Introduction

The strength and duration of erosive winds determine the deformation and migration of dunes1,2. However, whilst the impact of time-averaging methods on reducing the estimated sand transported by wind has been extensively studied3,4, there has been limited attention paid to capturing long-term wind data with shorter averaging time windows to enhance the prediction of dune migration. Research on wind-blown sand two-phase flow often involves averaging the wind speed functions to “simplify” turbulent fluctuations that are otherwise problematic because of the complexity of the flow field5. For example, studies have shown that sand flux is exponentially related to the fluid threshold friction velocity \({u}_{* }\) (usually considered to be a power of 2 or 3)1,6,7,8,9,10,11,12. However, the functional relationships rely on the average wind speed, which can impact the accuracy of estimates for both the actual wind speed, and the transport and deposition of wind-blown sand3,4. This is because, under the same conditions, there is a notable numerical difference between \({\left({\sum }_{1}^{n}{u}_{i}/n\right)}^{3}\) and \(\left({\sum }_{1}^{n}{{u}_{i}}^{3}\right)/n\), where \({u}_{i}\) represents the instantaneous wind speed and n represents the sample size. During extreme weather in particular, such as sandstorms, the wind-blown sand movement near the surface is more intense, and the deviation between the results calculated by the average wind speed using different time scales and the actual situation will be greater4,13,14. Therefore, capturing turbulent motion and sudden short-term intense storms in as much detail as possible is a necessary task for the accurate calculation of sand flux.

Although extreme wind events occur infrequently and for short durations, they remain significant factors of influence in the formation and evolution of aeolian landforms15. For example, the sudden disappearance or collapse of barchan dunes1,16, as well as the formation and evolution of coastal dunes17,18, shadow dunes19, mega ripples20, surface wave propagation and small dune derivation21, and some tall dunes and other aeolian landforms22, are closely related to strong-wind events. Field observations show that the short-term evolution of sand dunes often occurs during severe storms, resulting in steepening of the windward slope, more erosion of sand dunes, and even collapse17,18,23. Gao et al.24 analyzed the morphodynamics and movement of reversing dunes on the southern edge of the Taklamakan Desert, China, and found that, in areas of frequent wind reversals, the impact of extreme wind events blowing to the east, and westward wind that occurs during the daily cycle, cause the dunes to migrate against the transport direction of wind-blown sand particles. Bauer et al.25 observed that, during their three-hour measurement period, flurries accounted for 37.6% of the transport but only occurred for 7.6% of the time; whereas, for a particularly high-energy interval duration of 15 min, flurries accounted for 66.8% of the transport for 42.4% of the time. These findings suggest that strong-wind events may play a critical role in the overall transport regime.

A fundamental question is, what are the shaping patterns of typical aeolian landforms when acted upon by intense storms? Thus far, research on extreme wind events has tended to be limited to single events, such as near-surface turbulence properties26, structural characteristics of wind-blown sand27,28, dust emissions29, risk assessment30,31, and sand and dust storm warnings and assessment32,33,34. However, to the best of our knowledge, there is no systematic analysis in the literature of the relationship between different levels of strong winds and varying degrees of change in the morphology of dunes.

Results

Characterization of wind regime and sand flux

Taitema playa, located on the southeast edge of the Taklamakan Sand Sea, northwest China, experiences consistent control by strong northeastward winds throughout the year (Supplementary Fig. 1a–c). Figure 1a depicts wind rose and the distributions of sand-flux directions. Due to limited sediment supply and unidirectional flows, numerous barchan dunes have formed and migrated at high speed. These highly favorable conditions in the field make it a perfect region for our study. Consequently, we deployed instruments here to measure high-frequency time series of saltation and wind, as well as the morphological evolution of dunes (Fig. 1b; Supplementary Fig. 1d).

