Gully internal erosion triggered by a prolonged heavy rainfall event in the tableland region of China's Loess Plateau

Jiaxi Wang , Yan Zhang ,*, Kunheng Li , Ziqing Zhang , Chang Chen

a College of Soil and Water Conservation, Beijing Forestry University, Beijing 100083, China

b Forest Ecosystem Studies, National Observation and Research Station, Jixian, Shanxi 042200, China

Keywords:Gully erosion Mass movement Heavy storm Loess Soil and water conservation measures

A B S T R A C T Gully erosion is a severe form of soil erosion,but gully internal erosion processes are poorly understood,especially at the event scale.To investigate gully internal erosion intensity and understand the related gully development mechanism in an agricultural environment with gully head stabilization and vegetation restoration efforts, two successive field investigations were carried out just before and after a prolonged rainfall event in 2021 in the tableland region of China's Loess Plateau.Thirteen gullies were investigated and all experienced gully internal erosion, while most gully boundaries were stable during the heavy rainfall event based on the comparison of the UAV digital orthograph maps (DOMs acquired with Unmanned Aerial Vehicle) before and after the rainfall event.The proportion of gully internal erosion area to gully internal area of the 13 investigated gullies ranged from 3 to 55%,with average areal erosion proportion of the gully sidewall and gully bed of 21% and 36%, respectively.The erosion area of subdrainage units (SDUs) on the gully sidewall was positively correlated to the SDU area, average SDU slope gradient and vegetation type, while the erosion area on the gully bed was positively correlated to the gully area, gully depth and gully bed slope gradient.Gully internal erosion was not significantly correlated with gully drainage area because the connectivity between the upslope and gully areas was interrupted and the effective drainage area of the gully was obviously reduced by soil erosion conservation measures, including terraces on the upslope drainage area, shrub belts, and water barriers.Thus,gully internal erosion is still active under the heavy rainfall storm against the background of the ‘Grain for Green’ and ‘Gully Stabilization and Tableland Protection’ programs, and integrated measures for preventing both gully expansion and gully internal erosion must be further enhanced in the context of climate change.

1.Introduction

Gully erosion is often associated with a wide range of on-site and off-site environmental and ecological impacts, such as destroying land resources, changing catchment hydrology and reducing downstream reservoir capacity due to sediment deposition (Poesen et al., 2003; Vanmaercke et al., 2021).Rainfall and runoff are known to initiate and exacerbate gully erosion processes(Anderson et al.,2021;Castillo&Gˊomez,2016;Poesen et al.,2003;Torri&Poesen,2014;Valentin et al.,2005;Vanmaercke et al.,2016,2021).Given global climate change, extreme rainfall events are intensifying(Madsen et al.,2014; Moustakis et al.,2021; Yin et al.,2018), which may lead to an increased risk of gully erosion (Li &Fang, 2016; Valentin et al., 2005; Vanmaercke et al., 2016; Zhang et al., 2019).Consequently, there is a need for monitoring, experimental and modeling studies of gully erosion as a basis for gully erosion control under the effects of climate change (Poesen et al.,2003; Valentin et al.,2005; Vanmaercke et al., 2021).

