Determining soil moisture and sediment availability at ...

ClickHere

for

FullArticle

Determining soil moisture and sediment availabilityat White Sands Dune Field, New Mexico, from apparent thermalinertia data

Stephen Scheidt,1 Michael Ramsey,1 and Nicholas Lancaster2

Received 6 May 2009; revised 21 December 2009; accepted 6 January 2010; published 22 June 2010.

[1] Determinations of soil moisture and sediment availability in arid regions are importantindicators of local climate variability and the potential for future dust storm events. Datafrom the Advanced Spaceborne Thermal Emission and Reflection (ASTER) radiometerwere used to derive the relationships among potential soil erosion, soil moisture, andthermal inertia (TI) at the spatial scale of aeolian landforms for the White Sands DuneField between May 2000 and March 2008. Land surface apparent thermal inertia (ATI)data were used to derive an approximation of actual TI in order to estimate the windthreshold velocity ratio (WTR). The WTR is a ratio of the wind velocity thresholds atwhich soil erosion occurs for wet soil versus dry soil. The ASTER‐derived soil moistureretrievals and the changes through time at White Sands were interpreted to be drivenprimarily by precipitation, but the presence of a perched groundwater table may alsoinfluence certain areas. The sediment availability of dunes, active playa surfaces and themargin of the alluvial fans to the west were determined to be consistently higher than thesurrounding area. The sediment availability can be primarily explained by precipitationevents and the number of dry days prior to the data acquisition. Other factors such asvegetation and the amount of surface crusting may also influence soil mobility, but thesewere not measured in the field. This approach showed the highest modeled sedimentavailability values just days prior to the largest dust emission event at White Sands indecades. Such an approach could be extended to a global monitoring technique for aridland systems that are prone to dust storms and for other regional land surface studies in theSahara.

Citation: Scheidt, S., M. Ramsey, and N. Lancaster (2010), Determining soil moisture and sediment availability at White SandsDune Field, New Mexico, from apparent thermal inertia data, J. Geophys. Res., 115, F02019, doi:10.1029/2009JF001378.

1. Introduction

[2] The White Sands Dune Field is located in the TularosaBasin of the Rio Grande Rift in southern New Mexico andcontains a central complex of crescentic dunes (Cd) thatoverlie sediments of Late Pleistocene pluvial Lake Otero(Figure 1) [McKee, 1966]. East and south along the edges ofthe main dune population are parabolic dunes (Pd). Theaeolian system extends across the entire basin between theSan Andres and Sacramento Mountains [Langford, 2003;Kocurek et al., 2007; S. G. Fryberger, Geological overviewof White Sands National Monument, available at http://nature.nps.gov/geology/parks/whsa/geows/index.htm,2003]. Alkali Flat (AF), adjacent to the dune field, is acementation surface, stabilized by gypsum cement. Thelargest deflationary playa is Lake Lucero (LL), which is

commonly inundated, resting in the southernmost and lowestelevation in an area of active playa surfaces (Ac). These playasurfaces are between Alkali Flat to the east and the alluvialfans of the San Andres Mountains to the west. They continueto supply new gypsum sand to the dune field, but the majorityof gypsum sand has been derived from the deflation of LakeOtero during the Holocene Epoch [Langford, 2003]. Activesand transport is commonly observed on both dune andplaya surfaces today, and dust emission events have beendocumented in the field, as well as from orbital remotesensing data.[3] The White Sands Dune Field has been interpreted as a

wet aeolian system where soil moisture plays an importantrole in the sediment dynamics [Kocurek and Havholm,1994; Crabaugh, 1994; Kocurek et al., 2007; Langford etal., 2009]. In a wet dune system, the capillary fringe ofthe water table is at or near the surface. Accumulation ofevaporite minerals occurs with moist conditions, commonlycorresponding to a relative rise of the dense, brine‐richgroundwater table. During deflation, which occurs in dryconditions and a falling water table, the sediments becomemobilized and the source of gypsum for the White Sands

1Department of Geology and Planetary Science, University of Pittsburgh,Pittsburgh, Pennsylvania, USA.

2Desert Research Institute, Reno, Nevada, USA.

Copyright 2010 by the American Geophysical Union.0148‐0227/10/2009JF001378

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, F02019, doi:10.1029/2009JF001378, 2010

F02019 1 of 23

http://dx.doi.org/10.1029/2009JF001378

Dune Field. Therefore, sediment transport in this system issensitive to changes in the regional climate that affect thehydrology of the playa [Rosen, 1994; Fan et al., 1997]. Theperched water table fluctuates seasonally in response tochanges in precipitation, groundwater recharge/dischargeand evapotranspiration within the Tularosa Basin[Allmendinger, 1972; Allmendinger and Titus, 1973;Langford et al., 2009]. S. G. Fryberger hypothesized thatgroundwater is controlling dune morphology at WhiteSands. Langford et al. [2009] confirmed this using a com-bination of GPS topographic surveys and groundwatermeasurements of water table height and salinity, anddescribe the parabolic dunes atop a topographic high relativeto the lower crescentic dunes of the central dune field, anarea of deflation. This topographic high contains a lowersalinity groundwater lens recharged by precipitation, wherethe crescentic dune field is influenced by high saline brine

[Langford et al., 2009]. The parabolic field is largely sta-bilized by vegetation, but sand is mobile on isolated para-bolic dunes and ridges. Interdune areas are stabilized byvegetation and biotic soil crusts, where mobile sand hasbeen observed as widely scattered 2 to 3 mm patchesoverlain on the thicker biotic soil crust [Langford et al.,2009].[4] Several studies have examined the effect of soil

moisture on the erosion potential of sediment [Belly, 1964;McKenna‐Neuman and Nickling, 1989; Saleh and Fyrear,1995; Shao et al., 1996; Chen et al., 1996]. Aeolian ero-sion does not occur until the force of the wind exceeds theforces holding soil particles in place, including the cohesiveforce between particles due to soil moisture [Chepil, 1956].The velocity of the wind at which aeolian soil erosion takesplace is the wind velocity threshold (u*t). A parameterizationfor the large‐scale simulation of dust emission events hasbeen developed (i.e., the DREAM model) that attempts topredict the wet to dry erosion wind velocity thresholds fordifferent soils [Marticorena and Bergametti, 1995; Fécan etal., 1999; Nickovich et al., 2001; Peŕez et al., 2006a,2006b]. The threshold wind velocity has a direct relation-ship to surface roughness and soil moisture, which areaffected by the clay content because of its ability to retainsoil water [Fécan et al., 1999].[5] Measuring soil moisture on large scales is important to

hydrologic, aeolian and agriculture studies. However, field‐based investigations in arid lands are difficult for severalreasons. These areas typically experience a quick dry downfollowing rainfall, and soils in aeolian environments arecommonly very permeable. With respect to soil mobility,the upper few centimeters are the most critical for under-standing erosion potential and aeolian processes. Fieldmeasurements of soil moisture over a large geographicalregion are difficult and expensive, and require a significantamount of specialized equipment and man power. Estimat-ing soil moisture is desert regions is also highly problematicbecause of their remote location, difficult working condi-tions and the need for very sensitive in situ sensors that canmeasure the low water content of desert soils. Accuratemeasurements for soil moisture in playas with high salinityare also difficult for standard soil moisture probes.[6] Synoptic regional‐ to continental‐scale soil moisture

mapping has been successful using spaceborne data, par-ticularly with passive microwave remote sensing instru-ments [Jackson, 1993, 1997; Bindlish et al., 2003] such asthe airborne Polarized Scanning Radiometer (PSR) and thespaceborne Advanced Microwave Scanning Radiometer(AMSR‐E) where reasonable estimates have been obtained,even for vegetated areas [Njoku et al., 2003; Bindlish et al.,2006]. The future Soil Moisture Active Passive (SMAP)instrument planned for launch after 2011 will combine apassive microwave radiometer and high‐resolution activeradar and produce data with a spatial resolution of 1–3 kmand a temporal resolution of 12 h. It will also retrievemeasurements of the surface and vegetation roughness,which are useful for aeolian studies [Entekhabi et al., 2008].Even though radar and passive microwave remote sensingare effective in retrieving soil moisture at a high temporalresolution, the currently available AMSR‐E data with aspatial resolution of 25 km per pixel is limited for theinterpretation for smaller‐scale aeolian landforms (i.e., dune,

Figure 1. Location map of the White Sands National Mon-ument study area centered at 32.9°N, 106.3°W generatedfrom a panchromatic mosaic image acquired by LandsatETM+ between September 1999 and August 2002 (availableat http://ftp.glcf.umd.edu/data/mosaic/). Complete spatialcoverage of the region required two consecutive ASTERdaytime scenes, whereas the night data was generally cen-tered over the dunes and required only one scene. Importantgeographic areas are indicated: Lake Lucero (LL), AlkaliFlat (AF), and the White Sands Dune Field, which includesthe central core of crescentic dunes (Cd) and the parabolicdunes (Pd) to the south and east. Active playas (Ac) thatcontinually resupply sediment to the dune field are southof Alkali Flat. The dashed inset box 1 indicates the areashown in Figure 4. The other numbers on the map indi-cate the locations of the local meteorological stations for thedaily (2) and hourly (3) weather data.

