A multi-method approach to analyse changes in gully characteristics between 2009 and 2018 in southeast Nigeria

Gully erosion is the dominant environmental problem in southeast Nigeria and has led to loss of human and material resources. In this study, we evaluated changes in gully characteristics in southeast Nigeria and their potential drivers between 2009 and 2018 using a multi-method approach: analysis of high-resolution satellite imagery (2 – 5 m) and focus group discussions. Gully numbers increased from 26 to 39 and mean gully length increased from 0.39 to 0.43 km. We found that land adjacent to rivers had the highest concentration of gullies which is associated with an increase in slope angle from 10 to 58% up to 500 m from rivers. Regarding potential gully-drivers, land-use changes were observed. Non-vegetated lands increased from 58.6 to 144.7 km 2 between 2009 and 2018, while reductions in fallow lands from 281.2 to 57.8 km 2 were observed. Results from focus group meetings indicate there was no gullying in the area before the Nigerian civil war (1967 – 1970). War-time activities such as digging trenches and increased population density were said to have led to the formation of the oldest gullies in the area. Although war-time activities have ceased, meeting attendees believed that present land-use changes have increased the volume of surface runoff and thus enhancing gully erosion. Incorporating local knowledge in this study has therefore provided a valuable understanding on the key drivers of gullying that pre-dates the availability of freely available high-resolution satellite data.


| INTRODUCTION
Gully erosion is recognised globally as an important environmental problem with significant effects on the availability of land for cultivation, crop productivity and land degradation  Morgan & Rickson, 2003;Rickson et al., 2015;Vanmaercke et al., 2021;Zhang et al., 2002). Gully erosion is the dominant environmental problem in the southeast region of Nigeria and has led to the isolation of villages, severance of communication lines such as roads, as well as loss of homes, schools, human and material resources (Egboka et al., 1990). Several studies have been conducted on gully erosion and many factors have been found to enhance gullying, for example intensive rainfall (Afegbua et al., 2016;Obi & Salako, 1995;Ofomata, 1987;Vanmaercke et al., 2016), prolonged drought (Cerdà, 1997;Fleitmann et al., 2007), geology and lithology (Betts et al., 2003;Castillo & G omez, 2016;Ghimire et al., 2006;Parkner et al., 2007;Sonneveld et al., 2005), soil characteristics (Egboka & Nwankwor, 1985;Okagbue & Ezechi, 1988), nearness to rivers and roads (Frankl et al., 2019;Zabihi et al., 2018), land-use changes (Frankl et al., 2019;Panagos et al., 2020;Yibeltal, Tsunekawa, Haregeweyn, Adgo, Meshesha, Aklog, et al., 2019) and topography (G omez-Gutiérrez et al., 2015;. These results suggest the multiplicity of gully drivers and the complexity of gullying. Different research methods have been used to study gully erosion, for example, Okagbue & Ezechi (1988) and Iheme et al. (2016) adopted geotechnical approaches to understand drivers of gully erosion in southeast Nigeria. Their study found that soils in the area prone to gullying have higher sand content, high permeability, and low cohesion. High permeability may reduce surface runoff but facilitate infiltration and high internal flow velocities and seepage pressures, whereas low cohesion enhances dispersal by surface runoff (Okagbue & Ezechi, 1988). Ezezika & Adetona (2011) adopted a community-based method including interviews with stakeholders and informal discussions with experts on gully erosion to understand gully-drivers. Their study found that gully erosion is driven by human activities, especially, poor land-management practices. has a critical control on gullying as larger catchments produce higher volumes of runoff (Frankl et al., 2012;Dong et al., 2013;Vanmaercke et al., 2016;. However, considering the influence of changes in land use on gully evolution, it is also important to understand how these land-use changes in individual gully catchments influence gullying. One can say that no single research method provides all the answers to gullying, rather, research methods are chosen based on research questions of interest and data availability. As already stated, gullying is a complex process. It is challenging to study gullying over medium to long timescales (10-50 years), especially, in data-scarce regions, because of the lack of observations. To overcome this challenge, some studies on gully erosion have adopted multi-method approaches such as incorporation of community-based knowledge, analysis of remotely sensed data and quantitative analysis Nyssen et al., 2006;Tebebu et al., 2010). The significance of incorporating community knowledge in gully studies is fourfold: First, to eradicate any form of mistrust for science on the part of local population.
Second, incorporating local knowledge provides for comparison of scientific understanding with prevailing local expertise, thus, enriching scientific findings. This second significance forms part of the public debate model proposed by Callon (1999). Third, to avoid the adoption of nonnative 'top-down' approaches which are not effective in many instances in solving local environmental problems (van Aalst et al., 2008). Finally, it provides qualitative information on gullying which extends historical record and fills the existing gap due to unavailability of data (e.g., remotely sensed imagery of study area). Adopting a multi-method research approach to study gully erosion improves scientific knowledge of gully-driving processes by comparing scientific knowledge with local expertise. Changes in gully characteristics (form) are likely driven by the interactions among dominant drivers of gully erosion (process), hence, to understand gully dynamics, knowledge of interactions of the dominant drivers is important. To this end, the aim of this paper is to use a multimethod approach to ascertain the dominant drivers of the changes in gully characteristics in the study area between 2009 and 2018.

