Groundwater sapping as the cause of irreversible desertification of Hunshandake Sandy Lands, Inner Mongolia, northern China

Paleolake System and Its Chronology

The spatial organization of Hunshandake dunes and interdune depressions indicates that current stabilized dune ridges with transverse and roughly concentric lunette-type forms are controlled by locations of former shorelines. Shoreline dune deposits from the three largest eastern Hunshandake lake basins (Figs. 2 and 3) and their degree of cementation suggest that they formed during lake recession and desiccation, with aeolian sands burying older, laminate lacustrine sediments. These lacustrine/shoreline facies include dark-colored and strongly cemented fine-grained sediments with abundant diatoms, the most common of which are Pediastrum, Cyclotella, and Stephanodiscus, typical of floating algae in freshwater lakes (SI Appendix, Table S1). These freshwater paleolakes are open systems with connecting channels and well-defined inlets and outlets that reflect flow between the basins during highstands (Fig. 2 and SI Appendix, Fig. S1). Drainage networks derived from a Shuttle Radar Topography Mission (SRTM) digital elevation model indicate that the lakes were recharged by inflows from rivers with headwaters to the south, i.e., with a much larger catchment that produced surface flow during the wetter early-to-middle Holocene (see SI Appendix, Fig. S2).

Optically Stimulated Luminescence (OSL) dated highstands show that the earliest shorelines are early Holocene and formed at the beginning of Inner Mongolia’s Humid Period. A prominent deflation surface at the northern lake exposes lacustrine sands (11.4 ± 0.6 ka) (Fig. 3, section E), suggesting an early initiation of lakes in the region. Lacustrine deposits exposed on the lake’s eastern shore, OSL-dated to between 10.3 ± 0.6 and 9.3 ± 0.4 ka, and lacustrine sands in higher elevations that date to between 5.7 ± 0.3 and 5.2 ± 0.2 ka, indicate an increase in lake level from 1,324 to 1,340 m by 5 ka. Thick gray lacustrine sediments suggest a prolonged and relatively deep-water environment between 9 and 5 ka (see SI Appendix, Fig. S3). Pollen from lake sediments and ¹⁴C dating of preserved tree subfossils indicate the dominance of Betula, Picea, Abies, Pinus, and Quercus. However, during wet conditions in the early and middle Holocene, this area is characterized as temperate steppe environment, dominated by grasslands and trees near lakes and streams (8, 10, 12, 13).

Well-developed dark grassland-type paleosols (mollisols) at the southern edge of the Hunshandake, OSL-dated to between 6.93 ± 0.61 and 4.27 ± 0.38 ka (Fig. 3), also suggest a wetter climate. Lacustrine sands underlying this paleosol indicate an earlier wetland environment followed by soil formation that indicates a rapid transition to dry conditions at ca. 4.2 ka. Because the southern part of the Hunshandake was not impacted by ground water sapping until recently, it returned to green conditions again at ca. 2.8 ka and maintained this state for slightly longer than 1,000 y (Fig. 3, section P). Similarly, paleosols developed during the period between ca. 9.6 and ca. 3 ka in the western part of Hunshandake (6), where groundwater sapping has yet to occur. The total area of desertification since ca. 4.2 ka ago is >20,000 km² based on field observations and associated mapping.

Groundwater Sapping in the Hunshandake

A relict channel from the northern lake indicates that outflow before ca. 4.5 ka was directed 30 km north to megalake Dali (Fig. 2 and SI Appendix, Fig. S1). However, a lack of recessional shorelines, the presence of well-preserved terraces, and tributaries containing relict terraces and meander loops that grade eastward to an ancestral Xilamulun River ∼50 m above the current base level (see SI Appendix, Fig. S4) suggest rapid and possibly catastrophic lake level decline, with a switch to east-flowing surface drainage ca. 4.2 ka. An increase in the coarse sand fraction in a lake basin core to the southeast, originally interpreted as a state shift response to a drying climate (10), also suggests a rapid transition to drier conditions by ca. 4.2 ka.

Narrow V-shaped canyons incised into thick sand and silt deposited by aeolian, fluvial, and lacustrine processes (Fig. 2 and SI Appendix, Fig. S1) indicate that groundwater sapping (14, 15) or seepage erosion (16) by the headcutting Xilamulun River captured the groundwater table and resulted in rapid lake level lowering. This large-scale hydrologic piracy breached multiple drainage divides, permanently diverting water resources into the east-flowing Xilamulun River. The presence of lacustrine sediments east of the lake basins studied suggests that this is an ongoing process, as the Xilamulun River rapidly headcut during the Holocene. Once eastward flow was established the water table dropped ∼30 m to a level where surface flow was unsustainable, and rapid desertification ensued. Continued downcutting further lowered and steepened the water table’s eastern slope, leaving tributaries in our study area, as well as drainages associated with paleolakes to the east, as dry canyons with present drainage from springs at river level and through groundwater sapping (Fig. 2).