SCHEIDT ET AL.: WHITE SANDS APPARENT THERMAL INERTIA F02019F02019

2 of 23

interdune, and playas) at White Sands. The area of interestfor this study is roughly 26 × 26 km, yielding approximatelyone pixel from AMSR‐E per acquisition. The aeolian fea-tures of interest for this study are far below the spatial scaleof the currently available spaceborne passive microwaveremote sensing instruments.[7] Soil moisture has also been retrieved from apparent

thermal inertia measurements using visible and thermalinfrared data. Previous studies have used Advanced VeryHigh Resolution Radiometer (AVHRR) data to determineland surface thermal inertia [Xue and Cracknell, 1995].Relationships between thermal inertia and soil moisturehave been derived from the Moderate Resolution ImagingSpectroradiometer (MODIS) data [e.g., Cai et al., 2005,2007a]. Because relationships exist among thermal inertia,soil moisture and the erosion potential of sediment, aeoliansediment dynamics that are commonly controlled by thewetting and drying cycles can be directly extracted fromthese data. Quantification of aeolian sediment erosionpotential at a high spatial resolution using satellite‐deriveddata is particularly valuable to the validation of models thatpredict desert drying and dust cycles. The unique propertiesof the Advanced Spaceborne Thermal Emission andReflection (ASTER) radiometer provide this capability tomonitor and model the small‐scale spatial variability of anactive aeolian system. ASTER has been used to examinelarge‐scale aeolian systems after mosaicking data [Scheidt etal., 2008a], and the temporal frequency of ASTER wasleveraged to examine changes in sediment dynamicsthrough time [Katra et al., 2009; Katra and Lancaster,2008].[8] The approach presented here of determining the

ASTER‐derived sediment availability using the relationshipbetween thermal inertia, soil moisture and the erosionpotential of soil is unique in its ability to directly retrieve anapproximation of the WTR. At the ASTER spatial resolu-tion, the WTR is at the scale of the aeolian landforms foundat White Sands. This paper investigates these relationshipsfor both ASTER and MODIS data, but focuses on ASTERto document the small‐scale patterns of sediment availabilitythrough time. The main objectives are (1) to present amethodology for applying the Xue and Cracknell [1995]thermal inertia model to multispectral high spatial resolu-tion ASTER data, and compare the results to the highertemporal resolution/lower spatial resolution MODIS data,(2) to examine the spatial and temporal variation of thermalinertia, coincidently with soil moisture, the climate, and thegeomorphology of the White Sands aeolian system, and(3) to demonstrate the potential of remote sensing as apredictive monitoring tool to retrieve the WTR, which is aquantitative measure of sediment availability and erosionpotential.

2. Background

2.1. Aeolian Sediment Dynamics

[9] The prediction and monitoring of aeolian erosion isimportant for the management and conservation of naturalresources, agricultural land, and the prediction of air pol-lution. The initiation of movement of sand and mineral dusthas been extensively studied [Nickling and McKenna‐

Neuman, 1994; Greeley and Iversen, 1985]. The initiationof sand saltation and dust emission occurs once the shearstress exerted by the wind onto the sediment surface over-comes the frictional and cohesive forces holding that sedi-ment in place. Sediment that has a high potential for winderosion and subsequent transport is defined as having highsediment availability [Kocurek and Lancaster, 1999]. Sed-iment availability can also be defined as the percentage ofsurface soil particles for which the threshold wind velocityis exceeded under a given set of conditions [e.g., Nickovichet al., 2001; Shao et al., 1993]. The threshold wind speed orwind shear stress for entrainment of sediment depends on anumber of complex and interdependent factors includingthe: grain size, presence of roughness elements (e.g., vege-tation), soil moisture, soluble salts, crusting, and cohesion[Nickling and McKenna‐Neuman, 1994]. These factorsdetermine the availability of sediment for transport in abasin‐scale, sediment state model for an aeolian system[Kocurek and Lancaster, 1999]. Sediment is continuallyreworked within a single aeolian system on short timescales, resulting in considerable spatial and temporal varia-tion in sediment transport, storage and availability.[10] Soil moisture plays a major role in affecting the ero-

sion potential of sediment. Sediment availability decreaseswith increasing soil moisture because of the adhesive andcapillary forces that bind wet sediment particles together[e.g., Hotta et al., 1984; Sherman, 1990; McKenna‐Neumanand Nickling, 1989]. Soil composition, specifically the claycontent, is an important factor in determining the degree ofthe soil’s water retention and therefore the cohesion betweenparticles [Fécan et al., 1999]. The effect of soil moisture onthe erosion of wet sand and soil has been modeled theo-retically and empirically by several researchers [e.g., Chepil,1956; Belly, 1964; Shao et al., 1996; Fécan et al., 1999].Cornelis and Gabriels [2003] found that empirical modelsfrom Chepil [1956] and Saleh and Fyrear [1995] were goodpredictors because they use the soil moisture content of theupper most surface layer for the prediction of the thresholdwind velocity. Soil composition and textural characteristicsof an aeolian environment are expected to change inresponse to hydrologic and climatologic conditions, such asthe flux of rainfall, soil moisture and groundwater. In a playaenvironment, for example, these changes can occur rapidlyon the scale of minutes to days, where drought then becomesan important concern with respect to land use. The capacityof the surface to store water also varies spatially with theperiodic inundation and drying of playa lake beds, the influxof fine silt and clay [Reheis, 2006], and the crusting ofevaporite minerals [Langer and Kerr, 1966]. Dry conditionsfavor a higher degree of sediment availability, whereas wetconditions result in sediment immobilization [e.g., Reynoldset al., 2007].

2.2. Thermal Inertia Modeling

[11] Thermal inertia (P) is a physical quantity describingthe resistance of a material to change in temperature and isdefined as

P ¼ffiffiffiffiffiffiffiffiK�c

pð1Þ

where K is the thermal conductivity, r is the density and c isthe specific heat. However, these parameters cannot be


3 of 23

obtained from remote sensing measurements and thereforethey must be modeled or estimated from other directlymeasured data. A thorough review of determining thermalinertia from remote sensing data is presented by Cracknelland Xue [1996]. This thermophysical measurement has alinear relationship to changes in the soil density, particle sizeand moisture. Because wet soil or sand has a higher thermalinertia than dry soil, previous studies have used this propertyto estimate soil moisture from remote sensing data [Prattand Ellyett, 1979; Price, 1980, 1985; Zhang et al., 2002;Zhenhua and Yingshi, 2006; Cai et al., 2005, 2007a].[12] Apparent thermal inertia (ATI) is related to the actual

thermal inertia; however it can be generated directly fromremote sensing data. ATI is a relative measure of thereflected solar albedo to the difference in emitted brightnesstemperature over the diurnal cycle and is defined as

ATI ¼ NC 1� að Þ= �Tð Þ ð2Þ

where a is the land surface albedo over the visible/nearinfrared (VNIR) and the short‐wave infrared (SWIR) region,and DT is the difference in brightness temperature betweenday and night satellite overpasses. Additional scaling factorsand other constants (N and C) account for variations in solarflux with latitude and solar declination, and may or may notbe included depending on the study. The land surface albedoand temperature are the critical remote sensing parameters.For areas of similar surface albedo, a high ATI value resultsfrom a small DT between day and night, whereas a low ATIvalue results from a large diurnal change in temperature. Fora given DT, a higher albedo will result in a lower value ofATI. Studies using a TI model for geological interpretationinclude Watson et al. [1971], Kahle et al. [1976], Kahle[1977], and Gillespie and Kahle [1977]. These modelshave been used with various success using different datasets [e.g., Price, 1980, 1985], such as the Heat CapacityMapping Mission (HCM) [Price, 1977], AVHRR data[Xue and Cracknell, 1992, 1995; Wang et al., 2004], andASTER data for the purpose of thermal anomaly detectionin oil fields [Nasipuri et al., 2006] and oil spills [Cai et al.,2007b]. Various quantitative relationships between ATI andsoil moisture for a given soil density have been proposed[Ma and Xue, 1990]. For example, Cai et al. [2005, 2007a]used this relationship to determine soil moisture fromMODIS data, and the average difference of volumetric soilmoisture between remote sensing and in situ measurementsof the upper 5 cm soil surface in an agricultural test plot was<5% for a single point. A simple thermal inertia model byZhang et al. [2003] was used to model evapotranspirationfrom albedo and temperature data acquired from ASTER fora small drainage area in China [Huang et al., 2006].[13] The spatial resolution of ASTER is well suited for the

application of these surface temperature models [Coolbaughet al., 2007] and examining aeolian surfaces at the scalefound at White Sands. In another wet aeolian system (SodaLake, California), field observations of soil texture andASTER thermal infrared (TIR) data were used to examinethe spatial and temporal variability in composition [Katraand Lancaster, 2008]. Variations in hydrology were dis-cussed, including the frequency and timing of precipitationreceived at the study site. Soil moisture data for the coin-cident satellite acquisition were not available from remote

sensing data or directly measured in the field. Because soilmoisture can affect the emissivity and composition retrievals,soil moisture and TI data would be helpful in interpretingremote sensing retrievals of surface composition and theinterpretation of aeolian surface dynamics at Soda Lake[Katra and Lancaster, 2008].