Fig. 1: Regional setting.
figure 1

a Location of the study site within Taitema playa, Taklimakan Desert (39.46°N, 88.44°E). The white square represents the dune deformation monitoring area, and the yellow dot indicates the location of the monitoring instruments. b Arrangement of wind and saltation monitoring instruments. The 2-D ultrasonic anemometers are positioned at 0.10, 0.58, 0.78, 1.08, 1.28, 1.78, 2.28, and 3.28 m above ground level, with a measurement frequency of 0.5 Hz. The sand flux intensities were measured by three impact sensors with a sampling frequency of 10 min. The aeolian sediment transport profile was monitored by BSNE sand traps, which feature a vertical sampler opening of 0.02 m (width) × 0.45 m (height) and are situated at heights of 0.22, 0.42, 0.62, 0.92, 1.12, 1.62, 2.12, and 3.12 m. c Grain size distribution curves at different heights of sand traps collected during three strong-wind events. Solid lines are the mean sizes. d Sand flux profiles for 10 monitoring periods calculated from the actual sediment weight collected by BSNE. e Real-time monitoring of wind speed at 3.28 m above ground level and sand flux using SandFlow sensors (SF4, with a sampling interval of 10 min).

Full size image

Figure 1c illustrates grain size distribution curves of sand collected by Big Springs Number Eight (BSNE) saltation traps, showing a decrease in size with the monitoring height. Simultaneously, the average grain sizes vary during different monitoring periods. For P2, P6, and P9, the average grain sizes of sand collected by 8 different height traps were 91.28, 86.59, and 88.81 μm, respectively, reflecting the influence of varying wind conditions (refer to Fig. 2b(i)–(iii)) during these periods. However, these differences are not considered significant due to the inherently fine sediment on the ground surface (with grain size of 107 μm) of Taitema Playa. The sand flux profile in Fig. 1d was calculated using actual data collected by BSNE during the monitoring periods (P1–P10) (see ‘Methods’). Figure 1e provides a comparison of wind speed and sand flux based on field observations. The fluctuations in wind speed and sand flux represent the occurrence of frequent strong-wind events in the study area. According to our statistics, wind speeds exceeding the threshold shear velocity (u*t) contributed approximately 51.55% to the total time. The maximum instantaneous (0.5 Hz) wind speed recorded was 23.07 m s−1 at a height of 3.28 m, and the accumulated time of strong wind events, as defined in ‘Methods’, accounted for 6.87% of the total monitoring time. It is worth noting that the sand flux in Fig. 1e was monitored using sand sensors, which differs from the data source in Fig. 1d.

Fig. 2: The evolution of barchans is closely linked to high wind speeds.
figure 2

a(i)–(iii) Migration of barchans during specific time periods (P2, P6 and P10). The solid white lines in the images represent the outlines of the dunes at the beginning of each period, while the UAV images captured the dunes and their migration at the end of each period. b(i)–(iii) The wind speed fluctuations during three monitoring periods, with a frequency of 1 min, were calculated using the moving average method using 0.5 Hz sampled wind data. c(i)–(iii) Significant linear correlation between the Q/H and D of dunes (P < 0.1), with correlations of 0.90, 0.81, and 0.78, respectively. Here, Q represents the cumulative sand flux predicted using wind speed, which has been converted to a standard meteorological station height of 10 m using the law of the wall.

Full size image

Wind fluctuations and dune migration

We acquired 10 sets of high-precision morphology data using a UAV (see ‘Methods’) and quantified the morphological evolution of barchans. Our primary focus was on the migration distance (D) and height (H) variation for each period (Fig. 2a) (Supplementary Table 2) (Supplementary Data 1). We closely monitored high-frequency wind speed fluctuations (Fig. 2b) and calculated the cumulative sand flux (Q) for each period based on sliced wind-speed data. Figure 2c illustrates the significant linear correlation between the ratio of cumulative sand flux to dune height (Q/H) and migration distance (D) during corresponding time periods. All presentations and analyses in Fig. 2 are based on periods P2, P6, and P10. Our focus was specifically on typical barchans, excluding compound dunes and seif dunes.