Dominant gully erosion processes include gully head retreat,sidewall expansion and gully bed incision(Blong et al.,1982;Kirkby& Bracken, 2009;Sidorchuk,1999;Vanmaercke et al.,2016,2021).These processes are frequently observed to be associated with piping and tunnel erosion(Wilson,2011;Zhu,2012),landslides and gravity erosion (Dong et al., 2019; Xu et al., 2020).Overall, gully erosion appears to be a highly complex phenomenon in both its temporal and its spatial dimensions, requiring a detailed assessment (Castillo & Gˊomez, 2016).In previous studies, gully development was usually measured on successive field observations or satellite images (Frankl et al., 2012; Vandekerckhove et al., 2003;Wu & Cheng, 2005).Recently, widespread unmanned aerial vehicles (UAVs) have become a flexible and cost-efficient method for acquiring surface information (James & Robson, 2014).As a topographic modeling technique, structure-from-motion (SfM) photogrammetry combining the utility of digital photogrammetry with flexibility and ease of use derived from multi-view computer vision methods (Carrivick et al., 2016; James et al., 2017; Meinen &Robinson, 2020) has boosted the establishment of UAVs as a powerful tool for gully erosion research(Guan et al.,2021;Meinen&Robinson,2020;St¨ocker et al.,2015;Yuan et al.,2020;Zhao et al.,2021).Several previous studies have already adopted digital elevation models (DEMs) with as high as cm-accurate resolution calculated by point clouds derived from the SfM algorithm for erosion rate estimations, morphological analyses and gully development monitoring (Liu et al., 2016; Meinen & Robinson, 2020;St¨ocker et al., 2015).Many monitoring, modeling and prevention studies have focused on gully initiation and gully head retreat under extreme rainfall events based on UAVs (Anderson et al.,2021; Yuan et al., 2020; Zhang et al., 2019).However, changes in gully internal erosion processes, e.g., gully sidewall mass movement and gully bed incision, are poorly understood at the event scale because it is logistically difficult to be at the right place and time to capture an extreme rainfall event in field observations(Anderson et al., 2021).

Gully erosion occurs only when a threshold in terms of flow hydraulics, rainfall, topography, pedology and land use has been exceeded (Poesen et al., 2003).Drainage area, as a surrogate for surface or subsurface runoff, is one of the most important factors influencing gully erosion (Vanmaercke et al., 2016).Usually, soil and water conservation measures (SWCMs) in drainage areas can change hydrological connectivity and reduce effective drainage areas to prevent gully erosion (Arnˊaez et al., 2015; Wang, Zhang,et al., 2021); however, some studies have reported that SWCMs,e.g., terraces, can induce gully erosion (Arnˊaez et al., 2015; Wen et al., 2021).Therefore, under the condition of widely applied SWCMs on slopes, how gully internal morphology will change under extreme rainfall events is still unclear(Anderson et al.,2021).The Chinese Loess Plateau, which experiences severe soil erosion that leads to a series of ecological and economic problems,is a good example for understanding this issue (Chen et al., 2015; Li et al.,2017).As an important soil erosion process on the Chinese Loess Plateau, gully erosion could contribute 60-90% of the total sediment production in agricultural areas(Li et al.,2003;Wu&Cheng,2005).Since the‘Grain for Green’program(GFGP)was launched in 2000, substantial SWCMs have been implemented, resulting in significant land use changes and vegetation cover increases(Wang et al., 2015; Xia et al., 2019).In addition to the GFGP program, the‘Gully Stabilization and Tableland Protection’ program (GSTP) was launched in the 2010s to stabilize the channel and control gully erosion in the tableland region,and the GSTP includes shrub belts,water barriers, terraces, check dams, and other SWCMs (Wang,Zhang, et al., 2021).Gully head retreat rates have been decreasing since the implementation of the GFGP and GSTP in the tableland region (Wang, Zhang, et al., 2021).During the 2000s, the decadal average soil erosion rates of the Loess Plateau declined to a historically low status in the past 100 years; however, erosion rates increased between 2010 and 2016, mainly driven by extreme rainfall events(Li et al.,2022).Serious gully erosion in agricultural watersheds with vegetation restoration triggered by extreme rainfall events has also been reported(Wang et al.,2016;Yuan et al.,2020;Zhang et al.,2017,2019;Zhao et al.,2021).However,detailed investigations of gully internal erosion under conditions of vegetation restoration are still absent, especially at the storm-event scale.

