused by several factors, the most famous of which are climate change and unregulated urban growth (Tesfaw et al. 2023). Floods can result in human and economic losses, which causes a deterioration of the standard of living for people and disrupts even the efforts of countries to achieve sustainable development (Mokhtari et al. 2023). Experts and researchers, supported by decision-makers in various fields, have explored and valued modern methods and strategies for assessing and defining the frequency of floods as a first step, and as a second step proposing effective flood control strategies, and as a final step proposing effective strategies for protection and prediction from flash floods. Flood event analysis is important for assessing and understanding catchment hydrological response. To predict the hydrologic response of a catchment, we have performed hydrological modeling using a Geographic Information System (GIS) to extract the main inputs for the simulation.

We used a rainfall runoff model in our study to describe the relationship between rain and runoff. There are numerous models available, each with its own set of strengths, weaknesses, and operating conditions including: HEC–HMS model, Cross-wavelet analysis, SWAT model and Deep learning method for forecasting (Derdour et al. 2018; Sertel et al. 2019; Xu et al. 2022; Ghaderpour et al. 2023).

We selected Hydrologic Engineering Centers Hydrologic Modeling System (HEC–HMS), designed by the United States Army Corps of Engineers, one of the most well known of these models in the rainfall runoff process (USACE).

The purpose of the HEC–HMS mathematical model is to approximate the entire hydrologic phase of a dendritic catchment. In either natural or regulated watersheds, the model can simulate rainfall–runoff and watershed routing processes. It helps to predict peak flows, runoff volumes, and hydrographs in the basin (Mokhtari et al. 2016). The hydrographs produced by the program can be used in water supply studies such as flood forecasting, water distribution, urban drainage planning, and reservoir design. The shift in LULC coupled with irregular urbanization would undoubtedly contribute to the emergence of floods (Apollonio et al. 2016). The effects of changes in LULC in the catchment could be estimated using a combination of HEC–HMS and GIS to understand the increase in stream flow in river systems (Ramly & Tahir 2016).

Numerous studies have demonstrated the ability and efficiency of the HEC–HMS model to assess the impact of LULC change in various regions around the world (Ali et al. 2011; Zope et al. 2016; de Moraes et al. 2018; Koneti et al. 2018; Hu & Shrestha 2020; Azizi et al. 2021).

The HEC–HMS model was selected for this study because it was found to be accurate in evaluating the hydrological response of the catchment for event simulation. The purpose of this research is to evaluate the impact of land use change on the hydrological response of a Soummam Catchment with event rainfall. This research is critical because the Wadi Soummam watershed is one of the Soummam basin's raw water sources.

The Soummam watershed is primarily composed, on the left bank, of Oligocene formations intersected by lower Cretaceous formations and Miocene formations appearing in the downstream area along the riverbank. Generally, significant alluvial terraces cover the foothills, except in the region of Sidi-Aïch where the Cretaceous formations extend into the riverbed. On the right bank, it is mostly composed of middle and upper Cretaceous formations. The climate in the area is predominantly Mediterranean. The Soummam valley experiences a humid climate with slight seasonal temperature variations. On the plateaus of Sétif and Bouira, the climate is continental and dry, characterized by cold winters and hot summers. The southern part of the Sétif plateau ranges from sub-humid to semi-arid.

Predicting future changes in LULC and their effects on the hydrological cycle necessitates the use of various models and tools. These can include:

• - Remote sensing data and image analysis to track changes in land cover over time.

• - Land use change models that simulate future scenarios based on past trends, demographic and socioeconomic factors, and policy measures.

• - Hydrological models that calculate the effects of LULC changes on the water cycle, such as precipitation, evapotranspiration, runoff, and groundwater recharge.

The primary objective of this study was to demonstrate the applicability of the integrated CA-Markov hybrid model within a GIS software. The study aimed to assess the effectiveness of CA-Markov hybrid models in predicting land use changes, specifically focusing on predicting land use changes in the Oued Soummam watershed for the years 2030 and 2050. This prediction was carried out using the CA-Markov model and its impacts on hydrological responses were evaluated through the utilization of the HEC–HMS hydrological model.

