Playas are common in arid environments and can be major sources of mineral dust that can influence global climate. These landforms typically form crusts that limit evaporation and dust emission, modify surface erosivity and erodibility, and can lead to over prediction or underprediction of (1) dust-emission potential and (2) water and heat fluxes in energy-balance modeling. Through terrestrial laser scanning measurements of part of the Makgadikgadi Pans of Botswana (a Southern Hemisphere playa that emits significant amounts of dust), we show that over weeks, months, and a year, the shapes of these surfaces change considerably (ridge thrusting of >30 mm/week) and can switch among continuous, ridged, and degraded patterns. Ridged pattern development changes the measured aerodynamic roughness of the surface (as much as 3 mm/week). The dynamic nature of these crusted surfaces must be accounted for in dust entrainment and moisture balance formulae to improve regional and global climate models.
Playas are common landforms in arid regions and are one of the biggest sources of mineral dust emission (Washington et al., 2003; Prospero et al., 2002). They contribute to surface energy balances by modifying moisture and heat fluxes in the atmosphere (Bryant and Rainey, 2002). There is uncertainty as to how much playas contribute to dust emissions (Sweeney et al., 2011) and the extent to which they influence the hydrological budget, because when the water table is shallow, a protective and often heterogeneous mineral crust is typically formed (Reynolds et al., 2007). These crusts can form polygonal patterns that may be smooth and continuous, limiting subsurface exposure, or ridged and cracked, enabling unregulated exchange between surface and subsurface processes.
Uncertainty in dust emission predictions (flux and location) stems from three main controls on crust patterns. (1) Crusts influence the availability of surface material, which varies depending on crust moisture, type, and thickness (Nickling, 1984). (2) The salt chemistry and changes in moisture may result in the formation of weak, low-bulk-density sediment (mineral grains plus salt crystal; fluff), thereby increasing the supply of potentially emissive material (Reynolds et al., 2007), particularly on sulfate-rich playas (Buck et al., 2011). (3) As polygon ridges grow and degrade over time, they alter surface roughness and therefore the erodibility of the surface and erosivity of the wind through modifications to aerodynamic roughness (zo), a key parameter in modeling dust-emission potential (Raupach et al., 1993; Marticorena and Bergametti, 1995).
This paper assesses surface moisture, aerodynamic roughness, and crust pattern change on a large playa (Sua Pan) in Botswana over weeks, months, and one year. We first introduce the surface variability, and then use Fourier transforms to assess how surface patterns develop over time, ultimately linking surface change to atmospheric and subsurface controls through a conceptual model that identifies trajectories in crust patterning.
SITE LOCATION AND METHODS
Investigations were undertaken during August 2011 and August and September 2012 at 6 locations (sites L5, J11, D10, G6, I4, B7) on Sua Pan within a study area of 144 km2 centered at 20.5754°S, 25.959°E. Sua Pan is part of the 3400 km2 Makgadikgadi complex in Botswana. This playa undergoes ephemeral flooding and its surface consists of a polygonal salt crust that is predominantly halite, thenardite, and trona (Eckardt et al., 2008). The surface includes freshly formed crust with no or low topographic perturbations (continuous), well-formed polygon ridges (ridged), older broken, deflated ridges (degraded), or a mix of these three surfaces.
The surface elevations of the crust were measured using a Leica Scanstation terrestrial laser scanner (TLS) and converted to 100 m2 digital elevation models of 1-cm-grid resolution following the methods of Nield et al. (2013). Registration errors between monthly and weekly scans were <2 mm. Locations between years were registered using a differential GPS survey. Ridge height, width, and spacing were calculated by linearly detrending 1-cm-spaced transects and identifying positive height deviations (Nield et al., 2013). Two-dimensional Fourier transforms (following the method of Perron et al., 2008) were used to calculate radial frequencies and corresponding ridge spacings, which were normalized using the flat surface spectra (site L5 in 2011). Water-table depth was measured using a dipwell in 2 m wells. Relative surface moisture was calculated at each site using the mean intensity of the TLS return signal collected within a 2 m2 area at a distance of 12.5 m, normalized by the wettest site (site L5 in 2011), following the methods of Nield et al. (2014). High values indicate drier surfaces.