3. Methods

[14] This section briefly describes the data and the thermalinertia model used in this study. A detailed review of theASTER and MODIS data sets as well as the thermal inertia,soil moisture and the WTR models are found in Appendix A.The focus of this study is aeolian landforms on the spatialscale of tens to hundreds of meters, an ideal target forASTER. However, MODIS data are compared to ASTER inorder to assess the effects of high temporal versus highspatial resolution data.

3.1. Remote Sensing Data of White Sands Dune Field

[15] MODIS and ASTER collect coincident reflected andemitted radiance, but at much different spatial, temporal andspectral scales. ASTER acquires data at a repeat time ofabout 16 days, but this frequency depends on latitude andoff‐axis pointing. MODIS data are generally acquired twiceevery 24 h (once daily and once nightly) for the WhiteSands study area. ASTER was designed to acquire repeti-tive, high spatial resolution, multispectral data over theVNIR, SWIR, and TIR portions of the spectrum[Yamaguchi et al., 1998; Abrams, 2000; Pieri and Abrams,2004]. ASTER is composed of three subsystems with threechannels in the VNIR, six channels in the SWIR, and fivechannels in the TIR, and having spatial resolutions of 15 m,30 m, and 90 m, respectively. ASTER acquires routine dayand nighttime TIR data, as well as day time data in theVNIR and SWIR. However, after 23 April 2008 the SWIRdata were no longer available due to a malfunction in theSWIR subsystem.[16] From 2000 to 2008, there were 54 daytime and

70 nighttime, cloud‐free acquisitions of the White Sandsarea by ASTER. The satellite footprints of the day and nightdata acquisitions have different orientations and only theintersection of these over the White Sands study area wereused (Figure 1). This reduced the large data set to sevenday‐night image pairs for this study. Each image pair had aminimum time difference (Dt) of 36 h (Table 1). Ideally,satellite acquisitions should occur very close to the times ofmaximum and minimum diurnal temperatures for modelingATI. However, this is not possible with the ASTER orbits atthis latitude. Furthermore, with a greater Dt, the probabilityincreases of unexpected anomalies unrelated to the diurnalcycle of solar heating and cooling (e.g., rain, large windevents, etc.). The night image acquisitions occurred at about2210 local time and the day image acquisitions about 36 hlater. MODIS data can be acquired with aDt ≈ 12 h betweenday and night acquisitions, but only the MODIS dataapproximately nadir/coincident with ASTER were used inthis study to allow direct comparisons between the instru-ments. The calibrated data sets from the MODIS andASTER instruments were then used to calculate the keyinputs of the thermal inertia model: albedo (a) and the day‐night temperature difference (DT).


4 of 23

3.1.1. Temperature Retrieval[17] Day and night land surface temperatures (LST) were

extracted from ASTER TIR data using the emissivity nor-malization method [Realmuto, 1990; Kahle and Alley,1992]. The temperature difference (DT) was calculated foreach of the day/night image pairs from both the MODIS andASTER data. No manipulations or scaling calculations wereapplied to the temperature data, and the agreement betweenthe average DT of MODIS and ASTER through time waswithin the MODIS overall accuracy (±2 K). However,where compared to MODIS data, ASTER had a 15% largerrange in temperature (DT) data most likely due to higherspatial resolution.3.1.2. Broadband Albedo Estimation[18] Broadband albedo (a) was calculated directly from

surface spectral reflectance products of both MODIS andASTER data using the approach of Liang [2001]. Becauseof spatial resolution effects and differences in atmosphericcorrection techniques, ASTER broadband albedo was sig-nificantly higher than MODIS values (see Appendix A).ASTER broadband albedo was scaled to that of MODIS forthis study as a correction method for several reasons: (1) theMOD09 reflectance product was concluded to have a moreaccurate atmospheric reflectance correction than the ASTERreflectance product [Miura et al., 2008], (2) ASTERbroadband albedo values were overestimated (>1.0) in someareas of the dune field, and (3) the use of previously pub-lished relationships between the TI and soil moisture havebeen developed for MODIS data. After scaling, the ASTERbroadband albedo data have the same mean but still have alarger dynamic range (∼20%) than MODIS data, which isexpected for an instrument of higher spatial resolution.

3.2. Thermal Inertia Model

[19] The temperature difference and broadband albedoresults were used as input to the thermal inertia model givenby Xue and Cracknell [1995] that can be used for areasof variable soil moisture. The model was reproduced forthis study in the Interactive Data Language (IDL) for use inEnvironment for Visualizing Images (ENVI) software usingthe MODIS and ASTER data and metadata as inputs.Applying this model to remote sensing image data resultedin thermal inertia unit (TIU) maps at a spatial resolutionequal to that of the input data (i.e., 90 m pixel−1 forASTER). The data produced in this paper are reported anddiscussed as approximated TI data, which represents ATI atthe scale and units of real thermal inertia (P). Where com-

pared to MODIS results, ASTER data values had a 25%larger range in TI retrievals directly related to the largerrange of albedo values.

3.3. Estimation of Erosion Threshold Velocity RatioFrom Soil Moisture

[20] Soil moisture is determined as a function of thermalinertia and assumed soil density using a model for bare orsparsely vegetated surfaces given by Ma and Xue [1990](Figure 2a). For arid lands where the actual soil moistureis low (0 ∼ 15%), the density is assumed not to significantlyaffect soil moisture retrievals. For the data analysis here, adensity of 2.65 g cm−3 was assumed for the calculation ofthe TI. The resulting soil moisture values were then used asinput for a model that determines the erosion potential ofsoil. We use the operational parameterization of the winderosion threshold ratio (WTR) as a function of soil moisturefor semiarid soils, given by Fécan et al. [1999] (Figure 2b).

3.4. Climatic Conditions and Data

[21] Temperature and precipitation data were collectedfrom historical sources, but no known field work had beenconducted on soil moisture or aeolian sediment erosionduring any of the historical overpass dates and times.Weather data for the White Sands Dune Field were gatheredfrom local stations to determine if correlations exist amongprecipitation, modeled soil moisture and remote sensingretrievals, specifically for the time periods prior and coin-cident with the Terra satellite overpasses when ASTER andMODIS data were collected. Hourly precipitation data wereavailable from U.S. Air Force (USAF) station 722693located at 32°50′N, 105°59′W (http://cdo.ncdc.noaa.gov/),and daily data were available from National Climate DataCenter Cooperative (NCDC COOP) weather station 299686located at 32°47′N, 106°10′W.[22] Total precipitation in the White Sands area is about

22 cm yr−1, with the peak rainfall amount occurring inAugust and the minimum in May. Maximum rainfall in thisarea is in late summer to early autumn, with a secondmaximum in the winter. Precipitation events have a majorimpact on the results in two distinct ways: (1) they canproduce an error in thermal inertia calculations due tochanges in surface temperature and (2) they replenish thesoil moisture of the system. Examination of the hourlyprecipitation data for White Sands showed that precipitationevents did not occur between day and night overpass times,therefore no errors were expected. Time series of precipi-

Table 1. ASTER Acquisition Dates/Times and the Associated Relevant Climate Data for the Seven Image Pairsa

t1 t2 Td Tn tmin(LT) tmax(LT) Tmax_h Tmin_h Tmax_d Tmin_d � (%) dry (days)

1 8 Feb 2002; 1058 LT 6 Feb 2002; 2212 LT 286.9 275.2 0652 1631 290.8 270.2 288.0 265.0 17.2 12 7 Nov 2002; 1057 LT 5 Nov 2002; 2211 LT 289.1 278.0 0510 1630 294.1 274.1 294.0 271.0 22.6 103 23 Nov 2002; 1057 LT 21 Nov 2002; 2211 LT 288.0 280.8 0550 1510 295.8 271.9 295.0 266.0 21.8 124 4 May 2004; 1056 LT 2 May 2004; 2210 LT 300.2 286.9 0453 1630 305.2 280.2 301.0 277.0 19.6 225 8 Apr 2006; 1056 LT 6 Apr 2006; 2209 LT 291.9 289.1 0112 1630 298.0 279.1 298.0 273.0 14.5 426 27 Apr 2007; 1056 LT 25 Apr 2007; 2210 LT 300.8 289.1 0431 1530 ND 281.9 302.0 286.0 29.1 127 12 Mar 2008; 1056 LT 10 Mar 2008 2210 LT 290.8 280.2 0505 1630 295.2 275.2 294.0 270.0 17.9 25

aMODIS times are within several minutes. Here, t1 and t2, sensor overpass time (day and night, respectively); Td and Tn, atmospheric temperature(Kelvin; day and night, respectively); tmax and tmin, local time of the maximum and minimum air temperatures; Tmax and Tmin, maximum andminimum hourly (h) and daily (d) air temperatures (Kelvin) from local weather stations; �, average estimated volumetric soil moisture for the WhiteSands region from U.S. climate data; dry, the number of consecutive days without precipitation prior to the orbital overpass and acquiring data.