Devastating Impact on the Hongshan Culture

The consequences of a rapid climatic/hydrologic state shift on the Hunshandake’s mixed herding/agricultural Neolithic Hongshan and contemporaneous cultures were likely catastrophic. The Hongshan culture, established ∼6,500 cal B.P.—over 3 millennia before the florescence of the Chinese Shang Dynasty—was first recognized 300 km east of our study sites in the river valleys and hills of western Liaoning Province in northeastern China (17). We found previously unreported abundant remnants of Hongshan pottery and microlithic and macrolithic stone tools preserved in early-to-middle Holocene lacustrine deposits and interbedded sandy layers of former shorelines (∼9.0–5.2 ka) (see SI Appendix, Fig. S5 and Table S2) far to the west of the area traditionally assumed to be Hongshan.

The large number and variety of artifacts associated with these shoreline deposits suggest a relatively dense human population dependent on hunting and fishing. Artifacts indicative of post-Hongshan culture are absent at all sites examined until ca. 3.5 ka. This temporal lacuna suggests rapid abandonment and migration at the onset of desertification and indicates that these habitats were suitable only when conditions were exceptionally favorable during lake highstands. This conclusion is supported by recent studies that show a profound lack of artifacts in the region near the current Xilamulun River headwaters beginning ca. 4.2 ka and lasting for over 600 y (18, 19).

The Hongshan, possibly the earliest Chinese kingdom (17), has typically been viewed as a remote culture outside the key areas of early Chinese civilization. However, recent discoveries, including the Goddess Temple, jade artifacts with symmetrical designs, and remnants of sheep bones indicating trade with Mongolian sheepherders, suggest cultural complexity and stability, although the degree of complexity and links to other emergent Neolithic cultures in Mongolia and Central Asia are still under debate (17). In particular, the level of Hongshan jade-working is highly advanced and unique in early China (20). The first dragon-like symbol of Chinese culture is characterized in carved jade as a footless fish-like creature and may have originated with the Hongshan rather than with later Yellow River cultures (17). Other aquatic animals, such as fish, turtles, and ducks, are represented in jade carvings of the Hongshan culture (20). These aquatic-motif artifacts (see SI Appendix, Fig. S6) may indicate a cultural source related to the Hunshandake’s hydrologic excess in the early/middle Holocene.

Discussion and Conclusions

In a global context, a major climatic shift occurred at ca. 4.2 ka, marking the transition from an early and middle Holocene thermal maximum to late-Holocene colder and more unstable climate conditions (21–23). In semiarid regions of the world, this event is commonly associated with evidence of hazardous drought. Although there is no consensus as to whether drought events around 4.2 ka recorded in various parts of the world were indeed synchronous, extraordinary drought appears to have occurred on all Northern Hemisphere continents within a century or two of 4.2 ka (24). Severe drought is well-documented at ca. 4.2 ka in the Middle East and northern Africa (25–27). Drying at around 4.2 ka is also related to abrupt deforestation in the western Mediterranean region (28) and to the reactivation of dune systems on the margins of the Sahara (29). On the Tibetan Plateau, ¹⁰Be exposure ages on shoreline features of Lake Tangra Yumco indicate recession from a highstand of 143 m above the current lake at 4.3 ± 0.3 ka, associated with a weakening monsoon (30). Closer to the Hunshandake Sandy Lands, pollen records from the temperate steppe of northern China indicate a dominance of forests and a maximum in wet conditions between 8 and 4 ka, as well as noticeable decline in tree pollen at ca. 4 ka, associated with large-scale drying (8, 13).

The Hunshandake green/desert switch, similar to that found in the Sahara, occurred in conjunction with regional drying, as monsoonal flow decreased over the region (6, 8). However, unlike the Sahara, where a return to a green state may occur under a scenario with future wetter conditions and increased surface vegetation (31), reestablishment of a green Hunshandake is highly unlikely. We infer that in the Hundshandake, when the groundwater equilibrium state was altered with progressive sapping, the falling water table led to an eventual collapse in system hysteresis. The combination of factors at ca. 4.2 ka, including fluvial degradation by ∼30 m and a similar drastic lowering of the water table, which led to the disappearance of lakes, coupled with a precipitous drop in monsoonal precipitation, resulted in a dominance of aeolian processes. This produced the extreme aridification of the Hunshandake observed since 4.2 ka. We note that increasing precipitation alone will not reverse this hydrologic deficit because the prior system hysteresis has collapsed (3, 32) (Fig. 4). Because the groundwater table appears to be associated with sediments delivered to and stored in the basin on glacial/interglacial timescales, it is possible that current conditions may persist until enhanced erosion during glacial climatic conditions leads to basin filling and reestablishment of a higher groundwater table.

