Introduction

Flood occurrences in the context of climate change have swept through several facets of the world (Armah et al. 2010; Aznar-crespo et al. 2021). Notably, the generation of floods through riverine systems overflows, and overland flows under extreme rains has posed an ever increasing threat to natural resources and humanity (CRED 2018; Sofia and Nikolopoulos 2020). Significantly, the challenge becomes evident and rampant when the general land surfaces and drainage network systems are unable to discharge or convey excess rainwater after heavy downpours, as population growth with rapid urban land uses changes (United Nations 2009; Le Polain De Waroux 2011; Sofia and Nikolopoulos 2020; Liu et al. 2023; Nearing et al. 2024; Rączkowska et al. 2024). In recent times, driving climate change and its increasing global warming have been ranked as the worst acceleration of environmental and hydrological extremities; floods and droughts (Garg et al. 2023; Bullen and Miles 2024; UNESCO 2023a, b; Nearing et al. 2024). Heavy rainfalls and land surface changes also play major roles in flood frequency and magnitude generation (Tarasova et al. 2023; Fischer and Schumann 2024). Though some studies have ascribed flood occurrences to hydro-meteorology (Tabari 2020; Alifu et al. 2022), perennial flood impacts, severity, and risks are also driven by three (3) core phenomena; vulnerability, hazards, and exposure (Hudson and Berghäuser 2023; Srivastava and Roy 2023). Globally, there have been enormous reports on the losses of properties worthy of millions of dollars and lives to flooding disasters (Islam and Wang 2024). On average, the global community recorded annual economic losses of 40 billion euros to flooding events in 2009 (Gaume et al. 2009). Additionally, in 2022, 176 flood disasters were recorded while between 2002 and 2021, 168 flood disasters were recorded at an estimated economic loss of $ 44.6 billion and 41.6 billion, respectively (CRED 2023). The higher record of floods in 2022 as compared to the two (2) decades (2002 to 2021), confirmed the worldwide ever increasing rate of flood occurrence.

Further evidence showed that flood disasters have been rated as one of the leading natural disasters in African countries, with severe rainfalls as the major cause (CRED 2019; Ahadzie et al. 2021; Alimi et al. 2022; Danso et al. 2024). Within the period of 19 years (1994—2013), Africa economically lost $ 912 billion to floods (UNISDR 2015). Approximately, 12 per cent of Africans experienced food insecurity caused by flood events between 2009 and 2020 (Reed et al. 2022). Flood events between 2010 and 2019, also accounted for 64 per cent of natural disasters in the continent (CRED 2019). Subsequently, parts of Africa experienced food insecurity, after crops were washed off by floods in 2022 (WMO 2022). As of December 2022, the United Nations Office for the Coordination of Humanitarian Affairs report for the year (2022) indicated; a death toll of 1418, while 4398 citizens suffered various degrees of injuries, and 2.9 million were also displacedin West and Central Africa (OCHA 2022). Flooding was ranked 56% of natural disasters in Africa with 43% of fatalities and 92% of economic losses (WMO 2022). Additionally, Tramblay et al. (2022) analysis of 13,815 historical flood records (1981–2018, 37 years) across Africa showed that 75% of them occurred as a result of intensive rainfalls. These further confirmed the perennial increasing rate of floods in Africa as a result of heavy rainfalls (Alimi et al. 2022) and it is therefore paramount for more studies to understand the mechanism of flood occurrences.

In Ghana, perennial flood disasters have been a national challenge, especially in the major cities, rural areas, communities, and cities within low-lying areas and or along river channels of the country (Dekongmen et al. 2021; Yin et al. 2021). Rapid rural–urban and urbanization have been accelerating these challenges through, unapproved settlements on waterways, paved surface areas, and poor waste management in Ghana, at the expense of potential flooding areas, thereby exposing many citizens to perennial flood disasters (Danso et al. 2024). However, in the rural areas, it is mostly the farmers who suffer the consequence of floods due to their predominant farming of rice and some water-like crop in low-lying areas, wetlands, and valleys (Smits et al. 2024). This confirmed that the frequency and severity of floods and other natural disasters increase as economic development and population densities encroach on these highly disaster risky zones (UNISDR 2015). In Ghana, flood disasters have been ranked as the highest natural disasters with a statistical representation of 32% from 2010 – 2019 (Kelly and Addo 2023). According to the Emergency Events Database (EM-DAT) from 1968 – 2023, floods accounted for 534 deaths and affected over 5,062,732 citizens within various parts of the country. The lower part of the Volta Basin of Ghana is yet to recover from 2023 floods that rendered many citizens along the basin homeless (IFRC 2023a). However, heavy rains are expected to worsen extreme floods in Ghana at a yearly projected cost of $ 160 million (The World Bank Group 2021).

The perennial flood cases in the major cities in Ghana have led to severally flood studies (Amoako and Boamah 2014; Frimpong 2014; Asumadu-Sarkodie et al. 2015; Twum and Abubakari 2019; Kordie et al. 2020; Osei et al. 2021a, b; Yin et al. 2021; Yin et al. 2022; Twum and Abubakari 2019; Osei et al. 2021a, b; Danso et al. 2024). However, none of such studies have been conducted in the Upper West Region (UWR) of Ghana. Worryingly, within the past decades, several episodes of these flood events have occurred in the UWR; a report from the EM-DAT indicated 52 losses of lives, with 29 injuries, 291,573 affected people, and home damages recorded at 33,000 in September 1999. National Disaster Management Organization (NADMO) records from 2016 to 2020 in the UWR, reported 553.5 acres of crop losses, and an affected population of 9,385, with an estimated cost of damage of Ghc 1,458,872.5. The most unbearable floods in 2021 (Fig. 1), led to massive unexpected destruction of homes, farms, roads, and even loss of lives. Shockingly, perennial spillages of excess water from the Bagre Dam in Burkina Faso coupled with extreme rainfalls are causing humanitarian challenges in the UWR and other Upper Regions of the country (Armah et al. 2010; Smits et al. 2024). These perennial floods have the potential to heighten food insecurity and disease outbreaks (Andrade et al. 2018; Iqbal et al. 2022; Rehan et al. 2023; Ayenew and Kebede 2023).

Fig. 1
figure 1

(Source: Ministry of Roads and Highways Ghana)

Examples of flood disasters in 2021 in the region (a) Road washed off; b Daffiama-Sanwie Road; c Nadwoli –Wa Road; d Truck en route to Burkina Faso

Full size image

Improving the situation in the region requires the provision of flood risk terrains information, which is crucial to support stakeholders in developmental planning, policy and decision-making, that will prevent urban dwellers from building on waterways. However, in recent years advanced datasets and methods; modelling approaches and techniques have evolved around the globe for conducting flood studies. These methods include but are not limited to analytical hierarchy process (AHP) and Geographic Information System (GIS) (Danumah et al. 2016; Cabrera and Lee 2019; Swain et al. 2020a, b; Ozturk and Safety 2021; Ramkar and Yadav 2021; Wicaksono et al. 2021), Artificial Neural Network (ANN) and Machine Learning (ML) (Adel et al. 2014; Sayers et al. 2014; Dazzi et al. 2021; Tamiru and Wagari 2022), Artificial Intelligence (AI) (Riazi et al. 2023), soil and water assessment tool (SWAT) and HEC-RAS (Ogras and Onen 2020; Taraky et al. 2021), GIS and Remote Sensing (RS) (Osei et al. 2021a, b), Frequency Ratio model, Fuzzy Logic (FL) Model, regional flood frequency analysis (RFFA) (Komi et al. 2016; Forson et al. 2023; Debnath et al. 2024), Watershed Modelling System (Soussa et al. 2012), and hydrological model (HEC-HMS) and Hydraulic model (HEC-RAS) (Tripathi et al. 2014; Shaikh et al. 2023; Al-Areeq et al. 2023; Baci et al. 2024). However, most of these methods are often data and situational dependent and are mostly expensive to implement in developing countries (Samela et al. 2020).