Different wind speed levels drive different dune migration distances

We obtained a well-simulated liner relationship of cumulative sand flux by comparing low-frequency measurements with BSNE (Qa) to high-frequency model calculations (Q) over the time scales of monitoring periods (Fig. 3a). When the data of P8 were excluded (yellow dot in Fig. 3a), the significant correlation was higher (R2 = 0.83, at a significance level of P < 0.01). When P8 data were included, the correlation decreased (R2 = 0.61, at a significance level of P < 0.05). This difference is caused by the longer time intervals of P8 (81 days), suggesting that not only the strong winds quickly had filled the sand traps before collected and reduced the efficiency of BSNE (see ‘Discussion’), but also fully illustrates the importance of short-term strong wind events in evaluating the dune migration rate. Here, the amount of sand collected by the near-surface traps is less than the actual sand transport in P8, and there are sufficient reasons to consider it as an outlier and eliminate it. Additionally, the difference between Qa and Q values is partly due to the failure to capture near-surface sand transport caused by the height of sand traps.

Fig. 3: Quantifying the barchan migration response to varying levels of wind-blown sand.
figure 3

a High correlation of sand flux by comparing low-frequency measurements with BSNE traps (Qa) to high-frequency model calculations (Q). b Significant linear correlation between Q/H and D. c The contribution rate of different levels of wind speeds to the migration distance during different periods.

Full size image

Based on the saltation mass flux relation (see ‘Methods’), we employed the high-frequency wind speed data to predict the sand transport rate without long-term averaging. We established a strong linear relationship between Q/H and the migration distances (D) of all monitored barchans (Fig. 3b). Here, Q represents the cumulative sand transport during each observation period, and D signifies the distance migrated by each dune during that observation period. The establishment of this linear relationship aligns with Bagnold’s theory, which states that the migration rate of barchan dunes can be expressed as C = Q/γs H, where C is the dune migration rate, Q represents the sand transport rate at the barchan dune crest, γs is the bulk density of sand particles, and H is the height of dunes. Furthermore, extensive research has shown that mature barchan dunes typically maintain their height while moving forward. Therefore, it is reasonable to use the mean of the starting and ending heights of each monitoring period for H in this linear relationship, and the high-precision data from DEMs ensures the accuracy of H.

Different monitoring periods (P1–P10) encompass distinct combinations of wind speeds. Building upon the significant linear correlation between Q/H and D that persists under these varying conditions (Fig. 3b), we aim to address a fundamental question: what distance can a barchan dune migrate under different wind speeds? Our perspective is that, in each period, the sand flux generated by wind speed u ≥ ut contributes 100% of the barchan’s migration distance. For a specific height (H) of each dune, it can be reasonably assumed that if u ≥ 10 m s−1 accounts for 80% of the sand flux generated by u ≥ ut, it means that u ≥ 10 m s−1 also contributes 80% to the barchan’s migration distance.

Based on the aforementioned research results, we calculated the ratios of the cumulative sand flux Qi, generated by wind speeds of different levels ui (e.g., ui ≥ ut, 6, 8, 10 m s−1), to the total sand flux (Q) during the period. These ratios are equivalent to the contribution rates of barchan migration caused by different levels of winds to the total distance during that period. As depicted in Fig. 3c, colored scatter points represent various periods. Despite variations in wind conditions during each period, especially in terms of the interval and duration of strong-wind events, winds exceeding 10 m s−1 contributed an average of 75% to dune movement (as indicated by the red curve) throughout the entire monitoring period. Simultaneously, it is noticeable that the duration of high winds also influenced the evolution of sand dunes. As demonstrated by periods P6 and P10, wind speeds exceeding 10 m s−1 contributed 65% and 97% to the dune migration distance, respectively (Fig. 3c). This highlights substantial variations in wind-driven dune motion between these two periods, which can also be visually confirmed in Figs. 2b(ii), (iii).