In this study, based on two successive field investigations with UAV photogrammetry just before and after a prolonged heavy rainfall event,we intended to assess gully internal erosion intensity and understand the related mechanism of gully development in an agricultural environment with enormous efforts focused on gully head stabilization and vegetation restoration.The main objectives were to 1)estimate the erosion intensity of gully head retreat,gully sidewall mass movement and gully bed incision triggered by a heavy rainfall event; 2) investigate the effects of the geomorphological influencing factors on gully internal erosion under the conditions of gully head stabilization and vegetation restoration;and 3) discuss the effects of SWCMs and vegetation restoration on gully erosion.

2.Materials and methods

2.1.Study area

The study area is on the southeastern Loess Plateau located in Jixian County, Shanxi Province, China (Fig.1).There is a typical temperate continental monsoon climate with an average annual precipitation of 507.47 mm(1957-2019)(based on meteorological records from the China Meteorological Data Service Centre[CMDC],http://data.cma.cn/).Approximately 70% of precipitation events occur from June to September.The loess cover is approximately 150 m thick,and the main soil type is loessial carbonate cinnamon soil with weak alkalinity in the study area.The geomorphology of the study area is characterized by broad and flat tablelands threatened by surrounding deep gully systems(Wang,Zhang,et al.,2021).Gully erosion induced by concentrated flow from the tableland is the main source of sediment yield in the tableland region (Guo et al., 2020).Due largely to its flat topography and higher clay content in the loess and soils, loess tableland is very important for agriculture on the Loess Plateau.Most of the tableland in the study area was reclaimed as farmland or apple orchards.To stabilize gullies and control soil erosion,the GFGP was launched in 2000, and the GSTP was launched in the 2010s.Many SWCMs,including terraces, shrub belts, and small reservoirs, were constructed in the study area,and vegetation has been restored in the valley area due to the GFGP and GSTP(Wang, Zhang, et al., 2021).

A prolonged rainfall event occurred in the middle reach of the Yellow River, including the study area,from September to October 2021.According to the rain gauge records at the Jixian National Observation and Research Station, three single rainfall events occurred from 16 September to 14 October, with 97.6, 123.8 and 113.8 mm rainfall amounts and 5-, 8- and 5-day durations,respectively (Fig.2).There were three rainfall peaks on 18 September, 25 September and 4 October, with daily rainfall amounts of 63.4,45.8 and 33.2 mm,respectively.The maximum 60-min rainfall intensity(I60)was 7.2,8.2 and 7.4 mm h-1on 18 Sept.,24 Sept., and 3 Oct., respectively.Triggered by this strongest prolonged rainfall event on record in Shanxi Province,37 rivers in the province flooded,and the largest flood in the lower Fen River since 1954 was observed at the Hejin hydrological station(HJ),resulting in serious soil erosion and damage (https://backend.chinanews.com/shipin/spfts/20211011/3651.shtml).

2.2.Data acquisition and parameter calculation

2.2.1.UAV photogrammetry and data processes

Two successive field investigations were carried out just before and after the prolonged rainfall event in 2021(Fig.2).The first field survey was conducted on 8-14 September 2021.A UAV (DJI inspired 2 equipped with DJI zenmuse X4S camera) was used to acquire images in this survey with a 0.14 m ground sampling distance (GSD) and a flying height of 500 m, covering approximately 50 km2.After the rainfall events,we immediately implemented the second field survey on 13-15 October and selected some typical gullies for detailed photogrammetry by the same UAV, and the flying height was approximately 30-50 m with a 0.01-0.05 m GSD.In both the first and the second field surveys,UAV photogrammetry had an 85% frontal overlap and at least 70%side overlap to ensure point cloud and DEM quality.

Pix4Dmapper 4.4.12 software was used to automatically mosaic and process UAV images to generate digital orthograph maps(DOMs) and DEMs before and after the rainfall events.The same UAV was used in the two successive field investigations; however,DOMs and DEMs had a small shift due to the relative error of the on-board GPS between the two investigations.Therefore, twophase DOMs and DEMs were georeferenced with 3-5 ground control points (GCPs) visible in both (e.g., feature points of the roads, houses).Generally, the total RMS error of GCPs are<0.0001 m, which is far less than the GSD of the UAV.