Evaluating the complete possibility of stopping floods from occurring perennially may be artificially or naturally unachievable and elusive. However, the delineation of flood risk terrains is critical for flood adaptation, mitigation management, and sustainable development in the Upper West Region due to the perennial flooding. Therefore, this study, employed techniques such as; Jenks natural breaks classification (JNBC), and an integrated Geographic Information System (GIS)-based Analytical Hierarchy Process Multi-Criteria Decision (AHP-MCD) to delineate flood risk terrains, buffer zoning of communities along rivers/streams, evaluate historical flood records and population trends. Spatial distribution of rainfalls from 1992 to 2022 using Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) (Funk et al. 2015) was also visualized in a Jupyter notebook. Historical flood events data were also analysed to ascertain the trends of flood disasters in the region. The findings would set the ground for the development of a paradigm shift in policy that supports sustainable flood risk adaptations and mitigation measures, and water resources management in the region. This study is aligned with several studies conducted around the globe (Danumah et al. 2016; Swain et al. 2020a, b; Gupta and Dixit 2022; Nsangou et al. 2022; Shekar and Mathew 2023a, b; Ziwei et al. 2023).

Materials and method

Study area

This study focuses on the Upper West Region (UWR), Ghana, which is within longitude 1 25’’ W and 2 45’’ W and latitudes 9 30’’ N and 11 N, and it shares a boundary with the Savanna Region to the south, with Upper East Region and North East Region to the east, with Burkina Faso to the north and the west with Cote d’Ivoire (GSS 2013; Baddianaah et al. 2022). The region has total land coverage of 18,779.71 square kilometres (Fig. 2) and it comprises about 12.7% of the total landmass of the country. The agroecological zone of the region is within the Guinea Savannah and Sudan Savannah belt (Diabene et al. 2014) with common trees such as; dawadawa, manula, shea, baobab and neem. These trees provide economic value to the people as well as firewood for domestic use. Generally, the most cultivated crops in the Upper West include; nuts, tubers, vegetables and cereals (Diabene et al. 2014). Currently, the region has a population of 904, 695, which constitutes 2.9% of the total population of Ghana (GSS 2021). The region experiences continental tropical air masses (dry and hot) (Brulard 2011). It experiences a single rainy season, lasting for a period of 4 to 5 months (April to September), with an annual average rainfall ranging from 840 to 1400 mm. The rainfall tends to be heavy between August and September. The northeast trade winds which bring the dry season known as the ‘harmattan’ start at the recedes of the rainy period, and are characterized by cold and hazy weather conditions from November to March, hot weather conditions, which only seize at the onset of the rainy season. Temperatures in the region are generally high; with monthly mean temperatures ranging from 21 to 32 °C, which rise to about 40 °C somewhere in March—April and fall to 20 °C in December (GSS 2013;Saana et al. 2016; Abu et al. 2024). The topography of the region is almost gently flat with Lawra and the western part of Wa mostly referred to as the Wa-Lawra plains, with a height of land range between 250 and 300 m. The western part of the region is drained by the Black Volta River Basin (BVRB), while the central and eastern parts of the region are drained by the Red and White River Basins (Kulpawn River and Sisili River), with their various tributaries from the sub-catchments are discharging southwards into the White Volta River Basin in the southern. Soil structure, texture, porosity, coarse nature, clay composition, and organic matter control the permeability and water holding capacity (Ball 2011; Kerr et al. 2016; Irvin IV et al. 2023). However, the soil types found in the region were Acrisols, Arenosols, Fluvisols, Leptosols, Lixisols, Planosols, Plinthosols, and Vertisols. These soils are largely generated over granite with some fewer areas developed over lower Birimian phyllite. The geology features Basal Sandstone, Birimian Sediments, Birimian Volcanics, Dahomeyan, Obosum and Oti Beds, and Upper Voltaian (Abdul-Ganiyu and Kpiebaya 2020).

Fig. 2
figure 2

Study area (UWR) map

Full size image

Methods

To spatially delineate flood risk terrains over a watershed, topographical and or hydrological surfaces to support policy and decision-making requires datasets and tools. Succinctly, the details of each data and method of analysis employed in this study have been explicitly described in the sub-sections of the study. Figure 3 summarises the insightful steps employed in conducting this study. However, the flood risk terrain maps were developed using two techniques; Jenks natural breaks classification (JNBC) and GIS-based AHP-MCD. Jenks' natural breaks classification (JNBC) (Jenks 1967; Guoyi et al. 2023) technique was employed to first optimize the delineation of the five (5) classes of the study area, to initially demarcate flood-susceptible areas within the study area, which provided an insight into the flood risk terrain to be expecting. JNBC techniques algorithm has been employed in conducting flood hazard studies and other studies (Anchang et al. 2016; Toosi et al. 2019; Amirruddin et al. 2020; Mishra and Sinha 2020; Deroliya et al. 2022). This algorithm ensures the optimal arrangement of classes’ values within a group and between classes. JNBC optimization determines to reduce the intra-class means deviation while ensuring that the inter-class means are maximized (Jenks 1967). Each of the classes indicates an area within the study area that would likely suffer perennial floods during heavy rainy days or periods, as a result of a high volume of surface runoff inundation due to the heavy rainfalls (Toosi et al. 2019). There have been enough shreds of evidence of heavy rainfalls as the cause of severe flood disasters in the Upper West Region (UWR) (IFRC 2023b), as a result, Jupyter notebook was employed to visualize Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) (Funk et al. 2015) for 30 years (1992 to 2022). The average monthly and August rainfall events were visualized. In addition, the study analysed population growth and historical flood events in the study area to understand their trends. Most especially, a correlation matrix of rainfalls and flood events records was also carried out.