Verification of global typical barchan movement

Barchan dunes can be found in various locations on Earth. We selected three representative barchan fields to investigate the impact of strong-wind events on dune migration (L1, L2, and L3 in Fig. 4a; Supplementary Tables 3–5). Each dune field involves over 20 typical barchan dunes with different sizes (Fig. 4b–d). This analysis was performed using high-resolution remote sensing images and ERA5 reanalysis wind data (the fifth major global reanalysis produced by the European Centre for Medium-Range Weather Forecasts). It was found that, despite variations in the frequency and duration of strong-wind events across the three regions, all of them exhibit a strong linear correlation between Q/H (cumulative sand flux normalized by barchan dune height) and D (migration distance) of the barchan dunes (Fig. 4e). The results reaffirmed that the rate at which strong-wind events contribute to barchan movement is closely related to the ratio of sand flux generated by strong winds to the total sand flux during the monitoring period (Fig. 4f, g) (Results for L2 and L3 are shown in the Supplementary Fig. 6).

Fig. 4: Quantifying the contribution of strong winds to the dune movement of the global barchan sand sea.
figure 4

a Locations of typical barchan fields (L1–L3). b–d Dune morphology, wind roses, and the distributions of sand flux direction for L1, L2, and L3. e Significant linear correlation between the ratio of cumulative sand flux to dune height (Q/H) and the measured dune migration distance (D). f The contribution rate of different levels of wind speed to the migration distance of barchans during different periods for L1. g The migration distance caused by different dunes with different wind speeds in a specific period for L1.

Full size image

Taking L1 as an example, Fig. 4f shows that, for the same barchan dune, based on the linear relationship between sand flux and migration distance, we can conclude that, if the total sand flux is generated by wind speeds exceeding the threshold velocity ut (5.23 m s−1), then the wind contribution above that threshold accounts for 100% of the barchan’s movement. Similarly, over ten periods, sand flux generated by winds exceeding 10 m s−1 accounts for 32% of the total sand flux. Therefore, we believe that winds above 10 m s−1 contribute an average of 32% to the barchan’s migration distance (as indicated by the red line in Fig. 4f). However, the contribution rate of strong winds can vary depending on the wind energy during the specific period. For instance, during P7, due to more frequent strong-wind events, winds above 10 m s−1 contributed to 37% of dune movement. In this context, we can consider lowering the criteria for defining strong-wind events, such as considering the contribution rates of winds above 9 m s−1 and 8 m s−1 based on low-sampling-frequency wind data, which would result in dune movement contributions reaching 48% and 67%, respectively (Fig. 4f). By doing so, we can further investigate the error rates caused by low-sampling-frequency wind data when quantifying barchan movement in future studies. Lastly, although the total migration distance of different sand dunes varies, taking into account the specific height (H) of each dune, the calculation method of the contribution rate of different levels of strong winds to the dune movement distance remains unchanged (Fig. 4g).

Discussion

Leaving aside the details of sand transportation and fluid dynamics, we often describe and quantify the deformation and migration of dunes from a phenomenological perspective. Field and imagery data have become valuable sources of data to monitor changes in barchan morphology over time. The observation period, T, primarily depends on the available remote sensing image or field measurement time intervals, which becomes the denominator for calculating the barchan dune migration rate35,36,37 or generating interpolated migration rates for each dune. Taking the study of ref. 38 as an example, multiple field topographic surveys were conducted at irregular intervals, T (ranging from 24 to 167 days), depending on the season and wind activity, to generate monthly interpolated migration rates. The results were based on the assumption that the daily rate is constant throughout the period. This interpolation method is widely employed in current dune migration rate calculations39,40,41, including the assessment of annual migration rates from a single topographic survey conducted over several years or one year42,43,44,45. However, this interpolation inevitably overlooks rapid dune migration driven by outbreaks of wind events lasting less than the observation period T, hinting that assessments of dune mobility under climate change from the models with average migration rates must be taken with reservations46.