2.2.2.Gully extraction and parameter calculation

Thirteen gullies were selected based on two considerations: 1)gully edges were clearly visible on both UVA images, and 2) they were distributed in the study area as evenly as possible (Fig.1b).

Based on a comparison of the DOMs before and after the rainfall event (Fig.S1 and Fig.S2), we detected the gully head retreat and gully sidewall expansion area by interpreting the change in gully boundaries visually from the DOMs(Fig.3a and b);this method was used for gully boundary identification at cm accuracy in previous research (Marzolff et al., 2011; Wang, Zhang, et al., 2021).An area where vegetation was clearly damaged or had new erosion marks in the gully(Oostwoud Wijdenes et al.,2000)was identified as the gully internal erosion area.

Based on DEMs generated from UAV images,each gully internal area was subdivided into several geomorphological units(El Khalili et al., 2013).First, the internal area of each gully was divided into the gully bed and gully sidewall to calculate the area of the two parts (Agband Ags, m2).Then, subdrainage units on the gully sidewall (SDU) were identified by the D8 flow model based on the DEMs (Fig.3c) (Greenlee, 1987; Jenson & Domingue, 1988).The gully bed was taken as a single SDU.The SDU area (Asdu, m2), upslope drainage area of each SDU(DAsdu,m2),drainage area of each gully(DA,m2)and thalweg of the gully were determined based on the D8 flow model.For each of the 13 gullies, the gully internal erosion area (EA, m2) and the proportion of gully internal erosion area to gully internal area (ER, %) were calculated (Fig.3a and b).Similarly,the area of gully bed erosion and gully slope erosion and the corresponding proportion(EAb and EAs,m2;ERb and ERs,%)as well as the erosion area and the proportion of SDU erosion area to corresponding SDU (EAsdu, m2; ERsdu, %) were calculated.

For each gully,a profile was obtained based on the thalweg,and three transects perpendicular to the thalweg were extracted based on the DEMs before the rainfall events.The gully length (Lg, m),width (Wg, m), depth (Dg, m), and gully bed gradient (Sgb,°) were calculated based on the profiles and transects.The average slope gradient of the gully sidewall (Sgs,°), the average slope gradient(Ssdu,°)(Fig.3d)and the average profile curvature(Csdu)of the SDU were extracted based on the resampled DEMs(1×1 m).To evaluate the effect of the vegetation on gully erosion, the vegetation types inside the gullies before the rainfall events were interpreted as bare land,grassland,shrubland,and forestland based on the DOMs,and the land uses as well as conservation measures on the drainage area of the gullies were classified, which were verified with the field investigation.

2.3.Statistical analysis methods

Based on Pearson's correlation coefficient, we evaluated the effects of the SDU area and SDU drainage area on erosion.Based on the rank correlation coefficient(Spearman's rho),the effects of the gully area, drainage area of the gully, slope of the gully bed, gully width, depth, and length on the gully bed incision were evaluated in consideration of the small number of samples of the gully bed(n = 13).Using ANOVA, we compared the differences in the proportion of erosion area among the slope classes, SDU shapes (profile curvature), and vegetation types.The Levene statistic was calculated first for the homogeneity of variance test,and if p<0.05,Welch-ANOVA was used; otherwise, one-way ANOVA was used.

