1.1 Soil and water conservation on the Loess Plateau

Soil erosion is a prime concern for environmental protection on the Loess Plateau which is located in the upper and middle reaches of the Yellow River and spans an area of 640,600 km². Due to the high erodibility of loess and the incision of the Yellow River and its tributaries, the Loess Plateau is one of the most severely-eroded soil regions in the world (Wang et al., 2021). According to the Soil and Water Conservation Bulletin of the Yellow River Basin (YRCC of CMWR Yellow River Conservancy Commission of Ministry of Water Resources of People's Republic of China, 2024), the area of soil erosion on the Loess Plateau is 223,600 km², including a water-induced erosion area of 170,200 km² and wind-induced erosion area of 53,400 km². The classification of the water-driven erosion intensity on the Loess Plateau is shown in Table 1. Soil and water conservation practices have been extensively conducted on the Loess Plateau since the 1950s, mainly including engineering measures (e.g., construction of check dams) and biological methods (e.g., ecological restoration) (Wen & Zhen, 2020).

1.2 Development of check dams

The history of check dam construction on the Loess Plateau can be traced back to the Ming Dynasty in the 16th century (Ai & Yang, 2018; Wei et al., 2017). These ancestors secured a stable grain supply through “damming and reclaiming” (Chen et al., 2020). Concurrently, many gullies were transformed into terraced croplands, which significantly curbed soil erosion (Dang et al., 2020; Liu et al., 2014). By the end of 2019, there were 58,776 check dams on the Loess Plateau (CMWR Ministry of Water Resources of People's Republic of China, 2020a; Liu & Wang, 2020). Table 2 and Figure 2 show the distribution of check dams on the Loess Plateau.

1.3 Impact of inadequate flood control facilities

Check dams, particularly those built by local residents on the Loess Plateau before 1986, were typically designed and constructed without sufficient flood-control facilities. Most of these check dams do not contain spillways (Liu & Wang, 2020; Yu et al., 2019). Some large check dams have masonry spillways built on soil foundations without adequate bearing capacity. Horizontal concrete pipes buried in the dam for releasing the reservoir water have also been built on soil foundations, which have been designed to be unpressurized, being fragile against external hydraulic impacts. Damage or destruction of check dams by floods has been frequently reported, especially after extreme rainfall events that have been occurring more frequently in the Loess Plateau in recent years owing to climate change (e.g., Dang et al., 2019; Li et al., 2014; Yang et al., 2020). For example, 119 out of 217 check dams were damaged by a flood in the Jiuyuangou watershed on the northern Loess Plateau. This issue has been reviewed in detail in Section 3.

Debates on the pros and cons of building check dams on the Loess Plateau have been a long-lasting hot topic which has remarkably influenced the decision-making process for planning and constructing check dams on it. There were two periods of check dam construction surges: from 1968 to 1976 and from 2000 to 2009 (Chen et al., 2020; Liu et al., 2018), refer to Figure 3. Since 2009, there has been a long period of inactivity in terms of constructing check dams.

This study provides an overview of the various aspects of check dam construction in terms of its economic and environmental benefits and the countermeasures for dam safety controls over the past several decades, along with an outlook on its future development. Most of the references cited in this paper are from Chinese journals. They are readily accessible through web databases, such as CNKI. The digital object identifiers of the cited references, whenever available, have also been provided in the references section.

2.1 Land reclamation

On the Loess Plateau, sediment has been trapped in the gullies with the cascade check dams installed, creating arable land of 927.57 km² by the end of 2019 (Chen et al., 2020). Figure 4 shows the typical land reclamation due to check dam construction in the gulli. Table 3 summarizes the contributions of check dams to land reclamation on the Loess Plateau. In terms of quantity, 41,008 check dams have been completely filled by sediment, accounting for 69.77% of the 58,776 check dams on the plateau. In particular, the number of check dams completely filled with sediment built before 1986, between 1986 and 2003, and after 2003 were 28,198; 8624 and 4186, respectively. These numbers represent 87.97%, 62.15% and 32.59% of the check dams constructed during each corresponding period (Chen et al., 2020).

The land created by check dams is highly productive for agriculture, as opposed to terrace and slope land, owing to its leveled ground and high-water content (Fu et al., 2017; Wang et al., 2011a). Irrigation is ensured even during drought seasons because these lands enable the collection of precipitation in the gully. It is estimated that a dam land of 1 mu (1 mu = 667 m²) can replace the original slope land of 6–10 mu on the gullies, which will, in turn, be transformed into a forest and a grassland. With the implementation of the “Grain for Green” program in 1999, a total of 41.17 million mu slope land had been converted to forest and grassland. For example, in the Wangmaogou watershed in Yulin, Shaanxi Province, the forest area increased from 3% to 45% and the grassland area increased from 3% to 7% because of the new land created by the check dams (Chen et al., 2020).