Groundwater data also support our contention of a rapid and catastrophic state shift in the Hunshandake and continual lowering of the groundwater table and drying of the region. The mean tritium content of precipitation at Harbin, the closest International Atomic Energy Agency observation station, is ∼30 TU, whereas tritium in groundwater from the study area wells varies from 2 to 24 TU (see SI Appendix, Table S3). These tritium concentrations indicate that groundwater in the eastern Hunshandake is just 1 to 2 decades old, with a short residence time in response to an eastward sloping groundwater table, reflecting headwater capture that began at ca. 4.2 ka. As headcutting and groundwater sapping expands westward, we expect that the freshwater and brackish lakes and salty pans of the western/central Hunshandake will continue to drain, further depleting near-surface groundwater and intensifying desertification (Fig. 1).

During the last decade the Hunshandake, widely viewed as a severe instance of human-induced desertification, has been the focus of intense rehabilitation efforts (33). However, our finding of an ecosystem tightly linked to irreversible regional geomorphic and hydrologic change suggests that these efforts may have limited success. Consequences of increased headwater erosion and groundwater sapping in this region include continued ecological degradation and desertification. Our evidence suggests that Hongshan culture was devastated by the combined regional green/desert vegetation shift in the mid-Holocene, similar to that found in other monsoonal regions, but that that change was greatly intensified by the sudden north-to-east hydrologic shift ca. 4.2 ka, resulting from groundwater table drawdown by sapping. The Hunshandake’s current population continues to be adversely impacted by that change today.

Methods

Between 2001 and 2014 we investigated landforms and sedimentary sequences throughout the Hunshandake Sandy Lands to determine aeolian and lacustrine late-Pleistocene/Holocene chronology as well as interactions between human activity and environmental change and their effects on the desertification processes. We dug numerous sedimentologic sections to refine our observations and to interpret the genesis of the sediments before collecting sand samples for luminescence dating from key sections. Sediments were identified as being of aeolian or fluvial/lacustrine origin based on bedding, cementation, color, and texture.

Luminescence dating was carried out using the OSL signal from extracted quartz. Samples for OSL dating were processed under subdued red light in both the Xian Luminescence Dating Laboratory of the Institute of Earth Environment, Chinese Academy of Sciences, China, and the Luminescence Dating Research Laboratory, University of Illinois at Chicago. All OSL measurements in Xian were performed using a Daybreak 2200 automated OSL reader equipped with a combined blue (470 ± 5 nm) and infrared (880 ± 80 nm) LED OSL unit and a ⁹⁰Sr/⁹⁰Y beta source (0.107 Gy/s) for irradiations. All luminescence measurements were made at 125 °C for 200 s, with both IR and blue stimulation powers being nominally ∼45 mW/cm². Luminescence emissions were detected by an EMI 9235QA photomultiplier tube with two 3-mm-thick U-340 glass filters. The signal used for equivalent dose (D_e) determination was the integral for the first 5 s of the OSL decay curve. An Automated Risø TL/OSL-DA-15 system was used in Chicago. Blue-light excitation (470 ± 20 nm) was from an array of 30 light-emitting diodes that delivers ∼15 mW/cm² to the sample position at 90% power. A Thorn EMI 9235 QA photomultiplier tube, coupled with three 3-mm-thick Hoya U-340 detection filters transmitting between 290 and 370 nm, measured photon emissions. Laboratory irradiations used a calibrated ⁹⁰Sr/⁹⁰Y beta source coupled with the Risø reader. The signals used for D_e determination were integrated over the first 0.8 s of stimulation out of 40 s of measurement, with background based on emissions for the last 30- to 40-s interval.

D_e value is determined by Single-aliquot Regenerative-dose (SAR) protocol (34) in both the Xian and Chicago laboratories. To identify diatoms in the lacustrine sediments, general and regional diatom floras as well as genera-specific references were consulted for identification. Digital photographs of all identified taxa were taken during counting to aid taxonomic consistency. The content of tritium was investigated to understand the residual time of the groundwater, and the analytical method of tritium followed Chinese National Standards (35). Geographic information system-based paleogeographical modeling of the megalake region used topographic maps, SRTM/Advanced Spaceborne Thermal Emission and Reflection Radiometer elevation data and Landsat satellite imagery were analyzed using spatial analysis tools in ARCMAP 10.1. Identification of Hongshan artifacts was based on shape, color, decoration, and materials.

For further details of laboratory procedures and data analysis, please see SI Appendix, Additional Details of Methods.

Supplementary Material