3.Results

3.1.Gully internal erosion during a prolonged heavy rainfall event

The average gully width,gully depth,and gully length of the 13 investigated gullies were 57.53, 33.48, and 149.34 m, respectively.The inner gully area, ranging from 2340.21 to 23,051.36 m2, was divided into two parts, with the proportions of the gully sidewall area ranging from 62.1 to 88.0%and the gully bed ranging from 37.9 to 12%.The average slope gradients of the gully sidewall varied from 32.87 to 55.28°(Table 1).Before the storm, forests or shrubs occupied more than 50%of the sidewalls of the seven gullies.Four gullies had more than 50%grassland,and one gully had nearly 50%bare land on the gully sidewall(Fig.4a).The average slope gradients of the gully bed varied from 27.40 to 41.60°.Eight gullies had more than 70% forest, while others had mainly grass or shrubs on the gully bed (Fig.4b).The drainage area of these gullies ranged from 158.50 to 207,781.60 m2,with an average of 36,729.98 m2.The land use types of the drainage area were dominated by apple orchards,cropland for corn (Zea mays), and intercropping corn in apple orchards,and all of the drainage areas of the 13 gullies were terraced,with shrub belts on the boundary of some gullies.

After the prolonged heavy rainfall event, most gully edges were relatively stable(except for the severe gully head retreat of the No.9 gully with a growth area of 103.38 m2and many new tunnels formed on the upslope of the gully),while gully internal erosion was found in all investigated gullies, with the dominating erosion processes being gully sidewall mass movement and gully bed incision(Fig.5).The proportion of gully slope erosion area to gully slope area (ERs)ranged from 3 to 55%,with an average of 21%,while the proportion of gully bed erosion area to gully bed area(ERb)ranged from 6 to 99%,with an average of 36%, showing that the proportion of gully bed erosion area was larger than the proportion of gully sidewall erosion area (Fig.6).The internal gully area covered mainly by forests and shrubs had the smallest ER, while a large ER occurred where bare land or grassland was dominant in the gully area.

3.2.Influencing factors of gully internal erosion

3.2.1.Gully sidewall mass movement

Among the 193 SDUs delineated on the gully sidewalls of the 13 gullies,152 SDUs were eroded mainly by mass movement,with an average EAsdu of 142.5 m2and an average ERsdu of 30%(Table 2).No significant Pearson's correlation was found between DAsdu and EAsdu,but Asduwas correlated significantly with EAsdu at the level of 0.01 (r = 0.699, n = 193).

A significant correlation was found between ERsdu and Ssduat the level of 0.01 (r= 0.242**,n= 152),and significant differences were detected among the four slope classes with Welch-ANOVA(F = 3.548, p < 0.05) (Fig.7a), which showed an increasing trend of ERsdu with increasing Ssdu.The shape of the SDU was classified into 2 types:concave and convex.The average ERsdu was 31.1%and 28.8%in the concave and convex SDUs,respectively.However,oneway ANOVA showed that the average ERsdu between the two SDU shapes was not significantly different (Fig.7b).The ERsdu in different vegetation types in each SDU was calculated and counted(n=272)(Fig.7c).The average ERsdu values in bare land,grassland,shrubland, and forestland were 53.5%, 34.8%, 39.4%, and 30.8%,respectively.Welch ANOVA showed a significant difference in the average ERsdu among the vegetation types (F = 5.877, p < 0.01).Post hoc tests indicated that vegetation had an effect on gully sidewall erosion during the prolonged rainfall event.

3.2.2.Gully bed incision

The gully bed erosion area (EAb) was significantly correlated with the gully dimensions, i.e., AgWg, Dgand Lg(Table 3), but was not significantly correlated with the gully bed gradient (Sgb).However, the proportion of the gully bed erosion area (ERb) was correlated significantly with gully depth and gully bed gradient(Sgb) but not significantly with gully dimension parameters.The results of one-way ANOVA showed that the average ERb among the vegetation types was not significantly different (Fig.7d).These results indicated that the gully area, gully bed gradient and gully average depth were the important influencing factors of gully bed erosion during the prolonged rainfall event.