| Study area
Five Local Government Areas (LGAs): Ideato North, Ideato South, Isu, Njaba and Orlu covering an area of 534 km 2 in Imo State, southeast Nigeria were studied ( Figure 1). The climate of southeast Nigeria is tropical with rainfall decreasing from the coast inland. Rainfall generally is intense and ranges from over 2500 mm annually in the southernmost region to about 1500 mm in the northern borders (Igwe, 2012). Along the coast, mean annual minimum and maximum temperatures range between 21 and 29 C and in the interior, they range between 30 and 33 C respectively. The study area lies within the rainforest vegetation belt (Ezemonye & Emeribe, 2012) where the forests have been disturbed over the years due to demographic pressures. The soils are: cohesionless, have high sand contents and are very permeable with high infiltration rates of up to 3571 mm hr À1 (Chiemelu et al., 2013;Obi & Asiegbu, 1980;Okorafor et al., 2017).
Four geological formations underlay the study area: Imo Shale, Benin, Ogwashi-Asaba and Ameki Formations (Usman et al., 2014). These geological formations are sandy with intercalations of clay. Two major rivers, Njaba and Orashi have their sources within the study area ( Figure 1). Other streams such as Okpii and Ezizi also have their sources located within the study area.

| Gully mapping and geomorphic analysis
High-resolution imagery (with pixel sizes of 2 Â 2 m and 5 Â 5 m) covering the study area were obtained for 2009, 2014 and 2018 (Table 1) to map changes in gully characteristics. Gully mapping commenced in 2009 when freely available high-resolution satellite data of the study area became available. Fieldwork was conducted between April and   (Table 1). Gullies were found close to each other in all catchments where more than one gully was mapped (e.g., some gullies T A B L E 1 Satellite imagery used for gully mapping and land-use classification were <10 m apart in some catchments). Considering the 30 m resolution of the DEM and gully distance of <10 m from one another, the DEM may not likely delineate two different catchments for both gullies.

Tributary gullies (referred to by the villagers as gully-fingers). A gully
might flow directly into another and during mapping, more than one gully is identified but the same catchment feeds both gullies.
3. Extensive vegetal cover can make a single gully appear in two different parts, and thus, during mapping, more than one gully is identified, whereas the gully is a single continuous gully which had been apparently separated into different units by vegetation.
Gully catchments were named based on the location of the gullies in the LGA or autonomous communities. To measure gully head distance from rivers, a line was drawn from the gully head to the river. Another line was drawn from gully mouths to rivers, hence representing gully mouth distance from the river.

| Land-use classification
Satellite data listed in Table 1 were used for supervised land-use classifications. Land use was classified at the regional and catchment an example of land-use change; roads were digitised from the satellite imagery using a line shape file. The Euclidean distance between gully head and road was calculated representing gully head distance from road.

| Statistical analysis
Principal component analysis (PCA) was used to quantify the relative importance of gully drivers (maximum slope, relative relief, nearness to rivers and roads). Although geomorphic variables were collected at all gully heads, gullies that were not within 1 km of a river or road were removed from the PCA. This threshold was selected based on the average distance of gully heads from rivers and roads (Section 3.3). To ensure a uniform number of input variables for statistical analyses, only gullies that satisfied this condition of nearness to rivers and roads were added to the PCA. Where for example a gully was <300 m from a road but >3 km from a river, the gully was excluded from the PCA. Attributes from 29 gullies were included in the PCA. Because upslope contributing areas were not defined for some gullies that satisfied the above condition (nearness to rivers and roads) and judging from the nature of the satellite data used in this study (Section 2.3), two sets of multiple regression analysis were used to relate gully changes to gully drivers.