The outcomes of these predictions can help inform decision-making and to mitigate the effects of LULC changes on the hydrological cycle, such as increased runoff and flooding, decreased water availability, and altered water quality. It is critical to consider multiple scenarios as well as account for uncertainties and unforeseeable events. Furthermore, effective strategies for sustainable land use and water management require close collaboration among hydrologists, ecologists, planners, and other stakeholders.

The Soummam watershed is in northern Algeria, between longitudes 3°38′ and 5°38′ East and latitudes 35°38′ and 36°45′ North. The Soummam watershed runs along a north-east-south-west axis. It is composed of three main regions: the plateaus of Sétif, the plateaus of Bouira and the valley of the Soummam valley. It is limited

• to the North by the mountains of the great Kabylie (Djurdjura massif)

• to the East by the mountains of the small Kabylie

• to the South by the mountains of Bibans and Mansourah, to the South-East by the foothills of the Hodna;

• to the West by the Isser and Sebaou rivers. It has a very irregular shape.

The result of the confluence of two important wadis upstream of Akbou is the Oued Boussellam, which flows from the Setifian plateau (Fermatou) and is oriented toward the east, and the Oued Sahel, which flows from Bouira and is oriented toward the west. It flows into the Mediterranean Sea at Bejaia after running through the entire Soummam valley, which appears as a sinuous strip oriented South-West, North-East. It has an average flow of 25 m³/s, but its floods are violent and devastating. It drains a large catchment area of 9,125 km², and its minor bed develops in a valley between two mountainous massifs: Tizi-Ouzou to the north-west and Bejaia and the Bibans chain to the south-east.

Wadi Soummam's watershed is part of the Soummam basin (Figure 1(b)). It is situated in the estuary Soummam basin at an average altitude of 747.27 m. It has a drainage area of 1,059.21 km². Wadi Soummam flows for more than 75 km, reaching an altitude of 1,683 m, with the lowest point at the outlet at 11 m. The valley bottom has an average length of about 2 km and can be as short as 100 m upstream of Sidi-Aich and a widening of 4–5 km in the region of El-Kseur and the Bejaia plain.

HEC–HMS is a semi-distributed rainfall–runoff model that can simulate a single study catchment, as in our case, or several catchments at the same time.

The HEC–HMS model is made up of four main model components: the base model, the meteorological model, the control specifications, and the input data (Gebre 2015).

Rain should be considered for each event in the form of rainfall depth that fell on the watershed during the day where this event occurred, which we associate with one of the four NRCS distributions each time. In our case, we were restricted to the following dates: (01–10/05/2005) - (03–14/09/2005) - (21–31/05/2005) and (17–30/05/2005).

With recent advances in geographic information systems (GIS), it is now possible to integrate land use change models into GIS. Cellular automata and hybrid models are most commonly used for land use change modeling (Breuer et al. 2006). We use a hybrid CA-Markov model integrated into GIS software to predict future land use change based on historical spatiotemporal data. A primary goal of this study was to demonstrate the utility of hybrid CA-Markov models in predicting land use change. We intended to use a CA Markov Model to predict land use change in the Wadi Soummam Basin between 2030 and 2050.

and ⁠, j = 1,2, ……, m)) is the transition probability matrix.

The increase in forest area can be explained by a combination of active reforestation efforts, natural land conversion dynamics, policy changes, and urban planning strategies that incorporate green spaces. These factors collectively contribute to the observed and predicted growth in forest areas over the years. In particular, conservation and reforestation policies implemented by the government have played a crucial role. Targeted reforestation programs, funded by local and international initiatives, have helped restore degraded ecosystems. Additionally, education and awareness campaigns for local communities have encouraged active participation in forest protection and restoration. Natural dynamics, such as ecological succession, have also favored forest regeneration in certain areas. Sustainable agricultural practices and crop rotation have reduced the impact on agricultural lands, allowing some areas to revert to a forested state.

Finally, the integration of green spaces into urban planning strategies has not only improved the quality of life for residents but also contributed to the increase in urban and peri-urban forest areas. These measures have created ecological corridors that facilitate biodiversity and carbon sequestration, thus enhancing the environmental resilience of the region.