Fig. 3
figure 3

Flowchart of the study process

Full size image

Data acquisition

Historical flood records give a preliminary knowledge of present and future floods in a catchment. Therefore, in this study, both primary and secondary datasets from various sources including local departments and organizations were collected and used. Secondary data on historical flood disaster records from 2016 to 2021 were sourced from the Upper West Regional National Disaster Management Organization (NADMO). Population growth rate is a key factor in exposure to flood vulnerability (Nguyen et al. 2021). As a result, population data between 2010 and 2021 was also obtained from the regional Ghana Statistical Service (GSS) department, Wa, Ghana. These datasets were sourced to explore the trend of floods and population growth within the UWR. However, there are no generally accepted methods and techniques for classifying flood potential areas, delineation fo flood risk terrains, flood zoning etc. (Jodhani et al. 2021), as there are several methods and factors that influence flood occurrences (Bello and Ogedegbe 2015). Al-Areeq et al. (2023) noted a better performance of the 30 m Shuttle Radar Topographic Mission (SRTM) DEM in flood extend assessment as compared with 12.5 m ALOSPALSAR, 10 m Sentinel 1, 30 m Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and 90 m MERIT products. Therefore, to develop a spatial flood risk terrain (FRT) map of the UWR, the SRTM DEM from USGS was used to generate the various thematic maps. Woeful, the paucity of data in the region to model the hydrological and hydraulic dynamics in the region has prevented innovative modelling of early warning systems and designs. After carefully reviewing the literature (Okyere et al. 2013; Arabameri et al. 2018; Al-Areeq et al. 2023; Shekar and Mathew 2023a, b; Shekar and Mathew 2023a, b; Ahmed et al. 2024; Danso et al. 2024; Jemai et al. 2024), six (6) spatial topographical, hydrological, and geological flood conditioning sub-thematic factors comprising topographic wetness index (TWI), Drainage Density (DD), Elevation, Slope, Lineament Density (LD), and Aspect, were selected based on the data availability, and the nature of the UWR topography to explore areas susceptible to flood disasters. These spatial thematic factors selected, significantly control landscape forms and characteristics, and hydrological processes, that drive surface runoff (Danumah et al. 2016; Ahmad 2018; Toosi et al. 2019; Mishra and Sinha 2020; Karymbalis et al. 2021; Leandro and Shucksmith 2021; Addis 2022; Amen et al. 2023). As a result, several, studies conducted around the globe selectively combined different thematic factors to either model, delineate, and assess; flood risk areas (Liuzzo et al. 2019; Mishra and Sinha 2020; Swain et al. 2020a, b; Roy et al. 2021; Gacu et al. 2022; Al-Taani et al. 2023), and groundwater recharge potential zones (Andualem and Demeke 2019; Gómez-Escalonilla et al. 2022; El Ayady et al. 2023; Ishola et al. 2023; Kumar et al. 2023; Moharir et al. 2023; Shekar and Mathew 2023a, b; Tariq et al. 2023; Feujio et al. 2024), soil erosion (Ahmad 2018) and landslides (Poddar and Roy 2024), indicating that flood modelling or assessment is not static to one set of datasets. Ground truthing points (GTPs) of some historical and frequent flooding points were collected from the field with a GPS device. The CHIRPS from Digital Earth Africa sandbox (https://sandbox.digitalearth.africa/) were also analysed to highlight the trend of the spatial distribution of the rainfall behaviour during the flood seasons over the UWR. However, buffer zoning was also employed to capture communities likely to be inundated when the streams and rivers flow their banks during peak rainy seasons.

Preparing and comparing flood-causing Sub-thematic factors

After establishing the sub-thematic factors preparation, the Saaty method (Saaty 1977) of Analytical Hierarchy Process (AHP)-Multi-Criteria Decision (MCD) (AHP-MCD) which has been applied in several flood studies (Danumah et al. 2016; Cabrera and Lee 2019; Swain et al. 2020a, b; Gacu et al. 2022; Gupta and Dixit 2022; Al-Taani et al. 2023; Shekar and Mathew 2023a, b) was used to rank these factors through assigned weighted values based on their strong relative relevance and contributing to runoff inundation, overland flows, and flood risk within the territory (Amen et al. 2023). AHP-MCD was done through a pair-wise comparison matrix of the sub-thematic factors, based on an expert judgement and knowledge of the research terrain, on how the factors are influencing flood within the area (Gacu et al. 2022). The consistency of each sub-thematical factor was assessed to ensure they were consistent with each other or not, using Eqs. 1 and 2.

$${\text{CI}}=\frac{{\uplambda }_{{\text{max}}-{\text{n}}}}{{\text{n}}-1}$$
(1)
$${\text{CR}}=\frac{{\text{CI}}}{{\text{RI}}}$$
(2)

CR, CI, RI (Table 1), and λmax represent the consistency ratio, consistency index, random index, and matrix principal eigenvalue respectively. However, n (Table 1) also presents the number of sub-thematic factors being compared (Abdrabo et al. 2023; Ahmed et al. 2024).

Table 1 Standard values of random index (RI) (Donegan and Dodd 1991; Saaty 2002, 2004)
Full size table

The consistency ratio (CR) for a 3 × 3 matrix and a 4 × 4 matrix are 0.05 and 0.09, respectively. CR of 10% (0.01) in general, is for large matrices. The computed CI from the process was found as 6.254, which yielded a computed CR value of 0.04 for the six factors (n = 6, RI = 1.25), which is less than 0.01. The CR value calculated was within the acceptable range and it, therefore, indicated that the pairwise matrices were consistent with each other (Gupta and Dixit 2022). Table 2 presents the environment’s natural hydrological, geological, and topographical sub-thematic conditioning factors of floods. The rating of each sub-thematic factor’s class weight was determined by the Saaty, (1977, 2004) scale of 1 to 9. The scale values range from equally importance through to extremely important (see Saaty 1977). Therefore, classes that were adjudged as equally important and extremely important were assigned values as 1 and 9 respectively. For Saaty’s scale from 1 to 9 of the intensity of importance see (Saaty 1977, 2002, 2004). The AHP-MCD weightage values for all the sub-thematic factors; Drainage density, elevation, slope, lineament density, TWI, and aspect, were 0.27, 0.17, 0.09. 0.21 and 0.05, respectively (Table 2). The sub-thematic factors were integrated and overlaid to generate the final AHP-MCD map.

Table 2 Weightage assignment of each factor
Full size table

Lineament density

Lineaments are geological features such as faults, fractures, joints, streams and valley networks (Das 2017; Das and Pardeshi 2017; Slimani et al. 2023; Tegegne et al. 2024). The sum of the length of all lineaments (Li) over an area is known as the lineament density (LD). LD regulates the flow of both surface runoff and groundwater by serving as conduits (Slimani et al. 2023; Tegegne et al. 2024). Therefore, the lineament networks were first extracted from the DEM in a Geomatic PCI environment. The lineament density was further generated using the Line Density based on Eq. 3 and reclassified into five classes ranging from 0–54.94 (Fig. 4a) in the GIS tool environment.

$${{\text{L}}}_{{\text{D}}}\left(\mathrm{Lineament\ Density}\right)={\Sigma }_{{~}^{{\text{Li}}}\!\left/ \!{~}_{A}\right.}$$
(3)

where Li is the sum of the length of lineaments in m or km and A is equal to the area (m2 or km2).

Fig. 4
figure 4

Spatial thematic factors influencing flood risk terrains in the UWR: a Lineament Density(km/km2), b Elevation (m), c TWI, d Drainage Density (km/km2), e Slope (%), and f Aspect (degree)

Full size image

Elevation

Elevation influences the speed and flow direction of surface runoff rate, which encourages terrain susceptibility to flood risk. Terrains with low elevations, low-lying and flat slopes have a high risk of perennial flood occurrences. The elevation is, therefore, vital in determining flood risk terrains (Association of State Floodplain Managers 2020). Surface runoff moves from high lands to low lands; flat and lower terrains are more than susceptible to flood hazards (Shekar and Mathew 2023a, b) due to the rapid inundation of water. The elevation of the study area was extracted from the national elevation shapefile and reclassified into five classes ranging from 150 to 400 m (Fig. 4b).

Topographic wetness index

The Topographic Wetness Index (TWI) is an important flood conditioning characteristic factor in flood risk terrain mapping. TWI regulates surface runoff accumulation tendency within a point in a watershed, and the soil wetness or dryness, saturation, moisture, and water depth (Costache 2019; Chowdhuri et al. 2020). A terrain with lower TWI values indicates that it is less prone to perennial flood risk (Shekar and Mathew 2023a, b), while high TWI values are considered high-risk flood vulnerability (Rahman et al. 2023) because saturated soil influences overland flows leading to inundation (Shekar and Mathew 2023a, b). DEM was used to generate the TWI map using Eq. 4;

$${\text{TWI}}=\frac{{\text{In}}({\text{As}})}{\mathrm{Tan\beta }}$$
(4)

As and β are the upstream area (area/length_m2/m) and slope gradient (degrees) of the watershed respectively (Das 2019) Paul et al. 2019; Rahman et al. 2023). The generated TWI map was reclassified into five classes ranging from -0.99 – 13.5 (Fig. 4c).