Detection of surface moisture on the entire playa between 2000 and 2012 was based on normalized difference water index (NDWI) data extracted from Moderate Resolution Imaging Spectroradiometer (MODIS) MOD/MYD09A1 8 day data (500 m resolution) (Gao, 1996; Leon and Cohen, 2012). NDWI threshold values were validated using field observations of surface condition (i.e., flooded, moist, and dry) from various basin locations along with high-resolution remote sensing data (e.g., Advanced Spaceborne Thermal Emission and Reflection Radiometer, ASTER; Landsat). Processed data were consolidated to provide long-term and annual surface flood frequency data (Ff ; Figs. 1A–1C), and, for the 2010–2012 hydrological years, 8 day NDWI data were extracted for each study site (Fig. 1D).
Values of zo were calculated using 2 weeks of easterly wind speeds measured at 0.25 m, 0.47 m, 0.89 m, and 1.68 m with Vector Instruments cup anemometers (A-100R), averaged over 1 min intervals following standard law of the wall profile methods. Measurements below 3 m s–1 or with R2 values below 0.98 were discarded (Bauer et al., 1992).
SURFACE MOISTURE OBSERVATIONS
Surface water varied annually as a function of elevation, with topographically lower sites in the east (sites L5 and J11) being submerged more frequently over a 10 yr period than sites in the central (sites G6, I4) and western (sites D10, B7) areas (Fig. 1A). In the 2010–2011 wet season, prior to field measurements (Fig. 1B), only the western sites (D10, B7) remained relatively dry (Fig. 1D), while during the 2011–2012 wet season (Fig. 1C), only the center (G6, I4) underwent prolonged flooding (Fig. 1D). TLS moisture measurements (Table 1) indicated drying (higher values) on surfaces that did not undergo prolonged flooding in 2011–2012 (L5, B7, J11, D10), and a slight increase (lower values; 1%–3%) in surface moisture at flooded sites (G6, I4).
RIDGED PATTERN DEVELOPMENT
The combination of a high water table and flooding in 2011 reset the surface and enabled a flat, continuous crust to develop on the northeastern section of the playa (L5). After the wet season in 2011–2012, when no significant flooding occurred (Fig. 1D), the surface dried enough to initiate ridge formation (Fig. 2A). During 3–12 August 2012 (Table 1), ridge crests were thrust up by as much as 30 mm (Fig. 2E). While the mean surface change was 0.44 mm/day (Fig. 2H), most growth was near the ridge peaks, and the interridge areas remained at the same elevation, as indicated by the asymmetric growth distribution with a peak at zero and a tail extending to 2 mm/day. This initial growth period slowed in the latter part of August and early September (mean rate of 0.23 mm/day between 12 August and 14 September; Fig. 2G). By late September (17–26 September), the growth distribution became more symmetrical (Fig. 2I), ridges had reached their maximum stability, and heights began to reduce (−0.05 mm/day). Fourier transform spectra capture this pattern of ridge growth and decline and indicate increased persistence of ridge spacing in the 200–300 mm region throughout August and early September, and a slight decline in mid to late September (Fig. 2D).
SPATIAL AND TEMPORAL VARIABILITY OF CRUST PATTERNS
In a manner similar to that of the northeast crust at site L5, high levels of surface moisture toward the southeast (site J11; Fig. 3A) in 2010–2011 (Fig. 1D) reset the surface and formed a continuous crust prior to our measurements. The small moisture inputs and a diminished peak in surface water between August 2011 and August 2012 (Fig. 1D) were not enough to reset this surface between our measurement periods, but instead assisted in the development of a fast-forming, distinctive, and widely spaced ridged pattern in 2012 (similar to site D10).
In the drier southwest section of the playa (site D10), a well-defined ridged pattern continued to grow between August 2011 and August 2012. This growth extended the dominance of larger spaced (600 mm) ridges (Fig. 3B) because the playa did not undergo any substantial flooding during the wet seasons in 2010–2011 or 2011–2012 (Fig. 1D). The mixed pattern on the western side of the study area (site B7; Fig. 3E) followed a similar temporal trend in ridge development, but it had fewer ridges initially because it had more extensive flooding in 2010–2011 (Fig. 1D) and had a shallower water table (0.835 m and 1.815 m for sites B7 and D10, respectively; Table 1).
The smooth, degraded surfaces in the central areas of the playa (sites G6 and I4; Figs. 3C and 3D) indicated strong, smaller ridge spacing (between 100 and 300 mm) initially. This topography became muted after the substantial flooding around site G6 (Fig. 1D) during the 2011–2012 wet season, or transitioned through a mixed phase to a well-developed ridged pattern where standing water was not consistently present as at site I4 (Fig. 3D). The surface at G6 did not form typical polygonal ridges during 2012, but instead there was a general swelling in elevation of the entire surface. This swelling was evident to a lesser extent at the drier surface (site I4) and appears to be related to the thickness of the degraded crust, which limits capillary action and subsurface moisture exchange (Nickling and Ecclestone, 1981). Vehicle tracks on the surface (site I4; Fig. 3D) followed the crust development trajectory of degraded to mixed to ridged, and highlight the importance of small-scale topography in crust pattern development. During the wet season in 2011–2012, it is likely that these topographically depressed areas preferentially ponded water, initializing earlier ridge development in 2012. The forcing by moisture inverted the crust pattern, resulting in vehicle tracks becoming topographically positive after the wet season (Fig. 3D).