5 of 23

tation are reported here, which include (1) the total monthlyprecipitation trend during the study period (Figure 3g) withrespect to the ASTER overpasses and (2) the daily precipi-tation trends 48 days prior to each of the satellite overpasstimes (Figure 3a–3f). The number of consecutive dry dayswithout a precipitation event in the White Sands area, andthe predicted soil moisture data interpolated for the studyarea (bounded between 33.3 and 32.6°N and 106.9–106°W)are reported in Table 1. The predicted soil moisture datawere produced by a one‐layer, hydrologic model, referred toas H96, at a resolution of 0.5° across the entire United Statesfrom disperse meteorological measurement stations [Huanget al., 1996; van den Dool et al., 2003]. The H96 modelaccounts for precipitation, runoff, groundwater loss andevaporation, and is utilized for soil moisture prediction forthe real‐time National Drought Monitor [Svoboda et al.,2002].[23] Temperature data were used to determine the time of

maximum daily temperature or tmax for the thermal inertiamodel, aid in the interpretation of TI and soil moistureretrievals, and to detect a source of error in the TI retrievalsthat could be due to season or weather. An attempt was madeto find relationships between the data and MODIS watervapor retrievals as well, but these MODIS vapor data werefound to be unreliable because of inconsistent spatial cov-erage. Using the atmospheric temperature data from both thehourly and daily local weather stations, we determined thatthe study area experienced normal diurnal temperaturefluctuations during the satellite acquisitions. The satelliteoverpass times (t1 and t2) and the daily maximum (Tmax) andminimum (Tmin) atmospheric temperatures occur at roughlythe same time for each date, with one exception on 6 April2006 when the maximum temperature occurred at 0112local time. The atmospheric temperatures at t1 and t2 arereported here as Td and Tn (Table 1). Tmax is always greaterthan Td, and Tmin is always less than Tn. Although the LSTwave is expected to have higher amplitude, the average LSTchange is assumed to have tracked the average atmospherictemperature wave. The available climate/weather data show

that no abnormal local weather conditions (i.e., frontalweather systems) occurred during or between satelliteacquisitions of day and night image pairs, which would haveadversely affected TI calculation. From these available data,it was concluded that the DT is therefore unlikely to beconsidered a source of significant error in the calculationof TI.

4. Results and Analysis

4.1. Comparison of MODIS and ASTER Data

[24] The differences in VNIR reflectance between a dry(April 2006) and wet (April 2007) period were readilyapparent in the regional‐scale ASTER and MODIS colorcomposite images (Figure 4). The inundation of the largestplayas, including Lake Lucero, was observed in both datasets. The large‐scale changes in the albedo patterns of thegypsum sands and vegetation on the surrounding alluvialfans was also clear in MODIS data. However, the 15 mspatial resolution of ASTER was able to better resolve thefeatures and processes of interest within the dune and playasystem (e.g., the inundation of many smaller playas andalluvial drainages, dune and interdune surfaces, and changesin the spatial pattern of gypsum and evaporite deposits,especially in active playas).[25] The study area had full coverage in the ASTER and

MODIS reflectance data (AST07XT and MOD09 standardreflectance products, respectively). The broadband albedocalculations from MOD09 data had a difference of less than1% compared to the MOD43 broadband albedo wherespatial coverage of MOD43 was available. Comparison ofthe ASTER and MODIS albedo was difficult because of thedifference in spatial resolutions. AT MODIS resolution,ASTER and MODIS have the same mean and generalspatial pattern in the temperature and albedo data. The dif-ferences in the input data for the thermal inertia model werebest recognized by examining the broadband albedo andtemperature difference images calculated from MODIS(Figures 5a and 5b) and ASTER (Figures 5c and 5d). The area

Figure 2. The model relationships between soil moisture and thermal inertia and wind velocity ratio(WTR) as a function of soil density (g cm−3) used for the ASTER data in this study. (a) Thermal inertiaplotted as a function of soil moisture. The assumed density used for the study is 2.65 g cm−3 (reproducedfrom lookup tables by Ma and Xue [1990]). (b) The erosion threshold velocity as a function of soilmoisture and density [Fécan et al., 1999]. These relationships were combined in order to show thethreshold WTR as a function of thermal inertia.


6 of 23

Figure 3. Daily precipitation totals 48 days prior to each of the ASTER data acquisitions that were usedto retrieve thermal inertia and soil moisture in Figure 6. (a–f) Precipitation totals from station 296886. Thedays of the satellite overpass used in this study are shown by the vertical dashed lines. (g) Monthlyprecipitation totals for the entire time frame also compared with the satellite overpass dates (vertical lines).


7 of 23

of Figure 5 corresponds to the inset box from Figure 4dlabeled “zoom.” Temperature data from ASTER are 11times sharper and albedo data are 16 times sharper comparedto MODIS.[26] The results of the TI modeling using ASTER data

were expected to be more useful because of the ability todistinguish small‐scale thermophysical differences in aeo-lian geomorphology. Nonetheless, TI was retrieved using

both ASTER and MODIS data at their native resolution inorder to compare the results over time and assess the ben-efits of high spatial versus high temporal resolution. Thiscomparison was useful for several important reasons. First,the similar capability and accuracy of the instrumentsallowed for variations due to instrument effects to be sep-arated from real temporal and spatial trends. For example,the highest average TI values were calculated for the entire

Figure 4. ASTER and MODIS color VNIR data for a typical dry period (6 April 2006, left) and a wetperiod (27 April 2007, right). (a–b) The MODIS data product (MOD09) color composites (bands 2, 1, and4 in red, green, and blue, respectively). (c–d) ASTER color composites (bands 3, 2, and 1 in red, green,and blue, respectively). The typical extent of inundation of Lake Lucero can clearly be seen during thewet period (indicated by white arrows). ASTER data also show a higher level of detail in dune morphol-ogy and changes in the playas surfaces. The inset box labeled zoom denotes the area shown in Figure 6,whereas the smaller box denotes an area where parabolic dunes are resolved by ASTER but not MODIS.


8 of 23

image data on 6 February 2002 compared to other data.Because the same general temporal and spatial trends wereobserved by MODIS, an individual sensor or instrumenteffect was ruled out for these high values. Second, previousstudies that retrieve TI and soil moisture using the Xue andCracknell [1995] model relied on MODIS data at differentstudy sites [Cai et al. 2005, 2007a]. The retrieval usingMODIS for this study of White Sands demonstrates validvalues (∼400–2000 TIU) similar to those reported previ-ously and that the model is working correctly.

4.2. Spatial Patterns of Thermal Inertia

[27] TI retrievals of the White Sands dune and playaaeolian system from seven different dates between 2002 and2008 showed significant spatial and temporal variation(Figures 6a–6g). The data are shown as a subset of theWhite Sands Dune Field and playa area in order to focus onsmaller‐scale aeolian features (shown as an inset box on themap in Figure 4d). The majority of the ASTER TI data fallwithin the range of 400–1400 TIU with 95% of the datafalling within the range of 900–2000 TIU. The highest TI

Figure 5. Comparison of MODIS and ASTER derived image products acquired on 6 April 2006, whichwere used to calculate TI. (a) MODIS‐derived broadband albedo. (b) MODIS‐derived DT. (c) ASTER‐derived broadband albedo. (d) ASTER‐derived DT. Broadband albedo values range from 0 to 0.60 andDT values range from 25 K to 60 K. Both image pairs are linearly stretched equally for greatest imagecontrast.


9 of 23

Figure

6.TIderivedfrom

theASTERdata,which

correspondsto

arangeof

soilmoisturefrom

9%to

25%.Regions

inblue

aresurfaces

with

higher

soilmoistureandthereforeim

mobile

sediment,whereas

areasin

redaredriersurfaces

thatare

moresusceptib

leto

winderosionviasand

saltatio

nor

dustem

issions.(a–g)The

seventim

eperiodsdetailedin

this

study.

(h)The

averaged

imageproductof

allsevenim

agepairsshow

ingtheTItrendover

time.


10 of 23

was associated with very wet surfaces, such as those foundat the inundated Lake Lucero and playa (area A5). Fullinundation of areas like Lake Lucero is denoted by the darkblue color in Figures 6d and 6f, whereas the lake was par-tially inundated in 6b and 6c, and nearly or completely dryin 6e and 6g. Close examination of the ASTER TI imagesrevealed that many other small persistent playa lakes can befound within the region (e.g., Figures 6f, areas C1, D1, andE2), several of which occur as higher TI areas in the inter-dune areas (e.g., Figure 6b, area 4E). Several of theseinundated interdune areas correspond to the spectral char-acteristics of water, seen as the light blue color in theASTER VNIR reflectance. The western portion of the imagesubset represents the termination of the Andres Mountainalluvial fans abutting the edge of active playas and AlkaliFlat (Figure 6, portions of area A1, A2, A3, A4, and A5).This area has a pattern of consistently lower TI at the distaledges of the fans, and was interpreted as fine‐grained sedi-ments deposited at the edges of the playa.[28] Other regions have a spatially complex and tempo-