2.2 Environmental impact

The check dam system has transformed gullies into terraces, reducing the kinetic energy of channel runoff and altering the erosion pattern of the watershed. This means that check dams not only retain sediment but also reduce the gravitational erosion potential of gullies by raising the erosion level. As a result, the entire watershed on the Loess Plateau has become greener. For example, the annual maximum normalized difference vegetation index and the annual average fractional vegetation cover in Shaanxi Province increased by 32.2% and 20.8%, respectively, from 2001 to 2017 (Li & Wang, 2020; Yin et al., 2021a). Green China (2020) reported that the green area expanded northwards by more than 400 km from 2000 to 2018 in Shaanxi Province (Figure 5). The forest coverage and living wood growing stock had reached 41% and 424 million cubic meters by the end of 2020 (Zhang et al., 2024). In addition, 300 administrative villages in Shaanxi Province were recognized as “National Forest Villages” by the National Forestry and Grassland Administration. By implementing the “Three-North Shelter Forest” program since 1978 (Zhai et al., 2023), the “Grain for Green” Program since 1999 (Huang et al., 2023) and other soil and water management projects, the advancement of the Mu Us Sandy Land has been halted (Gao et al., 2022).

With increasing improvements in the ecological environment, the Loess Plateau has become home to wild animals. The Ziwuling forest area in Yan'an City is a key distribution area of the North China leopard, which is a national first-class protected wild animal. The crested ibis, which has very strict requirements for its habitat environment, was discovered near Dujiabian Village of Yan'an City.

2.3 Sedimentation retention

The impact of check dams on sediment retention and erosion reduction is a significant factor contributing to the recent decrease in sediment entering the Yellow River. The measured annual average sediment discharge from 2000 to 2012 at the Tongguan Hydrological Station on the Yellow River was 276 million tons, which was 1.316 billion tons less than that from 1919 to 1959. Preliminary studies indicated that check dams installed above the Tongguan Hydrological Station trapped sediment and reduced erosion by an annual average of 450 million tons, accounting for a 34% reduction in the sediment discharge in the Yellow River from 2000 to 2012 (Li & Liu, 2018). Ran et al. (2008) concluded that the sediment trapped by check dams accounted for 64.7% of the total sediment reduction due to all soil and water conservation measures in the Hekou–Longmen reach from 1970 to 1996. Wang et al. (2016) demonstrated that check dams and reservoirs accounted for 21% and 12% of the total sediment reduction in the Hekou–Tongguan reach (i.e., the middle reach of the Yellow River) between 1951 and 1999 and 2000–2010, respectively. Liu et al. (2018) compared the sediment trapped by check dams above the Tongguan section in the Hekou–Longmen reach and the upper reach of the Beiluo River from 1950 to 2014, as shown in Figure 6. The sediment retention of check dams peaked during the period from 1970 to 1989 and declined thereafter, which might be partly attributed to the saturation of the sediment storage capacity of some check dams and the remarkable sediment reduction by vegetation restoration with the “Grain for Green” Program since 1999 (Gao et al., 2024; Wang et al., 2016; Wang et al., 2017a; Zhang & She, 2021). Approximately 66%–80% of the total trapped sediments above the Tongguan section were retained in the Hekouzhen–Longmen reach and the upper reach of the Beiluo River, where the number of large, medium and small check dams accounted for 70%, 80%, and 91%, respectively, of all check dams constructed above the Tongguan section (Liu et al., 2018).

Table 4 compares the effectiveness of check dams in retaining sediments in two small watersheds during a typical extreme rainfall event: one is the Jiuyuangou watershed, which was harnessed by check dams, whereas the other is the Peijiamao watershed, which was basically untamed (Zhao et al., 2019). There are 241 check dams in the Jiuyuangou watershed, including 28 large, 42 medium, and 171 small dams. Their total storage capacity is 42.37 million m³, and their sediment retention capacity is 34.15 million m³, corresponding to a dam land development of 4.33 square kilometers. In contrast, the Peijiamao watershed is underdeveloped in terms of check dam construction. On July 26, 2017, an extreme rainstorm hit the northern Shaanxi Province, creating a catastrophic flood with a return period of roughly 100 years in this region. With roughly the same rainfall amount in these two watersheds, the maximum sediment concentration in Jiuyuangou was 170 kg/m³, compared to 382 kg/m³ in Peijiamao. The sediment transport modulus in Jiuyuangou was 75.8% lower than that in Peijiamao without soil or water management.