Neither EAb nor ERb were correlated significantly with the drainage area of the gully (DA) because the connectivity between the upslope and gully areas had been interrupted and the effective drainage area of the gully had been obviously reduced by multiple efforts to improve gully head stabilization in the study area.Roads and ditches were built along the gully edges to change the flow direction, which directly reduced the drainage area (see Fig.S1 in the Supplementary Material gullies).For some gullies, the effective drainage area became very small (Fig.5).Shrub belts were established along the gully boundary to increase infiltration and enhance the gully slope, thus reducing the effective drainage area.Terraces and orchards on the upslope of the gully reduced the runoff energy to the gully by changing the slope gradient and increasing infiltration.Therefore, the original geomorphology and drainage system were changed greatly so that the gully internal erosion was affected slightly by the upslope area and conditions(i.e., see Fig.S2 in Supplementary Material gullies).

Table 1 Basic information on the gully morphology and erosion.

Fig.4.The proportion of each vegetation type on the (a) gully sidewall and (b) gully bed in the 13 investigated gullies.

4.Discussion

4.1.Gully internal erosion processes triggered by a heavy storm

During this prolonged rainfall storm event, severe erosion occurred inside most of investigated gullies, with the dominating erosion processes being mass movement on the gully sidewall and gully bed incision; however, no obvious gully head retreat was observed for most investigated gullies.Gully sidewall erosion is influenced by both concentrated flow(Qin et al.,2018;Wells et al.,2013)and soil saturation(Xu et al.,2015,2020),which are related to gully sidewall retreat, tension cracks, mass movement and so on(Martínez-Casasnovas et al., 2009).Mass movement on sidewalls was the main process contributing to total sediment production(Martínez-Casasnovas et al.,2003,2009)and to gully initiation and development(de Bacellar et al.,2005;Kirkby&Bracken,2009;Zhu,2012).In this study, the average slope of the SDU on the gully sidewall, mainly ranging from 30 to 60°, was in the range most susceptible to slides on the Loess Plateau (Zhu, 2012).Regarding gully bed erosion, previous studies have shown that gully bed incision is an important process in gully development (Kirkby &Bracken, 2009; Simon & Rinaldi, 2006) that is mainly influenced by concentrated flow and surface runoff during rainfall events(Dong et al., 2018).Numerous experimental studies have been performed on gully bed incisions(e.g.,Bennett&Wells,2019;Chu-Agor et al.,2008;Qin et al.,2019;Wells et al.,2013),but monitoring research on gully bed incisions is still limited.In this study, such severe gully bed incision was found after a heavy rainfall event, in which some sparse trees were removed from the gully beds(Fig.5).Both the large runoff generated from this prolonged rainfall event and the easily eroded deep loess at the gully bed may be responsible for the active gully bed incision.In addition to gully wall mass movement and gully bed incision,gully development is frequently observed to be associated with piping and tunnel erosion(Bernatek-Jakiel & Poesen, 2018; Valentin et al., 2005; Zhu, 2012).In this study, tunnel erosion was also found in gully No.9, which was related to gully head retreat.

Gully internal erosion varies greatly in different parts of a gully,so we subdivided the gully internal area into SDUs using the methods presented by El Khalili et al.(2013) to explore the mechanism of gully internal erosion.An SDU is similar to a hydrological response unit (HRU) or an erosion response unit (ERU) at the catchment scale (Beighley et al., 2005; Fenicia et al., 2016; Flügel,1995; Sidorchuk et al., 2003).The SDU derived from highresolution DEMs provides a possibility for exploring the details of geomorphology and hydrology inside a gully.

Although cm-accurate resolution DOMs were used, only the gully internal erosion area on which vegetation was damaged could be identified; therefore, the gully internal erosion area may have been underestimated.Owing to the complex vegetation cover, we failed to calculate the accurate gully erosion volume with the DEM generated by UAV photogrammetry because many ground points were missing in the point cloud due to vegetation sheltering,which could result in uncertainty in the multiscale model to model cloud comparison (M3C2) method (Lague et al., 2013).Nonetheless, the erosion area was relatively accurately obtained from highresolution UAV photogrammetry,which allowed us to observe gully erosion dynamics at the storm-event scale.