The primary matrix is based on the calculated areas of the Landsat images' land use types. Table 2 shows the area statistics calculated for 2005, 2015, and 2020. Figures 6–8 depict the distribution of the land use class for these years, while Figure 8 depicts a graphical representation of the land use class for ease of interpretation and Figure 9 shows predicted LULC. Figure 8 highlights the progressive nature of the built-up areas as well as the declining nature of wetlands. P(o) = [473.39, 171.91, 385.74, 26.61] for further analysis with transition probability matrices in the generation of future LULC.

The entire watershed is clearly dominated by developed land, which includes residential and commercial services. This could be attributed to population growth, as well as other factors mentioned previously that draw people to the city center. In 15 years, the developed areas' land coverage has increased from 24.88 to 26.61 km². This rapid increase in built-up areas over a short period of time implies a reduction in the catchment's percolation and flood storage capacity, resulting in a recent increase in flood events and extent. This is concerning, which is why this study is necessary. Increased urbanization means less forest cover and wetlands.

The transition probability is defined as the rate of transition from one state to another over a given time period. It is calculated using the annual average transition rate of a specific land use and land cover type. Table 3 depicts a transition matrix for four land use types from 2005 to 2015. The transition probability of land use type in 2020 converted into land use type in 2030 and 2050 was calculated using Equation (11). Tables 3–5 show the primary transition probability matrix for four types of land use from 2005 to 2015, 2005 to 2020, and 2020 to 2040.

Based on the transition matrix shown in Table 3, the probability that forest areas remain forests over a period of 10 years is 44.9%. The probability of conversion from forests to grasslands and agricultural areas is 13.1 and 39.5%, respectively. For grasslands, the probability of remaining grasslands is 22.7%, while the probabilities of conversion to forests and agricultural areas are 34.3 and 42%, respectively. Agricultural areas have a 39% probability of remaining agricultural. The probability of conversion from agricultural areas to forests is 43.9%, and to grasslands, it is 15.6%.

Regarding built-up areas, they have an 8% chance of remaining the same. However, there is also a significant probability of conversion to forests (56.3%), grasslands (16.3%), and agricultural areas (19.4%).

Five storm events were selected in our study, two for pre-calibration and one for calibration and two for validation, these events correspond to the flow data of the stream data of the same date. The hydrograph was used to estimate initial flow based on discharge data obtained at the Wadi Soummam flow gauge. The impermeability of the study area was estimated using a GIS and a satellite image. Watershed factors such as initial abstraction and lag time must be adjusted to achieve the best match between simulated and observed flow. The model's output flows were calibrated using observed stream flow. The optimization option in HEC–HMS was used to calibrate the model. All parameters must have initial values at the start of the optimization. In HEC–HMS, starting values for the lag time were determined, and the initial abstraction in the research region was assumed to be 20.36 mm. To compare a calculated hydrograph to an observed hydrograph, HEC–HMS computes a goodness-of-fit index. The objective function of HEC–HMS algorithms is to find model parameters that provide the best value of an index (US Army Corps of Engineers 2018). The calibration and validation results, as well as the statistical parameters calculated from the simulation outputs for flow and volume, are shown in the following section of the results and discussion.

Following the successful calibration and validation of the HEC–HMS model, the volume and flow were calculated for different return periods (10 years, 20 years, 50 years, 100 years) for the concerned periods, namely 2005, 2015, 2020, 2030, and 2050.

To obtain optimal values for the model parameters, a calibration for the May 21–31, 2005 event was performed using the objective function on the peak flow.

Table 9 shows that with the new optimized parameter set, the model was able to correctly reproduce the peak flow for the events of September 23–26, 1994 and September 03–04, 2005. The performance criteria used to quantify the level of achievement of these various objectives are NASH, R², and RMSE. These provide overall assessments of the flood reconstruction. The model is validated based on the results in Table 11.

Nobody denies that climate change and land use alter hydrological processes and disrupt the natural flow of the environment. Thus, those in charge of planning and making decisions must understand how human activities like urban growth and deforestation and reforestation upstream of the watershed will affect the downstream environment.

This section attempts to repurpose the HEC–HMS model that was originally applied to the Wadi Soummam watershed in order to estimate its response to changing land use in 2005, 2015, 2020, 2030, and 2050.