| Focus group meetings
To learn about community knowledge on gullying, fill existing gaps due to unavailability of long term remotely sensed data and improve  (Table 2). With regards to land-use change in gully catchments, (see Table 1  These results show that whilst there is an overall change on a specific land use at the regional scale, land use is heterogeneous within individual catchments. Due to land use heterogeneity, the overall patterns of regional-scale land-use change are not manifested at smaller scales within all gully catchments, and thus the effect of land-use change on gully-driving processes is also heterogeneous. slopes were observed around the rivers with slope rises of 10 to 58.2% over distances less than 500 m from the rivers.

| Statistical analysis
Eigenvectors of the PCA indicate that along the first principal component, maximum slope and relative elevation have equal contribution while gully head distance from river and gully head distance from roads contribute towards the second principal component (Table 5)    Gully-head retreat rates of the 'old gullies' reported in this study are low compared to some results found in other parts of southeast Nigeria where gully retreat rates between 30 and 60 m have been documented (Egboka & Nwankwor, 1985;Hudec et al., 2005).  Figure 4 shows proximity of three visited gullies to roads in the study area. Associations between nearness to roads and gully evolution are well established (Collison, 2001;Frankl et al., 2012) and results from this study agree with referenced studies. Wrong termination of drainage channels by contractors often associated with road construction facilitates gullying. Nwankwor et al. (2015) noted that soils in southeast Nigeria were not easily erodible as is hitherto believed because most gullies in the region can be traced back to improper termination of road runoff concentration. Concentrated surface runoff from these drainage channels (especially when the endpoints of such channels are nearby bushes, as is observed in the study area) potentially initiates new gullies. In the first month of fieldwork (April 2019), no gully was identified at the Okwudor axis of the Owerri-Orlu highway. However, in June 2019, a gully had destroyed part of the highway ( Figure 4a). A villager informed us that the gully started due to construction of road drainage which was terminated in the nearby bush.

| Community/local knowledge of gully drivers
Focus group meeting participants informed us that one of the gullies identified in Obibi-Ochasi was less than 2 years old as at the time of our field visit. This gully (Figure 4b) was said to have started due to concentrated runoff from road drainage (road drainage is visible in Figure 4b) following abandonment of road construction. These examples support the findings of Nwankwor et al. (2015) regarding control of road drainage on gully formation in southeast Nigeria.

| Community knowledge
Although increase in demographic pressure has been suggested as a gully driver (Fanciullacci, 1978), the role of civil war first as a catalyst of increased demographic pressure and secondly, as a driver of gully erosion has not been documented previously in the study area. The Orlu area was considered a safe haven and provided shelters for displaced war refugees from other parts of southeast  Poesen, 2010;Poesen et al., 2003) and the susceptibility of soils of southeast Nigeria to dispersion by erosive forces due to their composition has been suggested (Egboka & Nwankwor, 1985;Idowu & Oluwatosin, 2008;Okagbue & Ezechi, 1988).

| The value of a multi-method approach
While traditional geomorphic methods (e.g., analysis of remotely sensed data and quantitative analyses) are important to establish correlations between gully-drivers and gully erosion, in some cases, they are insufficient to provide better understanding of gullying, for example gully commencement dates or forcing factors that led to gully initiation. As an example, in the study area, three highresolution satellite data were available, and within intermediate Within the study period (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018), reductions in fallow-cover across the study area were identified and gully characteristics also changed in response to these land-use changes. Although it has been suggested that while it is possible for land-use changes to occur in one direction across a region (e.g, reduction in tree-cover), it is possible for changes in land use in the opposite direction (e.g., increase in tree-cover) to occur within individual catchments. Finally, this work has shown the importance of within and between catchment heterogeneity, a widespread characteristic with implications for understanding and managing gully erosion.

ACKNOWLEDGEMENTS
We express our thanks to all the participants of focus group meetings in Amucha and Obibi-Ochasi. We are grateful to the Tertiary Education Trust Fund and the Geography Department, Durham University who supported the travel grant for fieldwork.