TEMPORAL VARIABILITY OF zo
In general, zo values (Table 1) across the study site relate to ridge dimensions, and were greater on surfaces with larger element heights (Nield et al., 2013). Temporally, zo values increased as the surfaces transitioned from flatter to ridged patterns. The higher resolution measurement sequence undertaken at site L5 (Table 1) demonstrates the rapidity at which the formation of a ridged pattern can induce an order of magnitude change in zo, increasing from 0.46 mm to 4.25 mm in 9 days (3–12 August 2012). This change in zo occurred as a result of a growth in the mean ridge height from 8.1 mm to 13.8 mm over the same period.
FEEDBACKS, CONROLS, AND DRIVERS OF SALT CRUST PATTERNING
Using high-resolution TLS sequences (days and millimeters), we build upon the conceptual postulations of Krinsley (1970), where polygonal ridge thrusting is driven by efflorescence. We find that ridge development (Fig. 4A; sites L5 and J11 trajectories) is a complex process likely controlled by flooding and interactions among subsurface, surface, and atmospheric moisture exchange (Groeneveld et al., 2010). Positive feedback accelerates ridge development because higher, exposed ridges increase evaporation, salt phase change, and thrusting. Over time, negative feedback diminishes ridge growth (Fig. 4B; site G6 trajectory) because ridges may collapse, either through structural weaknesses or mechanical and chemical breakdown by wind and/or rain. Under extreme external events such as flooding from river flow, rainfall, or upwelling, the surface may be reset at any stage of the formation-degradation cycle (Fig. 4C; site I4 trajectory). Continuous crusts in our conceptual model will have low evaporation and dust-emission rates, ridged crusts will have high evaporation rates but low dust-emission rates, and degraded crusts will have high dust-emission rates but lower evaporation rates than other crust types. Water-table depth, salt chemistry, and antecedent conditions are important controls on the extent to which a surface may progress through different pattern phases. For example, concentrations of thenardite and a high water table may increase the development of ridges and fluffy dust-emitting sediment, or interlocking halite crystals may decrease the availability of dust sediment (Buck et al., 2011). Chemical salt segregation within ridged and flat surfaces may also influence dust emissivity. The resulting surface patterns modify zo (Table 1) and therefore have important implications for dust emission and evaporation rates because zo is a component of both dust entrainment thresholds and turbulent heat fluxes (Deol et al., 2012; Bryant, 2013). For example, the increase in zo over 9 days in August could result in an increase in shear velocity threshold by as much as 350% using the Marticorena and Bergametti (1995) scheme, which could lead to an overprediction in dust emission modeling.
This work shows that detailed time series data on surface microtopography and groundwater and/or surface water status are required to fully characterize crust pattern formation and/or degradation. Our data suggest the following.
Ridge heights increase over time due to positive feedbacks between atmospheric and subsurface processes, unless external controls reset the surface.
Over time, ridge growth rates decrease due to negative feedbacks, limiting atmospheric and subsurface interactions.
On young crusts, ridge initiation and growth can occur quickly (>30 mm/week), changing the erosion potential by increasing zo.
Sediment supply and availability increase as the surface dries out, becomes cracked, and exposes freshly made, low-density, fluffy sediment.
This paper gives the first indication of salt-crust pattern variability, both temporally and spatially, for a major dust hot spot (Gillette, 1997), one of the largest paleolakes in the Southern Hemisphere. Ultimately, we show that it is important to include parameterization of salt-crust pattern dynamics (at the level of detail that we provide here) within climate models to better account for changes in water balance, dust entrainment, sediment availability, and salt accumulation in the geological record.
This study was funded by the Natural Environment Research Council (NE/H021841/1), the World University Network, and a Southampton SIRDF (Strategic Interdisciplinary Research Development Funds) grant. We thank the Botswana Ministry of Environment, Wildlife, and Tourism (permit EWT 8/36/4 XIV), Botswana Ash (Pty) Ltd., W.G. Nickling, the reviewers, and Iridis Southampton Computing Facility.