3.1 Flood incidents on check dams

As small-scale earth structures are mainly built by local peasants, check dams normally have low construction standards, and incidents caused by floods have been frequently reported (e.g., Guo et al., 2021; Liu et al., 2020; Yang et al., 2020).

3.1.2 Dam failure due to inadequate flood release facilities

Check dams, especially those constructed by self-organized local peasants before the 1980s, do not have spillways, which makes the flood control system very fragile against extreme weather conditions. Figure 7 shows a recent case of a check dam breach. The Qingshuigou Reservoir, completed in 2010, which supplies water to Zizhou County in Shaanxi Province, had a total storage and flood detention capacities of 370,000 and 280,000 m³, respectively. When heavy rainfall with a maximum 1-h rainfall of 52 mm happened on July 26, 2017, the Qingshuigou Dam overtopped and breached a height of 25 m (Zhang et al., 2019; Zhang, 2017). The lack of a spillway for this dam, which was used for normal water storage and supplied water to the nearby county, is the main reason for the dam's failure.

3.1.3 Dam failure due to breaking of the underlying conduit

Figure 8 shows a typical water release conduit, which is normally placed under the body of the dam. Since it is designed to work in open-channel conditions, hazards frequently occur owing to the failure to allow such working conditions by accidental inflow into this system, resulting in piping and dam breaches (Figure 9).

3.2 Flood risk assessment and mitigation in check dam watersheds

3.2.1 Quantifying flood risks for a clustered check dam system

The number of check dams on the Loess Plateau is so large that it is impractical to reinforce each dam to meet the requirements of the safety control standards (Guo et al., 2021). For a watershed that involves a group of check dams without adequate countermeasures against failure, an important approach for mitigating possible hazards is to develop a dam safety administration system that can identify flood-affected regions and issue an early warning. Risk evaluation of flood hazards due to overtopping by heavy rainfall, along with computer software, can contribute to this system.

Figure 10 shows the Wangmaogou check dam system, which exhibits complex series and parallel connections of 22 dams. When conducting a rainfall flood routing for each dam, it is necessary to consider the inflow released from the upstream tributary dams due to dam break or spillway discharge. More recently, this complex problem was successfully solved by proposing a refined analysis method to determine the critical rainfall threshold of a dam system (Chen et al., 2021; Zhang & Chen, 2019). It is observed that a check dam system has characteristics similar to human genealogy, that is, there is a backbone dam that controls the runoff of the entire watershed at the downstream terminal of the dam system, referred to as the “ancestor dam,” such as Wangmaogou 1# dam. From this dam, we can trace back upstream and find the “next generation” of dams. In this way, by tracing generation after generation towards upstream, the last batch of “boundary dams” without “descendants” is reached. Following this philosophy, a computer program was developed to identify whether check dams were overtopped or breached by floods under a given amount of rainfall. This is a refined algorithm that fully considers the possible formation of inflows owing to dam breakage and discharge in the upstream dam. We call this the “family tree method” (Chen et al., 2021).

Figure 11 shows the risk analysis results of 22 dams under a maximum 12-h rainfall of 179.6 mm (equivalent to a 200-year return period). The red-filled circles indicate the breached dams, and green-filled boxes indicate the surviving dams. It should be noted that the numbers represent the amount of inflow into the dam in units of 1000 m³. Those with and without underscores represent contributions from upstream dams and natural runoff in the watershed, respectively. For example, Dam No. 8 is marked with inflow values of 6.7 and 41.1, which means that the amount of natural runoff and released water from the upstream dams are 6700 and 41,100 m³, respectively.

3.2.2 Easy and quick dam breach flood evaluations and early warning

Over the past decade, the authors have been working on updating the existing dam breach analysis methods (Chen et al., 2015, 2019a, 2019b). The main improvements include (i) a hyperbolic soil erosion model, (ii) an empirical approach to model the lateral enlargement of the breach opening and (iii) a numerical algorithm that adopts velocity increments to allow a straightforward calculation of the breach flood hydrograph. Figure 12 shows the general framework of the updated theoretical methods and numerical algorithms. A computer program, DB-IWHR, has been developed to implement the algorithm, which is a spreadsheet supported by VBA programming. It is free to access and can be downloaded from the website geoeng.iwhr.com/geoeng/download.htm.