Fig.5.Examples of subdrainage units(SDUs)and erosion areas in gullies dominated by 4 vegetation types.(a)-(h)are digital orthograph maps(DOMs)derived from the UAV before and after this rainfall event.(a) and (b) show a typical bare land dominating gully, (c) and (d) show a typical grassland dominating gully, (e) and (f) show a typical shrub land dominating gully, and (g) and (h) show a typical forestland dominating gully.

Fig.6.The proportion of gully bed erosion area to gully bed (ERb) and the proportion of gully sidewall erosion area to gully sidewall (ERs) in the 13 investigated gullies showed that ERb was larger than ERs.

4.2.Effect of geomorphological influencing factors on gully internal erosion

Usually, drainage areas were widely used as surrogates for surface or subsurface runoff in previous studies of gully erosion(Montgomery&Dietrich,1994;Rossi et al.,2015;Vanmaercke et al.,2016).Given the low hydrological connectivity between the upslope and most parts of the gully internal area due to SWCMs in this study (cf.Section 4.3), the SDU area could roughly be used as a surrogate for runoff generated from the SDU.As expected, in this study, Asduwas correlated significantly with EAsdu (r = 0.699**).Slope gradient is usually considered a factor for gully erosion.Our findings suggest an increasing trend of ERsdu with increasing Ssdu,which agrees with the results of El Khalili et al.(2013).No significant difference in erosion area was found between concave and convex SDUs in this study.However,a previous study showed that the profile shape was particularly susceptible to shallow landslide occurrences on the Loess Plateau (Zhuang et al., 2018).This result may be attributed to the short slope length in the gully internal area of the tableland terrain.Gully area was correlated significantly with EAb (ρ = 0.802**, n = 13), implying that it was mainly caused by runoff generated from the gully area instead of the upslope drainage area under the conditions of the GFGP and GSTP.Another factor influencing this hydrological process is the slope gradient.As expected, ERb was correlated significantly with Sgb(ρ = 0.599*,n = 13).Furthermore, a significant correlation was found between ERb and Dg(ρ=0.775**,n=13).These results are consistent with the results of previous experimental studies, i.e., the higher the gully head height was,the more severe the gully bed scouring and incision (Wang, Li, et al., 2021; Zhang et al., 2018).These results indicate that the geomorphological characteristics of the gully itself have a great effect on gully internal erosion in the GFGP and GSTP contexts.

4.3.Effect of vegetation restoration and gully head protection on gully internal erosion

Although the high potential of vegetation to protect against gully erosion has been widely acknowledged(Li et al.,2015;Poesen et al., 2003; Valentin et al., 2005; Vanmaercke et al., 2021; Wang et al., 2022; Yuan et al., 2020), severe soil erosion was observed inside gullies in this study.However, the gully head retreat was limited, and the gully boundaries were relatively stable due to the many efforts made to stabilize gully heads.All of the investigated gullies had SWCMs on the drainage area and along the gully boundaries established by the GFGP and GSTP, e.g.,terraces, shrub belts, and water barriers.Reducing hydrological connectivity between upslopes and gullies will decrease peak flow onto gully slopes,which may mitigate gully slope erosion(Arnˊaez et al.,2015).SWCMs can decrease hydrological connectivity, which will reduce direct runoff into gullies and influence gully erosion processes(Bracken et al.,2013;Conoscenti et al.,2018;Fryirs,2012).For some circumstances,the runoff from the upslope area into gullies seemed to be switched off by the SWCMs so that the gully boundaries remained stable.However, for some other circumstances, runoff was diverted by the SWCMs from the gully head to the gully side so that parts of the gully slopes and gully beds were scoured seriously(e.g.,Fig.8).Runoff generated from both the drainage area and the gully slope accumulated onto the gully bed,leading to severe gully bed erosion.The results of this study showed that the proportion of the gully bed erosion area was larger than the proportion of the gully sidewall erosion area in most investigated gullies, eventhough the gully beds were covered by dense vegetation.