rally variable pattern of TI. Several of the active playa sur-faces have extremely low values in some areas (Figure 6g,areas B2–B4, C2–C4), which corresponds to very dry con-ditions and the high‐albedo gypsiferous evaporite. Theopposite trend was observed in the previous date of the timeseries (Figure 6f), when the same area was wet. Justnortheast of Lake Lucero is a circular area of consistentlylower TI (see Figures 6b, 6c, 6d, 6f, and 6g, area B4), but inFigure 6e this area has an average similar to that of itssurroundings. This circular feature is frequently rimmed by avery low TI area that was interpreted as a significantlyfluctuating amount of evaporite crust in the active playa. Forexample, in Figure 6e, this region has an average TI (greencolor) and a thin zone of lower TI (red color) borders thesouthern edge. This feature was interpreted as a high‐albedogypsiferous evaporite surface. Clouds cause anomalouslyhigher TI values, where an example is found at the topcenter of the image in Figure 6g (area C1). This propagatesinto the average TI image as an overestimation of TI and soilmoisture (Figure 6h, area C1), and demonstrates the modelssusceptibility to atmospheric conditions.[29] Dune ridges consistently have lower values of TI due

to the dry, mobile, high‐albedo gypsum sands, including thelarger parabolic dune forms at the southern edge of the dunefield (e.g., area D5 and E5). Large areas west of the dunesalso have persistently low TI values (see Figures 6b, 6d, and6g, areas B5 and C5). In other areas of the dune field, someinterdune areas persist as high TI regions despite the soilconditions in the rest of the image. These areas are inter-preted to have consistently higher soil moisture and inter-dune vegetation cover or biotic crust. An increase in theareal coverage of vegetation is expected to increase TIretrieval values because of water content in vegetation.Likewise, the lower albedo may also increase TI values forthese particular pixels if vegetation is a major scene com-ponent. Some interdune areas are covered by a clay‐richcrust that has a slightly lower albedo, causing an increase inTI retrieval. Unique to the composition of the White Sandsarea, the interdune areas and stoss slopes of the dunes alsohave a higher cohesion and bulk density due to the partialcementation of gypsum sand grains. The effect on TI is not

known where the hard and compact cores of these dunes areexposed by erosion of loose sand.[30] The average TI image (Figure 6h) shows the persis-

tent spatial patterns of the White Sands aeolian system. Forexample, the crescentic dunes in the south of the dune fieldhave a much lower average TI (100–200 TIU) than theadjacent interdune areas (400–600 TIU). Interdune areas areaffected by the interaction of the groundwater table with thesurface [Kocurek et al., 2007; Langford et al., 2009]. Playasurfaces were also found to have high standard deviations ofTI values, similar to the interdune areas. Several smallerplayas and persistently wet interdune areas with higher TIare easily identified throughout the dune field. The meanand standard deviation of the seven image dates were gen-erated and represent a range of soil moisture conditions. Thestandard deviation of each pixel is a relative estimate of thevariability of TI, and therefore, soil moisture variability. Asexpected, a higher standard deviation is observed on playaand interdune areas compared to dunes, indicating a higherdegree of variability in soil moisture through time in theseareas. However, the data constitute a limited temporal subsetand therefore the interpretation of longer‐term trends isdifficult. For example, exclusion of the image data shown inFigure 6a (a very wet time period) biases the statisticsshowing the greatest amount of variability in dry areas ofAlkali Flat. Exclusion of image data shown in Figures 6e or6g removes the influences of very dry conditions, and thewet areas appear more significant. Therefore, even thoughthe average of all seven images gives the best availablerepresentation of the average TI (and soil moisture), addi-tional data would further aid the refinement of the temporaltrends in the dune field.

4.3. Temporal Trends of Soil Moisture

[31] Independent in situ soil moisture measurements werenot available for these ASTER data acquisition dates, whichwould have provided ground truth and an accuracy assess-ment of soil moisture retrievals. The time series of ASTER‐derived soil moisture were compared to the modeled H96soil moisture at a regional scale. The meteorological stationsused by H96 are not located directly in the dune field, andthe gridded data products do not have the same spatialresolution to match ASTER or even MODIS data. The timeseries of H96 soil moisture were extracted from the mete-orological stations surrounding the study area and spatiallyinterpolated to give the best possible comparison of H96model data to ASTER‐derived data (Figure 7a). In additionto the differences between spatial resolution, obvious dif-ferences in these two data sets arise because of the verydifferent methods by which soil moisture is derived. How-ever, because the H96 model predicts soil moisture based ona specific set of hydrologic variables and for average soilconditions, some comparisons to the ASTER derived dataresult in some hypotheses of hydrologic forcing on soilmoisture occurring in an arid, wet aeolian system of dunesand playas. For example, the H96 soil moisture data can becompared to measured precipitation from the local weatherstations (Figure 3a–3g) giving possible insight into the effectson ASTER‐derived data (discussed below).[32] Image statistics show that the ASTER‐derived data

have a greater range of soil moisture values through time


11 of 23

than the H96 model data. ASTER‐derived data range from<10% to 100%, and these values are highly spatiallydependent as previously discussed. Average H96 data rangefrom a minimum of 14% on 6 April 2006 to a maximum of29% on 27 April 2007 (Figure 7a). Average ASTER‐derived soil moisture values were highest for 6 February2002, and the driest values were for 10 March 2008. Ingeneral, the comparison of the land surface soil moisturedata between H96 and the ASTER‐derived data for theseven ASTER acquisition dates are between 10% and 30%(periodically inundated playa and interdune surface areas areexcluded), but the data do not correlate well through time(Figure 7). These results are not surprising considering thegreater spatial resolution of ASTER, which captures a highdegree of local variation in soil moisture. The time seriesdata from ASTER are spatially variable and depend greatlyon the location in the White Sands system from which thetime series data are taken (Figures 7b–7d). The H96 soilmoisture values predicted for each of the dates examined inthis study were compared to the monthly totals of precipi-tation (from Figure 3g) in a simple least squares regression.H96 soil moisture predictions were directly related to theamount of precipitation (R2 = 0.92) if the 6 February 2002data point was excluded. The 6 February 2002 soil moisture

was the result of a localized precipitation event and notcaptured by H96.[33] The time series of ASTER‐derived soil moisture was

extracted from various regions of interest that representdifferent geomorphic settings within the aeolian system(Figures 7b–7d). Different trends of soil moisture wereexpected and were observed from these areas. Large playalakes, like Lake Lucero (LL), are periodically inundated,reaching complete saturation at times, and complete desic-cation at other times (Figure 7b). Other playa and interduneareas show similar trends, but do not reach complete satu-ration (Figures 6c and 6d). Alluvial fan (AL) and dune areashad the lowest variation of soil moisture (6% and 8%,respectively) and similar trends through time. Dune areasalmost always had lower soil moisture than their adjacentinterdune areas. Parabolic dunes at the southern end of theWhite Sands Dune Field had lower soil moisture and lessvariability than crescentic dunes located in the central part ofthe dune field (Figures 7c and 7d), although locally thesedifferences were small. The pattern of high interdune andlow dune soil moisture is much less than the soil moisturevariation found at LL, playas and AF. Dune and interdunevariability is also somewhat dampened in the time seriesbecause of the spatial averaging where regions of interest

Figure 7. Temporal trends of soil moisture for various geomorphic features and surfaces at White Sands.(a) The predicted average soil moisture extracted from U.S. climate data for the White Sands Dune region[Huang et al., 1996; van den Dool et al., 2003]. (b) Lake Lucero (LL), Alkali Flat (AF), and the lower SanAndres alluvial fans (AL) west of the dune field. The variation is high for Lake Lucero and Alkali Flat,ranging from complete inundation of the lake to very dry. The alluvial fan variation is low and moretypical of the average soil moisture conditions. (c) Central crescentic dunes and interdune areas.(d) Southern parabolic dunes and interdune areas. Higher soil moisture conditions were predicted for thecentral crescentic dunes and the majority of the interdune areas consistently had higher soil moistureretrievals than dunes.


12 of 23

were created for the time series. Not all interdune areasexperience the same wetting and drying cycles, and there-fore some differences between adjacent dune and interduneareas may be greater than others (Figure 6). Lower soilmoisture in the dune field may also be due to greater sandthickness. In general, the dunes sit on top of a sabkha sur-face that is connected to and interacts with the underlyinggroundwater table. Dune cores are commonly cemented atWhite Sands, and my also contain perched groundwater.Where sand thickness is low, the underlying moist soil willincrease TI retrievals relative to thick, dry sand cover. DuneTI and therefore soil moisture are at their lowest at the top ofthe dunes where these areas are well drained and not verywell connected to groundwater fluctuations. The southerndunes have the least amount of soil moisture variation (13 to16%) over time, not including the very wet conditionspresent on 6 February 2002 (see also Figure 6a).[34] The wet conditions on 6 February 2002 were con-

sidered to be anomalously high where compared to all of theother ASTER‐derived soil moisture data, but the H96hydrologic model did not predict wet soil conditions for6 February 2002. This was not due to an effect of highspatial resolution (e.g., capture of a typical playa lake sig-nature). The image wide ASTER‐derived soil moisture washigher than any of the other ASTER image data from anyother date (Figure 6a). Examination of the daily precipita-tion record from weather station 299686 at White Sandsshowed that a total of 23 mm of precipitation was recordedin the days prior from 24 January to 5 February 2002(Figure 3a). The majority of this precipitation occurred4 and 5 February for a total of 16 mm. Therefore, the localrainfall recorded at weather station 299686 better explainsthe resulting ASTER derived soil moisture on 6 February2002 and highlights the benefit of using ASTER observa-tions in this application.[35] Both ASTER‐derived and H96 data predicted rela-