Figure 13 shows an example of the breach risk analysis results for the Yangjiagou check dam located in Yulin City, Shaanxi Province. It controls a drainage area of 1.50 km² and has a total storage capacity of 515,000 m³. Currently, the sedimentation has occupied 225,000 m³ of its storage capacity. The dam is 25.0 m high with a crest length of 110.0 m. There are six residential areas downstream, labeled 0 to 5. Figure 14 shows the results of the dam-break flood analysis. It can be seen that the peak discharge of 528.52 m³/s was reached after 0.63 h following the breach of Yangjiagou check dam. Among the six risk points, the six houses of the No. 3 risk point were located at an elevation of 1090.54 m, lower than the corresponding flood elevation of 1094.62 m. The other risk points were located above the calculated flood inundation elevation. The estimated outcomes can serve as a useful reference for local administrations in planning risk assessments and disposal measures.

3.3 Novel water-release facilities

3.3.1 Geobag stepped spillway

Developing a new, reliable, and economical type of spillway is important for mitigating flood risks in check dams. The authors developed a new type of spillway to address this issue (Yu et al., 2019, 2023). Figure 15 shows the geobag stepped spillway built at the Sanlian check dam in Ya'an, Shaanxi Province. The dam has a height of 6 m and the spillway is 3.8 m wide with a slope of 2 on 1 (horizontal on vertical). Ten full-scale overflow tests were performed with gated and uncontrolled weir flows at different water heads on the weir. The energy dissipation ratio was greater than 67% under all circumstances (Figure 16). After the overflow tests, the structure of the geobag stepped spillway was stable with negligible deformation, indicating that this novel spillway was successful and can be used in actual check dam projects. The geobag stepped spillway has been adopted in three check dams on the Loess Plateau as a permanent flood-release facility.

Table 5. Typical check dam breach events on the Loess Plateau during catastrophic floods since the 1980s.

Time	Location	Watershed	Rainfall	Breached or damaged check dams
July 25–26, 2017	Yulin, Shaanxi	Qingshuigou, Majiagou and Shejiagou	Maximum 1-h rainfall of 52 mm and maximum 12-h rainfall of 211 mm	13 check dams were destroyed in Majiagou. The drainage channels of two check dams were damaged in Qingshuigou. Gullies on the dam and damage to the spillway were observed in one check dam in Shejiagou (Wang et al., 2017b).
August 16–18, 2016	Ordos, Inner Mongolia	Xiliugou and Hantaichuan	Maximum 1-h rainfall of over 50 mm and maximum 24-h rainfall of 248.8 mm	19 check dams, including 12 large, five medium and two small dams breached, all of which did not have spillways. The cascade dams of Youfangqu Nos. 1, 2 and 3 and Dawulanstaigou dam collapsed (Wang et al., 2019).
July 2013	Yan'an, Shaanxi	Yanhe River	Five rainfall processes with the frequency exceeding a 100-year return period, with the maximum cumulative point rainfall of 792.9 mm	516 check dams were damaged, most of which had been fully filled with sediment and did not have spillways (Wei et al., 2015).
July 15, 2012	Yulin, Shaanxi;	Jiuyuangou	Maximum 1-h rainfall of 75.7 mm	24 check dams (45 in total) were damaged to different extents (Gao et al., 2017).
September 19, 2010	Lvliang, Shanxi	Dianpinggou, Shizhang and Yangpo	Maximum 2-h rainfall of 185 mm and some areas experiencing 24-h rainfall exceeding a 200-year return period	The total water storage of 71 large and medium check dams reached 14.3 million m3. The culvert collapse and gullies in dams were observed in four check dams in Fangshan County. The Xize dam in Lishi District was breached due to overtopping (Wang et al., 2011b).
July 8, 2000	Lvliang, Shanxi	Wangjiagou	Maximum 30-min rainfall of 32.1 mm with a frequency less than one in 50 y	Four check dams were perforated, and the foundations of one dam spillway and one dam culvert outlet were locally scoured and suspended, with the sediment of 6973 m3 scoured away (Nie et al., 2000).
August 1994	Shaanxi, Shanxi, Inner Mongolia, Gansu and Ningxia Hui	Middle reaches of the Yellow River	4–7 regional heavy rainstorms with rainfall of 50%–60% more than that during the same period in the previous years	7542 check dams (8% of the total amount in the middle reaches of the Yellow River) suffered, to different extents, water-induced damage with a loss of 33 km2 dam cropland and a total investment of 374 million RMB for repairment (Wang & Wang, 1995).
July 22, 1989	Lvliang, Shanxi	Yujiagou and Liuyegou	Maximum 1-h rainfall of 90 mm and maximum 6-h rainfall of 178.8 mm with a 500-year return period	91 dams (205 in total) were destroyed, and the sediment of 743,600 m3 was scoured away in Xing County. The dam cropland with an area of 2505 mu was destroyed, and crops with an area of 1260 mu were flooded (Ma et al., 1990).