Table 2 The basic information of the subdrainage units (SDUs) on the gully sidewalls.

Fig.7.(a) The proportion of the SDU erosion area to SDU area (ERsdu) in different average slopes of the SDU (Ssdu) class; (b) ERsdu in the concave and convex profile curvature (Csdu); (c) ERsdu in the different vegetation types in the subdrainage area unit (SDU); (d) the proportion of gully bed erosion area to gully bed (ERb) in the different vegetation types in the gully bed.Two completely different characters indicate significant differences between the corresponding groups (a, b, ab).

Regarding the effect of vegetation on gully internal erosion,vegetation conditions inside gullies can indicate whether a gully is likely active (Golosov et al., 2018; Imwangana et al., 2015;Oostwoud Wijdenes et al., 2000; Wang et al., 2022).Before this rainfall event,many parts of the gully sidewall and gully beds were covered by forests and shrubs,indicating that they had been stable for a long time, while some small pieces of bare land on the gully sidewall implied that small mass movements occurred not long before the storm event.However,triggered by this prolonged storm event,severe erosion was found on the gully sidewall and gully bed mainly in the form of mass movement so that some forest on the gully bed was washed away, suggesting that the effects of high ground vegetation coverage on preventing erosion on steep slopes could be limited.Moreover, vegetation type may be more important than vegetation coverage for gully control (Chen et al., 2018).The results of this study showed that the average proportion of gully internal erosion area on grassland was significantly higher than that on shrubland and forestland.

This study showed that vegetation restoration by the GFGP and GSTP has had a great positive effect on preventing gully head retreat and mitigating gully internal erosion, but gully internal erosion may still be active as a result of heavy rainfall events.Therefore,integrated measures for preventing both gully expansion and gully internal erosion need to be further studied in the context of climate change.

5.Conclusion

Based on two successive field surveys with UAV photogrammetry in the scope of a prolonged rainfall event in the tableland region of China's Loess Plateau, severe gully internal erosion was found in most of the investigated gullies,but the gully heads did not retreat obviously,and most of the gully edges remained stable.The proportion of gully internal erosion area to gully internal area of the 13 investigated gullies ranged from 3 to 55%, with average areal erosion proportion of the gully sidewall and gully bed of 21% and 36%, respectively.The main erosion processes were mass movements on the gully sidewall and gully bed incision.Subdivision of the gully internal area into SDUs based on very high-resolution DEMs was proven to be a method for gaining insight into the factors of gully internal erosion.The results showed that the erosion area of SDUs on the gully sidewall was positively correlated to the SDU area,average SDU slope gradients,and vegetation types,while the erosion area on the gully bed was positively correlated to the gully area,gully depth and gully bed slope gradients.Gully internal erosion was not correlated significantly with the drainage area of the gully because the connectivity between the upslope and gully areas had been interrupted and the effective drainage area of the gully had been obviously reduced by multiple efforts taken to stabilize gully heads in the study area.The SWCMs established by the GFGP and GSTP,including terraces on upslope drainage areas,shrub belts, and water barriers, had a great positive effect on preventing gully expansion and mitigating gully internal erosion.However,gully internal erosion is still active under heavy rainfall storms.The findings of this study suggest that integrated measures for preventing both gully expansion and gully internal erosion need to be further studied in the context of climate change.

Table 3 Spearman's rho between gully bed erosion and gully morphological parameters(n = 13).

Fig.8.Examples of the flow path and the erosion area distribution.The background is the hill shade effect of the DEMs, and the DOMs of the gully were clipped.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This study was supported bythe National Natural Science Foundation of China (42177309) and the State Key Program of National Natural Science of China (42130701).

Appendix A.Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.iswcr.2022.12.003.