Drainage density

The drainage density (DD) forms the runoff connectivity patterns of the study area. It is the summation of the streams and rivers' length within a watershed/catchment divided by the total area of the basin (Rahman et al. 2023). Terrains with high DD values are highly susceptible to floods as well and terrains with lower DD are less prevalent (Getahun and Gebre 2015; Rahman et al. 2023; Shekar and Mathew 2023a, b; Kannapiran and Bhaskar 2024). To generate the DD, rivers/stream networks were first generated from the DEM and were used via the Line Density tool in the GIS tool under the Spatial Analyst tool. The generated DD was further reclassified into five (5) classes (Fig. 4d) ranging from low (less DD) to high (denser DD) (0 – 58.76 km/m2) using Eq. 5 below.

$$\mathrm{Drainage\ Density}\left({\text{DD}}\right)={\Sigma }_{{~}^{{\text{Lr}}}\!\left/ \!{~}_{A}\right.}$$
(5)

Lr is the total length of rivers and streams, while A is the total area of the watershed or catchment of the study area.

Slope

The slope is a major conditioning factor in spatial flood risk terrain identification (Karymbalis et al. 2021; Osei et al. 2021a, b; Ayenew and Kebede 2023). And how slope characteristics define the nature of an area’s extreme rains is critical. As a result, slope angles are important features in hydrological process modelling. It is a driving force of surface runoff and flow accumulation (Costache et al. 2021) that influences infiltration and soil erosion (Rahman et al. 2023) and high speedy runoff in high-land and waterlogging/flood inundation within low-lying terrains (Sachdeva and Kumar 2022). Generally, slope angles are important features in hydrological process modelling. Terrains within flat and low slope gradients are highly susceptible to the risk of flood while areas with higher sloppy terrain are less susceptible to flood (Liuzzo et al. 2019). To extract the slope of the study area, raster data of the digital elevation model (DEM) was used as an input. The region of interest (ROI) shape file was used as the boundary to digitize the DEM for the study. The spatial analyst tool in the GIS tool toolbox generated the slope from the DEM by calculating the maximum rate of changes in angle between each cell and its neighbour’s cell. The slope was calculated in per cent of the slope based on Eq. 6.

$$\mathrm{Degree\ of\ slope}=\uptheta ,\mathrm{ percent\ of\ slope}=\frac{\mathrm{Rise }({\text{Height}})}{\mathrm{Run }({\text{Length}})}\times 100$$
(6)

The produced slope map was reclassified into five classes ranging from 0–63.02% (Fig. 4e). Different colour codes were selected to designate areas liable to high and low flood occurrences. Red colour codes were assigned to areas with a high degree of susceptibility to flood occurrences. Green colour codes designated areas with moderate vulnerability of flooding and lower flood susceptibility were also assigned with yellow colour codes.

Aspect

Aspect shows the directions at which a slope is facing (Yu et al. 2023a, b). Aspect is measured in a clockwise direction in positive degrees ranging from 0 – 360, however, areas with zero or flat slopes are assigned negative ones (-1). The zero or flat aspects of terrains are the driving force of rapid surface runoff direction and accumulation, likely to result in floods (Sachdeva and Kumar 2022). DEM was used to generate the Aspect for the UWR, using the Surface tool in the GIS tool. The Aspect map was then reclassified into five classes ranging from -1 – 359.64 (Fig. 4f).

Population data

Equation 7 was applied in computing the population growth rate of the study;

$${\text{R}}=\frac{\left[{\text{Nt}}-{\text{NO}}\right]}{{\text{N}}0}*100$$
(7)

where: R = population growth rate, Nt = the final population size at time t, N0 = the initial population size at time 0.

Climate hazards group infrared precipitation with station data (CHIRPS)

The CHIRPS are satellite derived areal data average datasets, especially for rural, complex terrain and, semi-arid regions that are data-scared due to fewer rain gauge stations (Funk et al. 2015; Dinku et al. 2018). This was initiated through collaborative efforts with the United States Geological Survey Earth Resources Observation and Science (USGS-EROS) Centre to produce an up-to-date and complete reliable rainfall dataset for analyzing the trends and seasonal variations of the rainfall for monitoring and evaluation of hydrological occurrences, to support effective early warning systems development. The datasets have a record of over 30 + years spanning from 1981–01 up to date, with a spatial resolution of 0.050 with an approximation of 5.55 kms (López-Bermeo et al. 2022). CHIRPS was chosen due to its better performance as compared to other satellite products in Ghana (Atiah et al. 2020). Atiah et al. (2020) validated satellite products with 113 observation stations data over Ghana where findings showed a strong agreement with observed rainfall data and CHIRPS data as compared to the other products. A strong correlation of over 0.9 between CHIRPS and observational station data was also observed in the Upper Savannah agroecological zone as compared with the other agroecological zones in Ghana (Atiah et al. 2020). Ayehu et al. (2018) and Dinku et al. (2018), also reported better performance of CHIRPS products in East Africa as compared to other satellite products, this further strengthened the use of CHIRPS data in areas with observed data scarcity.

This necessitates the use of 30 years (1992–2022) of CHIPRS data to spatially visualise rainfall in the study area, at various spatial distribution scales; annual averages, five (5) year annual averages, monthly averages, and the month of August for each year.

Evaluation of FRT maps

The findings from this study will not be scientifically valid for policy and decision-making, if not evaluated using historical flood events (Cabrera and Lee 2019; Nsangou et al. 2022; Santosa et al. 2022; Upadhyay et al. 2022). Therefore, to ensure the reliability of the flood maps (both JNBC and AHP maps), ground truthing points (GTPs) of 85 historical flood points were collected within the study (Upper West Region (UWR)) after the 2021 direful flood events. The GTPs were overlaid on the maps and clipped to each class of the maps, to identify the number of points that were captured within each class. The captured points within each class were then calculated as per cent of the total points. Though there are several accuracy techniques such as; the area under receiver operation characteristic (ROC) curve known as AUC (Wang et al.2019; Arabameri et al. 2020; Pham et al. 2021; Rahman et al. 2021; Roy et al. 2021; Wubalem et al. 2021; Ghosh et al. 2022; Kumar et al. 2023; Shekar and Mathew 2023a, b; Tariq et al. 2023) for evaluating flood risk terrains, flooding areas, and potential zones delineation, however, application of this validation method was based on scientific studies and practices (Molinari et al. 2019; Amen et al. 2023). This is also imperative because, the methods presented have the potential to delineate FRTs and illustrate sensitive flood information to support flood mitigation strategies within the catchment, and therefore the findings have to be evaluated to ensure reliability and validity to boost the confidence of decision-makers (Molinari et al. 2019; Leandro and Shucksmith 2021; Addis 2022; Al-Taani et al. 2023).