tively wet condition for 25 April 2007. Observation of themonthly precipitation in Figure 3g shows the highest rainfallin the year prior to 25 April 2007, however this did not havean impact on surface soil moisture conditions. The 25 April2007 ASTER image data do not show widespread wet soilconditions, but several smaller areas of very wet conditionsexist, such as in playas, interdune areas and some alluvialoutwashes. Significant precipitation events occurred on14 and 35 days prior to the image acquisition (9 and 12 mm,respectively) and may be responsible for recharging sub-surface water and Lake Lucero (Figure 3e). This is inter-preted as a period in which the subsurface soil moistureremained persistently high and the groundwater table wasalso persistently high. The upper soil and sand layers are drydue to evaporation, rapid dry down and the persistent aridconditions. Low monthly precipitation was observed prior tothe 12 March 2008, 6 April 2006, and 2 May 2004 ASTERdata acquisition dates (Figure 3g). Monthly precipitationwas high in May 2004, but because the data acquisitionoccurred early in the month, precipitation during April 2004was responsible for forcing high soil moisture values shownin Figure 5d. Significant precipitation events in the monthprior to the satellite data acquisition on 2 May 2004 raisedLake Lucero levels during which time soil surfaces experi-enced a quick dry down due to evaporation (Figure 3c). Thiswas consistent with the dry soil conditions determined from

ASTER for the region. However, the ASTER‐derived datashow drier conditions on 12 March 2008 (Figure 6h) than6 April 2006 (Figure 6e). 6 April 2006 represents very drysoil moisture conditions, and 12 March 2008 shows thedriest conditions relative to all the other dates. H96 does notshow this trend, and the reason for this difference cannot beexplained by the available precipitation data. Evaporationrates and groundwater levels could possible lead to a betterunderstanding of the ASTER‐derived soil moisture. It islogical to conclude that duration and amount of precipitationhas a major control on soil moisture, as well the number ofdry days experienced since the last precipitation event(Table 1 and Figure 3). Otherwise, some of the dis-crepancies between the modeled soil moisture and theseASTER data may be related to differences in spatial andtemporal resolution. This suggests that whereas the H96data may be representative of the region, the dune field andunique surrounding aeolian environment probably responddifferently to hydrologic forcing.

4.4. Erosion Threshold Velocity Ratio

[36] The relationship between TI and the wind erosionthreshold velocity ratio (WTR) is not linear (Figure 2).However, the shape of the curve between the erosionthreshold velocity ratio and TI is similar and comparable tothose presented as a function of volumetric soil moisture[Fécan et al., 1999]. To observe the spatial pattern of ero-sion potential, the average TI data were recast into the dis-tribution of WTR values (Figure 8). The relationship variesfor different soil textures, and the erosion threshold velocityratios range between 1.0 and 2.0 for pure sand with soilmoisture between 0% and 5%. An erosion threshold velocityratio for gypsum sand has not previously reported in theliterature, but if it is similar to quartz sand, the modeledvalues range from 2.5 to 3.0, and correspond to modeled soilmoisture content of between 8% and 18% [Fécan et al.,1999]. The dunes and several areas of the playa surfacehave a wind threshold velocity ratio �2.5, indicating thatthese areas have the highest potential for aeolian transport orsediment availability relative to other areas of the dune field.The lower relative values of the WTR correspond to theposition of dune centers and/or crests compared to the highervalues found in the interdune. The variability of the averagewind velocity threshold ratio was observed along a transect(a – a′) oriented in the dominant wind direction (SW to NE)across a series of dunes (Figure 9). The dunes are easilyaccessible from the White Sands National Monument parkarea and access road. Several areas at the bottom of thealluvial fan adjacent to the dune field also have high sedi-ment availability. Other areas that have high sedimentavailability include an area west of the southern crescenticdunes, interpreted as sand mantling the surface suppliedfrom the eastern margin of Lake Lucero where gypsumevaporite is readily resupplied for transport. Interdune areasand the periodically wet playa surfaces appear to have theleast sediment availability from the averaged data, howeverduring dry conditions interdune areas may be activated inthe crescentic dune field (for example, during dry conditionsshown on 12 March 2008 in Figure 6g, and the windthreshold velocity ≈ 1.0–2.0).[37] The erosion susceptibility of the White Sands aeolian

system as derived by this approach was validated two days


13 of 23

after the last image pair in March 2008. On 14 March 2008,a large dust storm occurred and was captured by MODISAqua (Figure 10). Plumes from the playa and dune field canbe seen emanating from the area, inundating the city ofAlamogordo, NM and traveling over 200 km to the ENE. Atleast for this particular case, the ASTER‐derived WTRcorrespond very well and was shown to be a good predictorfor the subsequent dust emission event. Because of the highspatial resolution of the data, it can by hypothesized that thedust emissions originated from the aeolian features with lowWTR values, such as the active playas north of Lake Lucero,the edge of the alluvial fans to the west and the parabolicdunes as well. The low spatial resolution of MODIS makesit difficult to deduce the exact source of dust emissions fromfeatures imaged by ASTER. Later field observations on2 May 2008, when similar dry and windy conditions pre-vailed and aeolian processes were observed at multiplelocations at the White Sands aeolian system, provided someinsight into the probable surface conditions present at thattime of the large dust emission event. Wind velocities weremeasured and averaged between 10 to 15 mph, gusting togreater than 50 mph. Sand saltation was observed originat-ing from all dunes in all areas, the adjacent alluvial fans tothe west of active playas and the playas themselves. Severalpatches of friable surface crust composed of gypsum and siltwere observed to be eroded by wind from the playa surface,although specific dust emission sources could be seen onlyfrom a distance. The surface crusts observed on the playawere inflated (lower bulk density) from the lower, more

Figure 8. The unitless erosion threshold wind velocity(WTR) derived from the average ASTER TI image(Figure 7h). Areas that have a high susceptibility to aeo-lian erosion are shown in red, whereas areas in blue (higherTI) frequently have higher soil moisture and thus a lowersusceptibility to wind erosion. Note the higher WTR valuesfor the southern parabolic dunes as compared to the centralcrescentic dunes.

Figure 9. TheWTR plotted along the transect a − a′ (32.8103°N, 106.9364°W to 32.8143°N, 106.2619°W)across a series of dune and interdune areas at theWhite Sands NationalMonument access road. The image isa digital orthophoto quarter‐quad (DOQQ) acquired 4 January 2003 (available at http://online.wr.usgs.gov/ngpo/doq/MD‐DOQs.html). Arrows correspond to dune crest positions, where WTR values are lowerrelative to interdune areas.


14 of 23

moist soil conditions 2–3 cm below the upper surface. It isprobable that these types of surfaces appear as low TI andlow WTR in the ASTER‐derived images. The WTRparameter derived from ASTER does not distinguishbetween loose dry sand susceptible to saltation, a dustemission source or dry, low density, surface‐crusted sedi-ments. Under high wind conditions, it is likely that fine‐grained surface crust material (on dunes and playas alike)will break apart and become easily entrained by the wind togenerate saltating particles or dust emissions.

5. Discussion

[38] Aeolian sediment availability in the absence of veg-etation and stabilizing biotic soil crusts is highly dependenton the soil hydrology of any specific location. Evaporationrates are important in predicting soil moisture, which arestrongly influenced by wind. Several factors (i.e., precipi-tation and groundwater) affect the soil moisture in thisperched groundwater system beyond the influence of solarheating and evapotranspiration. The interdune and playa soilmoisture is affected by water table fluctuations due to var-iable precipitation anywhere within the basin. The WhiteSands aeolian system is situated between steep, high‐elevation mountain ranges and broad alluvial fans. Thisimplies that snowmelt or rainwater from a distance couldinfluence groundwater hydrology in the low‐lying basin.The subsurface variation in topography of the underlyingsabkha surface is also likely to affect the groundwater flowdynamics. A detailed examination of the groundwater con-ditions that impact soil moisture is beyond the scope of thispaper and has been previously explored by Langford et al.[2009], where groundwater at White Sands was found tocontrol dune morphology. Where local precipitation data ormodeled hydrology (i.e., H96 data) cannot explain wet soil

conditions through time, the influence of groundwaterfluctuations becomes a plausible scenario for explaininganomalously high soil moisture. In this case, the forcing ofprecipitation on soil moisture levels is apparent fromexamining the timing between precipitation events and sat-ellite data acquisitions, the number of dry days betweenprecipitation events and the total amount of precipitation.[39] Retrieval of the erosion threshold via TI modeling