The first scenario uses land cover to simulate the effect of rainfall events with different return periods on the discharge hydrograph at the Wadi Soummam station in 2005, 2015, and 2020. As a result, we replaced the event's average rainfall depth with depths estimated by statistical laws (Table 12).

According to Table 14 and Figures 16–18, we observe the effect of land use (LULC 2005, LULC 2015 and LULC 2020) on the variation of flows and volumes for different return periods. As an indication, the peak flow in 2005 is equal to 1,651.1 m³/s, in 2015 is 1,604.5 m³/s and in 2020 are 1,632.8 m³/s and for volumes, there is a remarkable variation between 2005 and 2020.The impact of rainfall events with different return periods on the flood hydrograph at the Wadi Soummam station is simulated in the second scenario using the LULC land use for the forecast years 2030 and 2050.

In Table 14, we see an increase in peak flow and volume for different return periods, demonstrating the effect of LULC on the hydrological response in the Wadi Soummam watershed. Knowing that we used the same rainfall amounts, we can clearly see that the peak flow has increased by about 39 m³/s (2030–2050) for the 10 year return period and by 70.8 m³/s for the 100 year return period.

The study conducted focused on the impact of land use changes on the future behavior of the Wadi Soummam watershed. This study is part of a series of research addressing the same topic. Previous studies have highlighted the importance of hydrological analysis and flood prediction in semi-arid regions. Among the previous studies, Derdour et al. (2021) examined the Ain Sefra watershed and successfully utilized the HEC–HMS model to simulate flood discharges. The model incorporated the storm frequency method, the SCS curve number for loss estimation, and the SCS unit hydrograph method for runoff simulation. The results demonstrated good agreement between simulated and observed values, indicating satisfactory performance in simulating runoff. In another study, Kastali et al. (2022) focused on the self-calibration of the HEC–HMS model, considering the uncertainty in the rating curve in the Allala watershed. By employing a Bayesian-based segmented BaRatin rating curve, the study revealed that accounting for rating curve uncertainty had a significant impact on model calibration parameters. This approach enhanced the model's capability to predict hourly flow hydrographs. Additionally, Haddad & Remini (2021) examined extreme runoff events during intense precipitation in the Koudiet Rosfa watershed using the HEC–HMS model. The study employed the Curve Number (CN) method to estimate excess precipitation and a parametric unit hydrograph model to calculate direct runoff. The model's performance in replicating observed hydrographs was evaluated using various measures, demonstrating its effectiveness in managing extreme runoff events. These studies underscore the importance of hydrological modeling and the potential of the HEC–HMS model in semi-arid regions. By accurately simulating runoff processes and accounting for uncertainties, these models contribute to improved water resource management and assessment of flood risks.

Changes in LULC are one of the factors that affect the water balance of watersheds by altering the magnitude of surface runoff. The Markov model tool was used to forecast LULC changes in 2030 and 2050, and the HEC–HMS tool was used to assess the impact of LULC changes on water resources in the Wadi Soummam watershed. The research included evaluating the HEC–HMS model, detecting LULC changes, and assessing the impact. The impact of LULC changes on water resources was successfully evaluated using available data. The model's overall performance was satisfactory. The NSE values were greater than 0.80, which is greater than the minimum requirement for using the model for further analysis. We simulated, calibrated, and validated the study using available flow data, and then demonstrated the current and future impact of LULC changes on peak flows and volumes for various return periods.

According to the findings of this study, changes in LULC have an impact on water resources in the study area. Changes in LULC have an implicit effect on hydrological response and will continue to have an impact on natural resource management and development. The built-up area in 2005, which accounts for 2.35% of the total area of the watershed, rises to 4.96% by 2050, while agricultural land decreases by about 7% between 2030 and 2050.

The predictions' outcomes can help inform decision-making and mitigate the effects of LULC changes on the hydrological cycle, such as increased runoff and flooding, decreased water availability, and altered water quality. It is important to consider multiple scenarios and to account for uncertainties and unpredictable events. Additionally, close collaboration between hydrologists, ecologists, planners, and other stakeholders is critical for developing effective strategies for sustainable land use and water management.

All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed by E.M. and B.A. The first draft of the manuscript was written by E.M. and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.