Results and discussion

Flood risk terrains analysis

The JNBC algorithm optimized the study area into five (5) classes of flood risk terrains (FRTs) as illustrated in Figs. 5a and 6. The five classes were calculated as 4% (856.278 km2), 7% (1466.685 km2), 12% (2418.642 km2), 35% (7014.96 km2), and 42% (8351.496 km2) representing; very high, high, moderate, low, and very low FRTs, respectively. The very high and high flood risk terrain classes were observed along the southeastern and eastern parts of the study area, that is within the Sissala East Municipality (SEM) and Wa East (WE) district of the region, which agreed with the low elevation and low slope nature of the terrains. These findings were consistent with other studies that asserted the high chances of extreme flood occurrences in low-lying areas (Das 2019; Ullah and Zhang 2020; Al-Taani et al. 2023). The moderate and low risk terrains were also observed along the patterns of the very high and high flood risk terrains and the western border of the Black Volta River Basin (BVRB). More illustrations showed that the low and moderate risk terrains dominated within five districts; Jirapa, Lawra, Nadowli-Kaleo, Nandom, and Wa West, which share borders with the BVRB. Further observation showed that the very high and high flood risk terrains within the southeastern and eastern part of the study area are along the major rivers; Kulpawn and Sisili banks, this affirms that floods are ever occurring along river conveyances (Sofia and Nikolopoulos 2020). The areas along the river banks also have gentle to flat slope terrains, that is low elevations and low slopes which are attributed to high susceptibility to flood risks and similarly were observed by Ullah and Zhang (2020) along the Panjkora’s river banks. This also underpins the fact that the gentle to flat nature of the study area is generally influenced by overland flows during peak rainy seasons, leading to the high records of perennial floods. This further bears out the frequent flood occurrences within low-lying areas in the study area. Therefore, the knowledge of the terrain behaviour is paramount to the rice farmers and stakeholders' economic activities (Nguyen et al. 2021) of their proximity to these vulnerable areas. Assertively, other studies attest the closeness of communities and citizens to and within these vulnerable areas facilitates frequent flood hazards (Nazeer and Bork 2021; Mwalwimba et al. 2024).

Fig. 5
figure 5

Flood risk terrains: a Natural Breaks (Jenks) method and b AHP-MCD method

Full size image
Fig. 6
figure 6

Flood risk terrains; Natural Breaks (Jenks) method and AHP method

Full size image

The GIS-based AHP-MCD approach was also employed in weighing and comparing the sub-thematic flood conditioning factors, and integrating and overlaying the factors over each other generated the final FRTs map (Das 2019; Liuzzo et al. 2019; Toosi et al. 2019; Wang et al. 2019; Ullah and Zhang 2020; Addis 2022; Al-Taani et al. 2023; Shekar and Mathew 2023a, b; Suwanno et al. 2023). The final output map was classified into five (5) categories of flood risk terrains, ranging from very high to very low. This was done based on the susceptibility of each class to flood occurrences. Figure 5b illustrates the spatial distribution of the FRTs over the study area, while Fig. 6 presents the FRTs in percentages. The five categorised classes were namely; very high, high, moderate, low, and very low with areas coverage of 0.707 km2 (0.004%), 3830.02 km2 (21%), 12,807.31 km2 (69%), 1827.011 km2 (10%), and 11.535 km2 (0.062%) respectively. Figure 5b shows that the majority of very high and high flood risk terrains are spatially distributed along the southeastern, eastern and north, while some are sparsely scattered within the whole region. This could result in scattering of extreme flood disasters over the region (UWR) since they are not confined to a specific location. Further observations clearly showed that the very high and high FRTs are heavily found along the stream/river networks, confirming that low-lying areas are susceptible to flood disasters (Pham et al. 2021; Amen et al. 2023; Shekar and Mathew 2023a, b; Danso et al. 2024). These findings were also consistent with a study by Rana et al. (2023). Additionally, the very high and high FRTs also coincided with the flat to low slope, low elevation (150–300 m), high drainage density (47.01–58.76 km2/km), and high TWI (4.52–13.5), which influence the high accumulation of runoff during rains. This also confirms that the topography plays a major role in conditioning flood susceptibility in low-lying terrains. The very low and low FRTs are found in areas with higher slopes, higher elevations, and low drainage density, especially, within the highland areas in the region. The moderate FRTs were discovered as the dominant class, spatially distributed all over the study area, and accounted for 69 per cent of the total study area. Furthermore, critical observations also noted that the majority of the very high and high FRTs are sandwiched within the major river catchments (Kulpawn River and its tributaries) and alongside the Sisili River in the region. This indicates that the communities, districts, and farms are in perennial flood danger during the rainy periods. This has also been verified by the historical perennial flood events in some communities and districts in the region. That is, those communities, districts, and municipalities which are heavily drained by the Kulpawn River and Sisili River and their minor tributaries during rainy seasons; most especially Wa East, Sissala East, Sissala West, Daffiama Bussie Issa and the eastern part of Jirapa. This revealed that low-lying areas, most especially those along the rivers/streams are risky zones, due to overflows of their banks. However, due to the nature of agricultural activities in the region, farmers mostly plant crops such as rice within low-lying; waterlogs, waterways, valleys, rivers/streams banks, and near dugouts’ areas, and this increases the perennial exposure of the farms to flooding, inundate and washing away of crops. The results from this study, significantly affirm that the very high and high FRTs in the study area are not homogeneous, and are distributed all over the region. This is also sign that the moderate FRTs are likely to experience unprecedented extreme floods, due to spillover and overflows from the very high and high flood risky areas (Amen et al.2023). These observations, further confirmed Ghana’s flood prone areas are always along rivers/streams, near rivers, waterways, poor drainage networks, encroachment of waterways, and river basin sites, which are often serving as high runoff retention areas (Okyere et al. 2013; Yin et al. 2021, 2022; Gaisie and Cobbinah 2023; IFRC 2023a).

The knowledge and information of flood vulnerability areas, risky terrains, flood drivers, and their impacts are key factors for strategic flood disaster management, policy, and decision-making (Nazeer and Bork 2019; Li et al. 2022). Delineation of flood risk terrains and flood studies are rare in this part of the country, despite the perennial flood disasters. However, apart from drought, flooding has been the most disastrous problem, affecting thousands of lives and properties. Hence, the study aims to scientifical delineate these flood risk terrains in the region in support of farmers, stakeholders, and NGOs, and to boost sustainable developments in the region.

Validation of flood risk terrain maps

The ground truthing points (GTPs) of historical flood events were overlayed and clipped to each of the FRTs maps. The JNBC map classes recorded; 31.76% (very low), 16.47%, (low), 15.29% (moderate), 24.71% (high), and 11.76% (very high) flood risk terrains, respectively. Quantitatively, 31.76% of the flood historical points were recorded in the very low flood risk areas, while the very high-risk regions recorded the lowest points with 11.76%. The generation of the JNBC map did not take into consideration other flood influencing factors which limited its ability to delineate a truthful representation of flood occurring areas within the study area. During the field data collection, it was also observed that several communities, and suburbs within the Wa Municipality are encroaching on, and sandwiching stream/river networks within the city, which probably led to the 31.76% of GTPs within the very low class. During the evaluation of the AHP-MCD map reliability, it was observed that none of the GTPs was found within the very low and very high flood risk terrains classes. However, the low, moderate, and high classes recorded; 20%, 71.76%, and 8.24% of the GTPs, respectively. It is also arguable that the very low (11.54 km2) and very high (0.71 km2) risk terrains in terms of area were insignificant which could probably be the reason why no GTPs were found in their classes. The higher number of GTPs found within a class shows the susceptibility of that class to flood events (Toosi et al. 2019). Therefore, 71.76% of the historical flood events recorded within the moderate class, revealed that the study area (UWR) is largely dominated by moderate flood risk terrains. The validation also further showed that flood disasters are not homogeneous within the study area. This confirmed the reliability of the findings to assist and support stakeholders in flood planning, decision-making, and policy (Amen et al. 2023).