from spaceborne thermal infrared remote sensing measure-ments has some limitations. The TI data from multiple datesare at the same scale of real thermal inertia units (TIUs) andwere comparable to each other in this study. Despite theefforts put forth to model real thermal inertia from remotesensing, retrievals from any satellite‐based system are stillan approximation and subject to the same limitations[Scheidt et al., 2008b]. The relative values of ATI could alsobe related to soil moisture and wind threshold velocity ratio;however, this study using ASTER builds on previous workusing the MODIS‐derived values of soil moisture frommodeled TI data. The real thermal inertia of the land surfaceis affected by the heterogeneity of material properties (i.e.,K, r, and c). The remote sensing TI retrievals, which shouldbe but are not always proportional to the real thermal inertia,are affected by composition (e.g., reflectance and emissivityvariation), topography and surface roughness, vegetationand variable atmospheric conditions. Specifically for theWhite Sands Dune Field where dry sand has been strippedaway, the algorithm may not detect the difference in bulkdensity between dry cemented dune surfaces and wet sedi-ment. These areas are small however, and occur seasonally.Based on this study, in order for the derivation of sedimentavailability from modeled TI data to be useful, it must beassumed that (1) all environmental effects are described orassumed to be negligible with respect to soil moisture,(2) soil moisture contributes a high degree of variability tothe relative values of TI retrieved by the algorithm, and(3) sediment availability is significantly controlled by soilmoisture. Some of the exceptions and important points(including temporal‐scale, subsurface moisture affects, soilcomposition and texture) for the interpretation of the dataare discussed below.[40] The approach presented works well for the overall

relative field conditions on the dates presented in this study,specifically for the main dune complex that consists ofcrescentic dunes and interdune areas (Figure 6). Areas out-side the main dune field, such as in the parabolic dune area,Alkali Flat and the alluvial fans to the west, raise someconcern. In these areas, erosion potential appears to be afunction of other factors in addition to soil moisture. Forexample, the water table in the parabolic dune area is lowerand does not exert a significant influence on soil moisture.The low values of WTR for these dunes suggest that theyare active, however this area is considered a stabilizedvegetation system and the dunes do not appear to be mobile[Kocurek and Lancaster, 1999] especially interdune sedi-ments stabilized by biotic soil crusts [Langford et al., 2009].These dunes may be quite dry and far less mobile thancrescentic dunes, and the sediment state of surfaces does notnecessarily reflect the state of the dune field as a whole orthe mobility of the dunes in general. Active sand transport

Figure 10. On 14 May 2008, two days after the period oflowest predicted WTR values in this study, a large duststorm was capture by MODIS Aqua from the White SandsDune aeolian system. This image shows the MODIS albedo.The image is stretched linearly to prevent saturated pixels inareas of gypsum ground cover and dust. This allows thevisualization of the dust plume structure emanating fromthe whole of the aeolian system. Winds are from the west‐southwest.


15 of 23

occurring from the sediment supply generated by LakeLucero and other active playas feed the central dune core aswell as the parabolic dunes. This sediment supply is avail-able only in pulses with the periodic dry down of the playaareas. The WTR results generated here appear to capturesand transport atop areas where stabilized interdune areasoccur in the parabolic dune field area. Because these areasare stabilized by biotic crusts, they cannot be interpreted assource areas of sediment, rather an indication of a sedimenttransport/bypass region, which may help to maintain para-bolic dunes. Alternatively, the WTR may simply indicatethat a high potential for aeolian erosion exists for thesedunes in the absence of vegetation, but the actual transportrate is far less. Alkali Flat is a hard gypsum‐cemented sur-face and likely not an erosive surface. However, it wasmodeled as having a low WTR during dry periods. Underthese conditions, this area will not be an area of sedimentaccumulation, rather an area of sediment transport/bypass.Alluvial fans may contain both vegetation and cemented (orarmored) surface characteristics. This is certainly true for theupper portion of these areas, but alluvial fans adjacent to theplayas have a sufficient supply of loose sand and silt sizedparticles available for transport by wind. As mentionedpreviously, saltation and dust emission were observed fromthese areas under dry windy conditions. In summary, cau-tion should be applied in the interpretation of the WTRresults because other factors can influence soil erosion.[41] The timing of the data acquisition is also an important

variable influencing the interpretation of the results and theircontext. Long periods of time exist between each of theimage dates and several wet and dry cycles have occurred.Each TI image in Figure 6 represents a 36 h time differencebetween day and night image acquisition, and we assumethe results are consistent with a 1/2 period (w/2 = 12 h) ofthe LST wave, ideal for determining a remote sensingderived retrieval of TI. Therefore, the data here reflectpseudo daily TI from different times of different years. Thistime scale is useful for examining the sediment states thatoccur at the White Sands Dune Field and playa aeoliansystem as a whole, as well as within the system. Eventhough presented as a time series representing 6 years, thesediment state and configuration of wet and dry surfacesvaries on several temporal scales. Both the daytime andnighttime satellite acquisitions occur at roughly the sametime of day, which allows for a direct comparison betweendifferent dates. Atmospheric conditions, such as the thermalstructure of the atmosphere (i.e., such as a temperatureinversion in early morning), vary through the day. Similarly,soil moisture in the upper cm of the land surface variesthroughout the day (e.g., precipitation, dew formation). Ifthe same dates are used, but images are acquired at differenttimes, the results would be different. Fortunately, the dayacquisition takes place well into the LST wave and the nightacquisition occurs early enough that surface moisture vari-ability from day to day is probably similar, favoring drierconditions.[42] The depth at which soil moisture affects the results of

the WTR is not known, but it is hypothesized to be sensitivethe only the upper few centimeters of sediment. The soilmoisture and the rate of temperature fluctuation at the sur-

face are partially driven by solar heating and cooling. Anassumption is made that the retrieved TI represents anintegration of soil moisture to some depth within subsurfacesoil, and we are observing the surface using optical remotesensing that is mostly sensitive to the uppermost sedimentsurface (∼100–200 mm). The amplitude of the daily LSTwave is expected to be greatest at the sediment surface,decreasing rapidly with depth [Sabol et al., 2006]. Heatingand cooling may occur rapidly on the time scale of minutesto hours, fluctuating with atmospheric humidity, the for-mation of dew, heat waves and wind. Pore space at thesediment surface will lose or retain moisture faster than thesubsurface soil, and in a wet aeolian system, it is notuncommon to find this soil moisture to be stable and per-sistent in both the playa and dune subsurface. What is notknown from the remote sensing measurements is the degreeto which the spatial variation of subsurface moisture cor-relates with the sediment surface. A reasonable assumptionwould be that the geomorphology, soil composition andcementation are significant factors in determining the soilmoisture depth profile. For example, the uppermost sedi-ment surface of a relatively wet playa surface will dry outquickly in the day during intense land surface heating,rapidly forming a friable surface evaporite or silt crust. Thesurface within a few centimeters may be quite erodiblewhere this occurs and therefore have a low wind thresholdvelocity. Subsurface soil moisture, which could lower theDT due to the higher heat capacity of the moist subsurface,would give a relatively higher TI and soil moisture retrieval,resulting in a falsely higher wind WTR with respect to thedry skin. Therefore, caution must be taken if applying theresults of these TI retrievals to the bulk of the dunes.However, this study was concerned with sediment avail-ability and dust emission events. These two factors would bemost concentrated in the uppermost sediment layer andtherefore the use of this technique may actually be moreappropriate and more accurate than other techniques (i.e.,microwave‐based approaches) which measure the deeperlevels of soil moisture.[43] The surface albedo may also fluctuate with soil

moisture on short time scales for soils of variable compo-sition, including a playa environment. For example, as theplaya surface dries and minerals precipitate, the albedo ofthe sediment may increase. Where soils are different in totalevaporite mineral content, the amount of highly reflectivecrust will also vary. Soils that have the same moisturecontent should have a similar thermal inertia, but the dif-ference in albedo independent of the thermal inertia willaffect the ATI calculation, lowering the TI retrieval for thehigher‐albedo soil surface. The albedo of a clay or silt richsoil may not fluctuate with soil moisture as much as gyp-sum‐, carbonate‐ and bicarbonate‐rich sediment. At WhiteSands, where gypsum forms as a highly reflective evaporitecrust, these areas can be expected to have relatively low TIretrievals and a lower WTR. However, the cohesion of thesoil is also an unknown parameter, and the remotely senseddata provide no means to detect how much this cohesioncontributes to the strength of the surface soil and the resis-tance to erosion. If the surface formation of high‐albedogypsum is indicative of a stable surface crust, the WTR will


16 of 23

be underestimated. Future and more detailed field observa-tions can determine soil strength properties and better cor-relate sources of erodible sediments with the orbital data.