3.3.2 Pressurized conduit

Recently, prestressed concrete cylinder pipes (PCCP) have been widely used in civic water supply and drainage projects and large-scale water conveyance projects (e.g., Cheng et al., 2020; Hu, 2009). They can withstand high internal water pressures, thus mitigating the risk of a dam breach. Figure 17 shows an example that adopts PCCP for releasing water from the check dams in Yan'an, Shaanxi Province.

4.1 Actions on further developments

In the early 1950s, Ning Qian, a renowned sediment expert from China, observed that most of the sediment deposited in the main channel of the Yellow River consisted of coarse particles larger than 0.05 mm (Qian et al., 1978). Table 6 summarizes the landforms, hydrology, and land use in the high and coarse sediment yield zone located in the middle reaches of the Yellow River (Li et al., 2021; Yin et al., 2021b).

Table 6. Landforms, hydrology, and land use in the high and coarse sediment yield zone in the middle reaches of the Yellow River.

	Landform: Main landforms are loess tablelands, ridges and hills with dense and deep gullies and steep slopes. The topsoil mainly consists of the erodible loess, which is a loose aeolian deposit of yellowish silt-dominated dust.
	Hydrology: Located in the arid and semi-arid region, the multi-annual average precipitation in the zone is around 450 mm, which decreases from the southeast to the northwest. Greater than 60% of the whole-year precipitation occurs usually in the form of heavy rainstorms in the summer (i.e., from July to September). The Yellow River and over ten tributaries go through. The topsoil (i.e., loess) is eroded and transported into the Yellow River.
	Land use: Main types of land use are grassland, forest, and farmland. The annual average vegetation cover area and fractional vegetation cover (FVC) increased by 10,000 square kilometers and 18.5%, respectively, which partly results in the remarkable reduction in the soil erosion from 1999 to 2020.

This high and coarse sediment yield zone covers an area of 78,600 square kilometers, where the sediment transport modulus and coarse sediment transport modulus exceed 5000 and 1300 tons per square kilometers per year, respectively. The region's landforms, soil characteristics, river systems, and climatic conditions result in the most severe water-driven erosion within the Yellow River's middle reaches.

Specifically, the concentrated coarse sediment yield zone spans an area of 18,800 square kilometers, with a multi-annual average sediment transport of 255 million tons per year. This zone primarily encompasses the watersheds of the Huangfuchuan, Qingshuichuan, Gushanchuan, Shimachuan, Kuye, Tuwei, Jialu, Wulong, Wuding, Qingjian, and Yan Rivers. These tributaries form eleven sub-reaches between Hekou and Longmen on the right bank of the Yellow River.

The coarse sediment concentrated yield zone is the main source of the soil erosion on the Loess Plateau and, consequently, the continuous sediment siltation in the main channel in the lower reaches of the Yellow River (Li et al., 2023; Liu et al., 2023). Therefore, during China's “14th 5-Year Plan,” the construction of check dams in the Yellow River will focus on this region, and more than 3000 check dams will be planned. At present, the first phase of the sediment-retaining project is in full swing, including 2559 small and medium sediment-retaining dams (Chen, 2022; Hui et al., 2024).

4.2 Promotion of the green corridor on the Loess Plateau

Combining check dam construction with terrain reclamation has become a new endeavor to promote soil and water conservation on the Loess Plateau (Chen, 2022). As shown in Figure 18, for example, by pumping water from the Liujiaping reservoir using solar energy, the Zhaojiagua loess terrain in Yulin, Shaanxi Province, has been transformed into 18,000 mu forests and orchards. This pioneering study has attracted extensive attention.

4.3 Upgrading the flood release facilities

To improve the construction quality of the stepped geobag spillway in check dams as described in Section 3.3.1, the approach of the pumped concrete mattress could provide a reliable and economical cast-in-place alternative to the geobags filled with the compacted loess (Yu et al., 2023). Figure 19 shows an example of the pumped concrete mattress construction for channel lining.