Buffer zoning of stream network

Buffer zoning (bz) plays a key role in regulating flood risks along streams and rivers (Osei et al. 2021a, b; Kurugama et al. 2023). During the heavy rainy seasons, most streams and rivers overflow their banks to some extent which provokes inundation and flooding communities and farms within their catchment. It is well noted that in most communities in the UWR, farms sandwich wetlands, along streams and valleys, and near dugouts (commonly called dams in the region). At the end of the 2021 flood disasters, NADMO reported 121 communities in some municipalities and districts (Sissala East Municipality (SEM), Sissala West Municipality (SWM), Wa East (WE), Wa West (WW), Nadowli-Kaleo, Lawra, Lambussie/Karni, and Wa Municipality (WM)) in the region as flood risk communities. Therefore, to further identify the number of communities likely to experience fluvial floods in the UWR, a buffer zoning of the stream/river networks in the study area was, therefore, important to further observe the pregnant communities that would be inundated whenever these streams/rivers overflow their banks into their flood plains and their nearby subbasins. A buffer of 0.5 and 1 km was selected to observe communities likely to be inundated within the flood risk zones when rivers/streams overflow their banks. With about 1952 community points in the UWR, it was observed that 395 (Fig. 7a) and 927 (Fig. 7b) communities were found as sandwiched between 0. 5 km and 1 km respectively. It further revealed that these communities within the buffered ranges are sited within the moderate, high, and very high flood risky terrains (both JNBC and AHP-MCD maps). It is paramount to understand that these communities are more susceptible to high flood hazards than those outside the buffer zones. Notwithstanding the several flood conditioning factors; settlement encroachment on wetlands, heavy rains, waterways, low-lying areas, runoff retention areas/runoff converging points, rapid population growth etc. (Yang et al. 2022) and climate change impacts (IPCC 2023b). These findings were also consistent with Toosi et al. (2019) and Osei et al. (2021a, b) studies, which revealed that communities within the buffer zones are more prone to flood disasters than those outside the border. This further highlights a sign of future flood disaster severities in the informal urban settings and communities in many developing countries (Juma et al. 2023). As it has already been reported, flood disasters in developing countries are expected to increase due to rapid infrastructure encroachment on waterways and climate change that owes its impact to severe rainfalls (Anwana and Owojori 2023; IPCC 2023a, b). Therefore, buffer zoning is a key component towards delineating retention zones for sustainable development and planning of cities and communities (Cabrera and Lee 2019; IPCC 2023a, b). However, there is an arguable assumption that cities' and communities' exposure to flooding disasters is constant as the population keeps increasing and settling around and near risky and vulnerable areas (Aerts et al. 2018; Tesselaar et al. 2023). These findings highlight the importance of integrating buffer zoning into the policy in the communities, towns and municipal planning.

Fig. 7
figure 7

Buffer zoning (bz) communities: a communities within 0.5 km bz and b communities within 1 km bz

Full size image

Historical flood disasters in Upper West Region (UWR)

Figure 8 presents historical flood event records over six (6) consecutive years (2016–2021) from the National Disaster Management Organization (NADMO) in the UWR. The report indicated significant flood disaster experiences, which affected over 1,344, 6,806 and 6,190 people in 2016, 2018, and 2021, respectively. Floods in August 2021, were recorded as the worst disaster, that has ever happened in the region, with an estimated cost of 1,183,660.00 Ghana Cedis. It was also further observed that the affected population rose with the economic losses, especially in the years; 2016, 2018, and 2021, which were hard hit years. The statistics (Fig. 8) revealed that the population affected in 2018 was higher than that in 2021, however, the economic losses cost in 2021 were also higher than in 2018, due to the massive destruction of many infrastructure; such as roads and culverts, crops, and houses/homes. The poorer settlers in vulnerable areas are largely affected due to their inability to afford safer places and arable lands for agricultural activities within the rural communities (Singer 2018).

Fig. 8
figure 8

Cost (Ghs), affected population, and farmlands (Acres)

Full size image

This therefore showed, that population growth rate is also a key influencer of flood occurrences as human activities; agriculture, infrastructure, businesses, roads and settlements encroach on natural waterways, low-lying areas, wetlands, valleys, runoff retention/detection areas and increased impervious areas, and reduce the natural drainage systems (Singer 2018; Kaiser et al. 2021; Pham et al. 2021; Gaisie and Cobbinah 2023). The number of crop acres destroyed in 2021 (Fig. 8) is the confirmation of agricultural activities sandwiching wetlands, waterways, low elevations, valleys, and rivers/streams. Most especially, rice farming is the most predominant in those areas. The population of the UWR has steadily increased from 702,110 in 2010 to 901,502 in 2021, indicating a growth rate of 28.40% per year (Fig. 9b). This growth is likely to continue in the future, as the region is experiencing a significant influx of people due to factors such as rural–urban migration, high birth rates, establishment of new educational faculties and improved healthcare, as well as economic activities. Therefore, analysing the population trends in the region is an indispensable aspect of future flood hazards/risks assessment and sustainable management. Figure 9a, b, and c), shows that the population in the region and the districts has been increasing over the 10 years (2010–2021). Nevertheless, the population data was estimated between 2012 and 2020 except for 2021 data, where population data was recorded through the National Population and Housing Census (PHC) project (GSS 2021).

Fig. 9
figure 9

UWR Population 2010–2021: a population map, b regional trend, and c Districts' population growth trend

Full size image

Statistically, there has been about a 28.40 per cent population increase between 2010 and 2021.

$$Rate\ of\ Change=\frac{901502-702110}{702110}\times 100=28.39896882\mathrm{\%}\sim 28.40\mathrm{ \%}.$$

Observations from Fig. 9c, also revealed that districts; Jirapa, Lawra, Lambussie Karni, and Nandom saw a drop in population growth, along with Daffiama Bussie Issa (DBI), which also experienced a little drop in its populations, according to the figures recorded in 2021. However, other districts maintain a constant increase. Significantly, Wa Municipality, which serves as the regional capital, maintains a consistent increase in population. Population growth is a major contributing factor to flood occurrences of hydrological events (Rahman et al. 2021), particularly in regions and terrains with poor infrastructure and inadequate flood control measures, as well as in this era of climate change (Saksena et al. 2020). Population increases translate into urban sprawl and an increase in human activities encroachment on waterways, low-lying areas, and wetlands (Rahman et al. 2021). In the case of the Upper West Region of Ghana, the following factors may have increased the region's vulnerability to floods, though climate change could play a mastermind too. Figure 8, highlighted a steady increase in population and flood events. These observations further showed that the increase in population decreases the surface runoff retention areas and the natural drainage capacity due to the conversion of natural drainage areas for development projects and encroachment on wetlands as a result of rapid urbanization, these observations were consistent with a study conducted by Seemuangngam and Lin (2024) on the influences of urbanization on urban floods.

Spatial rainfall distributions and floods

Monthly average rainfalls (1992–2022)