It is well known that the higher the resolution of topographic data, the more accurate the estimation of the dune migration rate will be. Obviously, it is almost impossible to observe wind and topographic data at the same frequency with current monitoring technology. The time resolution of acquired topographic data is often much lower than that of wind data, making it challenging to accurately quantify the relationship between wind and dune migration. Even if the observation T is shortened as much as possible, such as to one year, a quarter, or a month37,47, the high contribution rate of strong-wind events with low probability and high energy to dune migration is inevitably ignored. This includes rare events, like an extremely large sandstorm occurring once a year and lasting for 10 h. While there are assessments of highly active dune evolution and sand flux during a single sandstorm in the very short term15,48, much of the research is limited to the short-term evolution of coastal dunes. Moreover, the resolution of existing remote sensing data is insufficient to sample both large dunes that do not propagate much over a year, and small ones that cannot be followed over long periods of time21.

This study utilized high-frequency sampled wind and intensively monitored topographic data to establish a significant correlation between sand transport and the distance of barchan migration. Building upon this foundation, we placed a greater emphasis on the crucial contribution of short-term strong-wind events to the formation and evolution of dunes. The paper addresses this challenge by amalgamating multiple wind events within the study area and utilizing Q as a key indicator to intuitively reveal the distance barchan dunes migrate under different levels of wind speed. In essence, we captured multiple strong-wind events which responsible for sand transportation, enabling more accurate predictions of sand flux and contribution estimates of migration distance without smoothing wind speed and interpolating migration rates.

Additionally, there are two other reasons why we chose not to utilize the measured BSNE data. Firstly, the study area is prone to frequent strong winds, and the efficiency of BSNE decreases with increasing wind speed49. The small particle size of the sand source leads to the generation of eddies at the sampler entrance due to strong winds, which, in turn, diminishes the capture of sand particles. Secondly, by directly correlating high-precision wind speed data with well-established sand flux models, we can achieve a more accurate quantification of sand dune movement. This approach will provide more effective and scientifically guided solutions for wind and sand control engineering.

This study provides a more precise answer to the scientific question of ‘what distances do barchans move at varying levels of wind?’ In doing so, it also holds significant implications for a more scientifically grounded explanation of the mechanisms underlying the impact of wind on the evolution of barchan dunes.

Methods

In situ measurements of wind and saltation

We deployed instruments (39.46°N, 88.44°E) in the study area to record high-frequency time series of wind and saltation (Fig. 1b; Supplementary Fig. 1d). The wind profile was measured using eight 2-D ultrasonic anemometers (Wind Sonic) placed at heights of 0.10, 0.58, 0.78, 1.08, 1.28, 1.78, 2.28, and 3.28 m above ground level, capturing wind data at a frequency of 0.5 Hz via a datalogger (Campbell CR6). Sand flux intensity was measured using three impact sensors (Sand Flow SF4) capable of detecting saltating grains larger than 0.15 mm, collecting data at 10-min intervals with a datalogger (Campbell CR 1000X). The aeolian sand flux profile over the playa surface was determined using a vertical array of Big Spring Number Eight (BSNE) sand traps, each equipped with a sampler opening of 0.02 m (width) × 0.45 m (height), positioned at heights of 0.22, 0.42, 0.62, 0.92, 1.12, 1.62, 2.12, and 3.12 m. In practice, we excluded the data from periods when the sand traps was filled due to very strong wind events or long-time wind action. Additionally, to ensure data validity, we physically visited the study area multiple times during each monitoring period to collect sand from the traps.