Figure 10 presents the spatial distributions of monthly average rainfalls over the 30 years (1992–2022), while Fig. 11 shows the average rainfall for the months of August in each year. The rainfall plot scale ranges approximately around 350 mm/month to below 50 mm/month (Fig. 10). The results (Fig. 10) clearly, revealed that the study area experienced seven (7) months of no or little rainfall. This indicates that October to April serves as the long dry period in the study area as reported by Abu et al. (2024). Arguably, due to the persistent high temperatures that can hit from 40 to 41 degrees Celsius coupled with heatwaves during March and April, rainfalls in April amount to little to nothing before the commences of the rainy season in May (Dery et al. 2023; Abu et al. 2024). The spatial distribution clearly showed an indication that the dry season intensely lasts between November and March as the rains completely recede in October. The observations further illustrated that the rainy seasons in the UWR are between May and September. Additionally, the results highlighted that the rainfall peaks from June to September (JJAS) with exhibitions of non-homogeneity (Fig. 10), these observations were consistent with the study area's climatic conditions (Dery et al. 2023; Abu et al. 2024). However, the month of August exhibited a strong heterogeneity of rain distribution across the northern part of the region (UWR), largely with the northeastern part of the Sissala West Municipality (SWM) and the northwestern of the Sissala East Municipality (SEM), recording high nucleus of rains magnitude. Consistently, these two municipalities have recorded perennial flood disasters during the rainy season in August. These observations are also in line with the 2007 floods, which were reported to have massively affected populations in both Wa East, Wa West, Sissala East Municipality, and Sissala West Municipality (SWM) (United Nations 2007; Centre Emergency Response Fund 2008). To extensively rule out the fact that the month of August experiences the most perennial floods in the region. Figure 11 also presents the average rainfall results for each August within the 30 years (1992–2022). The total number of months (Augusts) from 1992–2022 analysed was 30 (Fig. 11). The overall mean of the 30 months was computed to be 239.35 mm. The results highlighted some Augusts with higher rainfall depths than others, which is either above or below the overall mean (239.35 mm). The graph showed that, only six (6) Augusts with average rains below 200 mm per month. Illustratively, Augusts in some years; 1996, 2007, 2010, and 2021 also recorded average rainfall peaks over 300 mm. It can be seen from the graph that, the August of 2007 recorded an average rainfall value above 300 mm, where the massive flood disaster occurred in the same August in 2007 (Centre Emergency Response Fund 2008). Significantly, the average rainfall peak amount in August 2021 is approximately 350 mm the highest observed value within the analysis period, this, therefore, confirms why the floods in 2021 were worst of all, after the 2007 flood disaster. All these rainfall findings are consistent with the historical flood disaster records (from 2016–2021) reported in the study area by the NADMO. Additionally, Asumadu-Sarkodie et al. (2015), United Nations (2007), and Centre Emergency Response Fund (2008) reported that the terrible 2007 flood occurred as a result of three (3) weeks of extremely intensive rainfall between August and September 2007. This led to the loss of lives, livestock, properties (homes and schools), foodstuff, water supply systems and roads were also destroyed. The government in 2007 pronounced the study area and other regions (Northern and Upper East regions) as a state of emergency due to the terrible nature of the situations of flood destruction that had occurred. All observations point out August as the most flooding disaster risky month in the study area and generally, the northern parts of the country (IFRC 2023b). This also strengthens the fact that climate change influences droughts and floods through the hydrological cycles (UNESCO 2023a). Further indications from the neighbouring regions (Upper East and Northern) of the country also reported August rains as disastrous. Especially, heavy rainfalls in August of 2018 and 2019 coupled with the perennial spillages of Bagre Dam also caused an unforgettable flood woes in the Upper East region (UER) where Sissala East Municipality and Wa East district share borders (IFRC 2018, 2019). These analyses are consistent with the high nucleus of rainfall magnitude in the Sissala East Municipality and the Sissala West Municipality (SWM) (Fig. 10) and the Augusts average rainfall (Fig. 11). Subsequently, rainfalls forecast on May 5, 2023, by the Ghana Meteorological Agency (GMet) reported, heavy rainfall peaks above normal depths in the UER, while observations in the Upper West, Savannah, North and North East regions ranged from normal to above normal depths till early weeks of September (IFRC 2023b). The current findings affirm the manifestation of flooding during period of heavy rainfalls in the study (Centre Emergency Response Fund 2008; Owusu et al. 2017; IFRC 2018, 2019, 2023b; IPCC 2022). It is also noted at the recede of the 2022 rainy season (GMet 2022), FAO (2022) reported over 80 k hectares of crops had been washed off by floods under influence of heavy rainfalls between August and October 2022 in the northern parts of the country with the study area inclusive.

Fig. 10
figure 10

Spatial distribution of climatology average rainfall (over 30 years) for each month

Full size image
Fig. 11
figure 11

Average monthly rainfall for each August: 1992–2022

Full size image

The findings further reflect the shreds of evidence of heavy and intensive rainfall contributions to flood disasters, which has been widely reported around the globe (Das 2019; Anwana and Owojori 2023; Ncube et al. 2023; Pandey et al. 2023; Plataridis and Mallios 2023). West Africa sub-region alone has over the past decades recorded several flood occurrences during the rainy seasons (Ntajal et al. 2017; Miller et al. 2022). Flooding has been projected to worsen as climate changes through its propelling conditions of extreme rainfalls (IPCC 2022). To buttress the synergy of rainfalls and flooding, the IPCC (2023a, b) further reported an increasing correlation between rains and floods. The findings demonstrated a trend of the relation of how rains in the study area keep nurturing floods in the vulnerable villages, communities, districts and municipalities in the region.

Spatial patterns of average rainfall distributions

Figure 12, also presents the variations of the spatial distribution of annual average rainfall patterns over 30 years (1992–2022); ranging on a scale from 60 mm/yr to 110 mm/yr. From Fig. 12, the presentation highlights the low and high magnitude rainy areas distributed over the study area. The least rainy years were observed to be 1992, 2005, 2013, and 2017 in the study area (Fig. 12). The northwestern and middle parts of the region experienced the least amount of rain in 1992 and 2005 respectively. The southern part of the region also experienced the least distribution of rainfall in 2013 and 2017. The findings further revealed that the northeastern parts; both the Sissala East Municipality (SEM) and Sissala West Municipality (SWM) experienced a high depth of rainfall in 1994. Significant rainfalls were observed to have partly distributed over the region in 2003, 2006, 2010, and 2012. Additionally, a high magnitude of rains was also noted over the region in 1999, with much being felt in both the Sissala East Municipality (SEM) and Sissala West Municipality (SWM). The results revealed strong spatiotemporal observations throughout the region in 2018 and 2019, though more observations were seen along the SEM and SWM, and the Wa East. Significantly, the high rainy trends were observed in 1999, 2018, 2019, and 2021 within the scale of 90 – 100 mm/yr. This further showed that the seasonal rain trends during the 30 years have been not homogeneously distributed over the study area (UWR). These presentations highlighted the rainfall dynamics information over the region, providing a spatiotemporal pattern of the rainfall behaviour in the municipalities and districts. Furthermore, these presentations aligned with the understanding that rainfall trends are influenced by climate change dynamics, that affect the presence of water, and the strengthening of floods and droughts (Wani et al. 2017). The heavy rainy seasons are the key factors inciting floods, and due to the inability of rivers, streams, channels, and low-lying areas to discharge a high volume of surface runoff flows, results in inundations of their catchments and or riparian communities (Nahin et al. 2023). The findings were observed to be consistent with the historical flood events reported by the NADMO in the study area. Additionally, the high spatial distribution of rainfalls, largely in the 2016, 2018, 2020, and 2021 affirms flood disasters reported.

Fig. 12
figure 12

Spatial distribution of annual average rainfall (mm/yr)

Full size image

Figure 13 represents 5 years categorization of spatial annual average rainfall distribution (1993–2022). From the analysis, both Figs. 12 and 13 visualizations show distinctive similarities of heightened rainfalls along the eastern SWM and northwestern SEM. The NADMO has so far identified several communities as flood prone communities, with 9 and 15 communities in the SWM and SEM, respectively. Notably, from the 1990s to 2022, parts of the study area experienced a series of perennial flood events that were largely attributed to the severity and the intensive nature of rainfalls during the rainy seasons. Perennial flood manifestation between 2016 and 2022, in the study area has been under an upward rise in flooding events, particularly linked to the spatial distribution of the rains in the region. As illustrated in Fig. 13, the highlights revealed that a less substantial amount of rain (75–85 mm/5yrs) was observed within the middle and southern parts of the region in 2017–2022. On the contrary, the historical flood events documented show destructions of flood episodes all over the middle belt and the southern parts. This is a confirmation of fewer rains resulting in flood perennial floods mostly during seasons. As global climatic patterns have disrupted the ecological system of the study area, extreme rainfalls in the rainy seasons become intensive, manifesting severe natural disasters; floods and longer droughts in the dry seasons. This has underscored the need for sustainable mitigation measures to address flooding and drought impacts in vulnerable communities within intensive rainy zones. Unforgettable, in August 2021, the study area experienced unusual rainfall between the 12th and 13th of August, which resulted in devasting flooding, that swept away farms and homes, and disrupted roads, with almost all the districts and municipalities being blocked from the regional capital. The national road linking Ghana and Burkina Faso through the region was also massively cut off, as several culverts were swept away. This signified the severity and intensity of the rainfall within two days that led to the terrible flood disaster. The spatial highlights of the visual rainfall distributions (Fig. 13), provide critical insightful information on the amount and patterns of rainfalls experienced over the study area during the period of the analysis.

Fig. 13
figure 13

Spatial patterns of rainfall distribution (1993–2022): Five (5) year annual average(mm/5yrs)

Full size image

As delineated in Fig. 13, the communities that felt within the heavy rainy areas (80 to 95 > mm/5yrs) were exposed to the consequences of multiple flood episodes. This therefore affirms that when the rivers and streams become incapacitated of channelling out a high volume of rainwater results in floods in the communities (Nahin et al. 2023). This further strengthens the fact that rainfall is one of the key causes of flood disasters in the Upper West region. Though the intensity of rainfalls alone cannot qualify it as the main cause of flood disasters (Garba and Abdourahamane 2023), however, several flood disasters over the years within the region and country, and around the globe have been linked to heavy rainfalls (Paeth et al. 2011; Okyere et al. 2013; Asumadu-Sarkodie et al. 2015; Kwang and Osei Jnr 2017; Cabrera and Lee 2019; Sacré Regis et al. 2020; Osei et al. 2021a, b; Alifu et al. 2022; Danso et al. 2024; Nagamani et al. 2024; Rączkowska et al. 2024). It has also been noted that heavy rains as an extreme weather event often result in natural hazards with floods alike (World Economic Forum 2024).

Correlation matrix of historical rainfalls and flood disasters

The correlation matrix (Fig. 14) was also used to measure the magnitude of August’s heavy rainfalls and their association with historical flood events. Therefore, the average Augusts rainfalls (2016–2022), and the six years of historical flood records available data (affected population, farmlands, and cost of disasters) were used as the elements of proportionality to compute the correlation matrix to confirm the Augusts rainfalls relationship with floods, that have a strong association on elements within the study area. The correlation matrix is based on a scale range of -1 ≤ r ≤  + 1, where r could represent -1, 0, and + 1, as a negative proportionality of each two elements, no correlation/non-association, and proportionality of each two elements respectively (Castleman and Wu 2023). Schober and Schwarte, (2018) indicated that a strong correlation is from 0.7 to 0.89, a very strong correlation ranges from 0.99 to 1.0, while a moredate correlation also ranges from 0.40 to 0.69. The results (Fig. 14) showed a strong correlation of the rainfalls with the affected population, farmlands (acres), and cost (Ghc) with r values of 0.82, 0.78 and 0.87, respectively. The economic losses (Cost (Ghc)) and the affected population exhibited a very strong correlation of 0.99, while farmlands (acres) and the affected population recorded a moderate correlation of 0.6. However, the cost (Ghc) and farmlands (acres) also reported a moderate correlation with an r ≤ 0.67. These analyses further strengthen the relationship between rainfalls and flood events' effects on citizens and the general ecosystem.

Fig. 14
figure 14

Paired correlation of rainfall and flood effects

Full size image

Conclusion and recommendations

The main goals of this study are to delineate flood risk terrains, buffer zoning of communities along rivers/streams, evaluate historical flood records and population trends, and visualization of rainfall behaviour, which are proactive measures for the management and mitigation of perennial floods in the Upper West Region. The study primarily proceeded with Jenks natural breaks classification (JNBC) (Jenks 1967) technique and GIS-based Analytical Hierarchy Process (AHP) multi-criteria decision (AHP-MCD). JNBC technique employed only SRTM DEM to first optimize the areas likely to be susceptible to flood. The GIS-based AHP multi-criteria decision algorithm also incorporated six geospatial flood conditioning factors comprising Lineament Density, Elevation, Topographic Wetness Index, Drainage Density, Slope, and Aspect. This was done based on literature and the availability of datasets. Buffer zones of 0.5 km and 1 km were also created around the streams/river networks to determine the number of communities that will be inundated during fluvial floods. The outcomes of JNBC algorithm reclassification indicate that the UWR can be categorized into five classes; very high risk to very low risk terrains covering 4% (856.278 km2), 7% (1466.685 km2), 12% (2418.642 km2), 35% (7014.96 km2), and 42% (8351.496 km2), respectively. The performance of the JNBC flood risk terrains map generated was evaluated with ground truthing points (GTPs) of 85 historical flood points, this was done by overlaying the GTPs on the map, while computing the number of GTPs on each class of the map as a percentage. The very low flood risk terrains recorded 31.76% of the GTPs while the very high flood risk terrains recorded the lowest GTPs with 11.76%.

The AHP-MCD integration of the six sub-thematic flood conditioning factors also characterized the total study area into five classes; very high, high, medium, low, and very low with coverages as 0.707 km2 (0.004%), 3830.02 km2 (21%), 12,807.31 km2 (69%), 1827.011 km2 (10%), and 11.535 km2 (0.062%), respectively. During the validation, the moderate flood risk terrains class attained 71.76% of the GTPs as compared with the other classes, thus signifying that the classes generated were accurate for decision-making in the study area, and largely indicated the study falls within moderate flood risk terrains.

The spatial outlines of the very high flood risk terrains and high flood risk terrains classes of both the JNBC and AHP-MCD maps are situated along the southeastern and eastern fringes of the study area, that is, Sissala East Municipality (SEM) and Wa East district (WED), and primarily sited along the major river networks, agreeing that low-lying areas are more susceptible to flood disasters in the study area.

The agreement of the GTPs (historical flood points) and the flood risk terrains generated by the AHP-MCD method, improved the acceptance of the results for flood policy formulation and communication, sustainable flood mitigation strategies and management, and as well as the public perception of flood disasters (Houston et al. 2019).

From the selected buffer ranges; 0.5 km and 1 km, communities sandwiched between the buffered ranges were classified as vulnerable to flooding as compared to those outside the ranges, during an overflow of the rivers and streams.

The spatial visualization of the rainfall patterns distribution highlights the important role rainfalls play in flood assessment. The CHIRPS data analysed revealed the distribution of the heightened rainfall over flood risk terrain communities. The historical flooding records over the years, months and communities were consistent with the spatial rainfall distributions especially the month of August.

This study contributes valuable insights into spatial flood and extreme rainfall patterns distribution in the study area.

It emphasizes that the town and country planning, municipalities and district assemblies ensure that buffer zoning rules are strictly enshrined in community, town, and city planning, and the allocation of areas for broad farming.

Farmers need to plant short maturity period seeds and crops that can be harvested before floods commence or that stay in flood waters for a period. Most especially rice and maize cropping in the region, though rice is a hydrophobic crop, however, the flood water resistant variety should be encouraged in the region, mostly for those farming along the risky areas.

The findings underscore, further studies using machine learning algorithms to analyse satellite imagery data for inundation through the integration of advanced scientific hydroclimatic modelling and simulations and stakeholder engagement around the communities to prepare resilient proactive measures to tackle future flood events.