Saturated flux on a flat sand bed

Wind data collected in the Taitema playa were used to predict the sand flux (Fig. 1b). Table 1 displays a set of relevant variables and results obtained using the following formulae. Wind data of three representative dune fields were obtained from the ERA5 reanalysis. Here, wind data include the speed \({u}_{i}\) and direction \({\theta }_{i}\) at height z and different times \({t}_{i}\), i ∈ [1; N]. The shear velocity can be extrapolated by using the law of the wall,

$${u}_{\ast }^{i}=\frac{{u}_{i}\kappa }{\log (z/{z}_{s})},$$
(1)

where z is the height at which the wind data are measured. Using the formula calibrated by Iversen and Rasmussen50, we can evaluate the threshold shear velocity for motion inception as follows:

$${u}_{* t}\, = \, 0.1\sqrt{\frac{{\rho}_{{{\mbox{s}}}}}{{\rho}_{{{\mbox{f}}}}}{{\mbox{gd}}}}.$$
(2)
Table 1 Aeolian dynamic parameters
Full size table

By substituting formula (2) into formula (1), the threshold shear velocity at any height z can be obtained. Therefore, the threshold wind velocity (Table 1) at the standard height of 10 m for the meteorological station in the study area can be calculated.

The saturated sand flux on the playa was computed from the relationship proposed by Lettau and Lettau8:

$${{Q}}_{{{{{{\rm{flat}}}}}}}({u}_{\ast })=\Bigg\{\begin{array}{cc}C\sqrt{\frac{d}{{D}_{g}}}\frac{{\rho }_{f}}{g}({u}_{\ast }-{u}_{\ast t}){u}_{\ast }^{2} & {{{{{\rm{for}}}}}}\,{u}_{\ast } > {{u}}_{\ast t},\\ 0 & {{{{{\rm{else}}}}}}.\end{array}$$
(3)

According to the methods, we obtained aeolian dynamic parameters for each period in the study area (Table 2).

Table 2 Sand flux for each period in the Taitema playa
Full size table

Definition of strong wind events in this study

For the 169 wind days from April 24, 2021, to October 10, 2021, 57.4% of recorded wind speeds were above the threshold speed (10 m height calculated by the law of the wall). In this study, we employed the universal concept of strong winds, which refers to winds capable of causing significant erosion and extreme events, such as sandstorms, in arid and semi-arid regions. It is worth noting that this definition differs from meteorological classifications like Gale (17.2–20.7 m s−1) or Strong Gale (20.8–24.4 m s−1). Our criteria for defining a strong wind event were as follows: wind speeds at a height of 10 m equal to or exceeding 10 m s−1; a duration of at least 30 min; and a minimum interval of 5 h between consecutive strong wind events. This definition allowed us to determine the duration of strong winds for each period (Table 2). We excluded short-term winds that lacked the potential for sediment transport and accounted for the impact of wind speed fluctuations.

Extraction of barchan morphometrics

We selected an entire dune field, containing barchans of different sizes and stages of evolution, within 1.4 × 1 km of the study area (Fig. 1a). We acquired high-resolution (7.9 cm/px) digital elevation models (DEMs) and orthophoto images, 10 times, using an unmanned aerial vehicle (UAV, DJI Phantom 4) before and after the transit of strong-wind events (i.e., sandstorm).

The monitoring periods (P1–P10) are outlined in Table 2 and depicted in Fig. 1e. Wind data were time-sliced according to field operations. Throughout these periods, we meticulously documented the birth, growth, migration, disappearance, propagation, and further evolution of the dunes’ structure (Supplementary Figs. 2–4; Supplementary Data 1). We maintained uniform flight parameters (Supplementary Table 1) for each measurement and analyzed the point cloud data acquired by the UAV using ArcGIS software. This allowed us to quantify the 3-D morphological parameters of the dunes (Supplementary Fig. 3; Supplementary Data 1). Finally, we calculated the successive migration distance and other morphological changes for each dune on the Taitema playa. The morphometric data for three other representative sand seas (Fig. 4b–d) were from satellite imagery (Table 3). Dune height was estimated by multiplying the horizontal length of the slipface (measured from the brink to the base of the slipface, in map view) by the tangent of 34°, representing the angle of repose for aeolian sand.

Table 3 Migration monitoring of barchan dunes around the world
Full size table

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.