6. Conclusions

[44] The approach presented here using ASTER‐derivedTI to estimate soil moisture and the wind threshold velocityratio (WTR) of an aeolian dune system at a high spatialresolution from optical remote sensing data is unique. Fur-thermore, the data are sensitive to the surfaces most directlyinvolved in saltation, sediment transport, and dust emissionevents. For the study of White Sands, small‐scale variabilityof soil moisture was retrieved without the use of sparseweather station data or the limitations of a hydrologicmodel. However, modeled soil moisture and precipitationdata sets were important in determining the hydrologicforcing that affects soil moisture, and to explain the eventsthat occur between satellite data acquisitions. These datahelped to partially explain variations in the modeled soilmoisture and correlated them with the timing of precipita-tion events. Trends in soil moisture were difficult to explainusing the monthly precipitation, but the daily precipitationdata provide a finer temporal scale from which to comparethe soil moisture maps. Dry down of surface sedimentsprobably occurs quickly, and soil moisture responds toprecipitation on the scale of days rather than months withrespect to these ASTER results. In wet aeolian systems suchas White Sands, soil moisture exerts an important control onincreasing the WTR, sediment availability and aeolian ero-sion. The results for White Sands represent the best possibleTI retrievals at this spatial resolution, and local weather dataverified that normal diurnal cycles occurred with little to noclouds, precipitation or frontal activity interfering with theDT estimation of each image date. Based on previousstudies and field observations at White Sands, spatial andtemporal variation of TI are strongly affected by soilmoisture, which is in turn affected by precipitation. Themodel for retrieving TI and, subsequently, soil moisture andthe WTR, has limitations and is in need of further refine-ment using ground truth and accuracy assessment at thespatial resolution of the ASTER instrument. Overall soilmoisture retrievals (10–25%) were higher than expected forarid lands and dunes (0–10%), reinforcing the need forcalibration of the model, incorporating field measurementsof soil density, volumetric soil moisture, and soil composi-tion (e.g., percent clay). However, these higher percentagesdid not impact the spatial patterns or the overall calculationof the sediment availability.[45] The potential exists to use this approach to estimate

sediment availability and soil moisture in other playa, dune,and sparsely vegetated environments (e.g., fallow agricul-tural fields), especially if these regions are influenced bysubsurface water. A need also exists to conduct these studiesfor different soil compositions with different albedo, asWhite Sands has uniquely high albedo. Where small‐scalespatial variability is less important, such as the large dunesystems of the Sahara, MODIS can be leveraged to examineaeolian sediment availability at a lower spatial resolution buthigher temporal resolution. This study demonstrates a viableapproach to soil moisture mapping that may be conductedwith ASTER, MODIS or future VSWIR‐TIR instruments

such as the Hyperspectral Infrared Imager (HyspIRI). Thespatial resolution of HyspIRI is planned to be 60m for all thewavelength regions and the temporal resolution for the TIRwill be as low as 5 days, which will provide much betterspatial resolution for ATI studies. Additionally, apparentthermal inertia modeling using these instruments wouldcomplement future high spatial resolution SMAP retrievals ofsoil moisture, vegetation and surface roughness of arid lands.

Appendix A: Derivation of the Wind ThresholdVelocity Ratio From Apparent Thermal InertiaA1. Remote Sensing Data of White Sands Dune Field

[46] This appendix covers import details about the ASTERand MODIS data sets used, the application of the modelsused to derive thermal inertia, soil moisture and the windthreshold velocity ratio (WTR).[47] Several studies have compared ASTER and MODIS

data products, such as retrievals of reflectance and radiance[Miura et al., 2008], emissivity and albedo [Zhou et al.,2003], and emissivity [Jacob et al., 2004] and temperature[Liu et al., 2007]. Good agreement was found betweenMODIS and ASTER temperatures [Jacob et al., 2004]. Theresulting MODIS and ASTER reflectance values and veg-etation indices compared well in the radiometric compati-bility study of Miura et al. [2008]; however, (1) the valuescompared were resampled to a spatial resolution of 5 km and(2) the study concluded that the method of the atmosphericcorrection for ASTER reduced the overall quality of thestandard reflectance product. A detailed discussion of thedifferent atmospheric correction methods used to generatestandard reflectance products is given by Miura et al.[2008]. The radiative transfer code and approach to cor-recting scattering terms are different, as well as the sourcesof ozone and water vapor, and the lack of an aerosols cor-rection for the ASTER reflectance product. The basic dif-ferences between the instruments relevant to this study aresummaries in Table A1, and the separate treatment of tem-perature and reflectance data of the two instruments arediscussed.

A1.1. Temperature Retrieval

[48] The LST and emissivity product (MOD11_L2) pro-vided temperature retrievals without complete spatial cov-erage of the White Sands Dune Field (Figure A1, left). Thisprocessing artifact is due to a false detection of cloud pixelsbecause of the highly reflective gypsum composition of theplaya and dune surface. A similar effect occurs over areas ofmountain snow in the Sacramento Mountains just east of thedune field. Day and night land surface temperatures wereretrieved using the emissivity normalization method[Realmuto, 1990; Kahle and Alley, 1992] from the calibratedtop of atmosphere (TOA) thermal radiance from bands 29,30, and 31 of the MOD021KM data product. Approaches tothe accurate geolocation of MODIS data to achieve evensubpixel accuracy are discussed by Wolfe et al. [2002], butthe geographic coregistration between MODIS day andnight temperature data for White Sands was found to beaccurate at the 1 km spatial resolution. The AST_09T TIRstandard data product has been radiometrically, geometri-cally, and atmospherically corrected [Thome et al., 1998].Coregistration between day and night ASTER data was


17 of 23

found to be poor [Fujisada et al., 2005], with a spatial offsetof similar land features by as much as 500 m in some scenes.The nighttime ASTER TIR data’s coregistration accuracywas quality checked. Errors were corrected by coregistering

the data to daytime data with an accuracy of at least half ofan ASTER TIR pixel (±45 m). Day and night land surfacetemperatures (LST) were extracted from ASTER TIR datausing the emissivity normalization method [Realmuto, 1990;

Table A1. Details and Characteristics of the ASTER and MODIS Instruments and Data Products Used in This Study Are Compareda

Instrument Characteristic ASTER MODIS

Swath width 60 km 2330 kmScanner push broom (VNIR/SWIR); whisk broom (TIR) whisk broomSpatial resolution at nadir VNIR 15 m VNIR 250 m

SWIR 30 m SWIR 500 mTIR 90 m TIR 1000 m

Radiometric resolution VNIR: NEDr � 0.5% VNIR: 57 � SNR � 1087 DNb

SWIR: NEDr � 0.5–1.5% SWIR: NEDT = 0.05–0.25 KTIR: NEDT � 0.3 K TIR: NEDT = 0.25–0.05 K

Geolocation accuracy ±15 m ±50 mRadiometric accuracy 4% 2%Radiative transfer code MODTRAN 6S codeOzone NCEP TOVS NASA TOMSWater vapor NCEP GDAS MODIS water vaporAerosol No correction; r = 0 MODIS aerosolsAtmospheric correction accuracy 14% for r < 0.1; 7% for r > 0.1 5–9% for clear to high rSpectral albedo product N/A MOD43 (V5)Reflectance product AST07XT MOD09 (V5)VNIR band passes used x1 (0.52–0.60 mm) x1 (0.62–0.67 mm)

x3 (0.78–0.86 mm) x2 (0.84–0.87 mm)x3 (0.46–0.48 mm)x4 (0.54–0.56 mm)

SWIR band passes x5 (2.15–2.18 mm) x5 (1.23–1.25 mm)x6 (2.18–2.22 mm) x7 (2.11–2.15 mm)x8 (2.29–2.36 mm)x9 (2.36–2.43 mm)

Thermal radiance product AST09T MOD021KM (V5)TIR band passes used x10 (8.125–8.475 mm) x29 (8.400–8.700 mm)

x11 (8.475–8.825 mm) x31 (10.78–11.28 mm)x12 (8.925–9.275 mm) x32 (11.77–12.27 mm)x13 (10.25–10.95 mm)x14 (10.95–11.65 mm)

aTable after Miura et al. [2008].bSNR, signal‐to‐noise ratio.

Figure A1. (left) The standard MODIS product, MOD43 broadband albedo, has missing data in theWhite Sands study area because of the quality constraints used to create the cloud‐free data product.The area of missing data retrievals from the standard albedo product is indicated on the map with a solidline, whereas the hatched areas represent missing temperature retrievals from the MOD11 standard LSTand emissivity product. The gray inset box indicates subset used for Figures 5–8 in the main text. (right)ASTER albedo data were scaled to MODIS data: the frequency distributions of MODIS reflectance fromthe MOD09 product (solid line), ASTER reflectance from the AST07XT product (short dashed line), andthe resulting scaled ASTER reflectance (long dashed line with asterisk).


18 of 23

Kahle and Alley, 1992]. The temperature difference (DT)was calculated for each of the day/night image pairs fromboth the MODIS and ASTER data according to the TImodel described below. The TIR bands used from theASTER and MODIS satellites are compared in Table A1.

A1.2. Broadband Albedo Estimation

[49] The preferred data input for narrowband to broad-band albedo determination is the MOD43 spectral albedoproduct [Liang, 2001]. The 1 km spatial resolution spectraland broadband albedo data from MODIS is generated fromcloud‐free, atmospherically corrected, multiangle reflec-tance data over a sixteen day period [Schaaf et al., 2002].However, for all daytime satellite overpasses in the WhiteSands Dune Field area, spatial coverage of the albedoretrievals is limited, leaving gaps in the most critical area ofthe study (Figure A1, left). These data were falsely flaggedas cloud by the standard processing algorithm and ignored.In version 5 of the MODIS data, retrieval only occurs ifthere are at least three noncloud acquisitions for a givenlocation. The processing constraints on the MOD43 spectralalbedo product are intended to produce a high‐quality,cloud‐free product, and these constraints can be relaxed inorder to generate data over bright targets that would nor-mally be classified as cloud (C. B. Schaaf, personal com-munication, 2009). However, even with a relaxation of thesequality constraints allowing for the retrieval of albedovalues, these data would be averaged over multiple datesand viewing angles over a several day period. Consequently,because of the time constraint needed for Dt in the TImodel, the accuracy of the TI retrieval would be adverselyaffected. TI retrieval is dependent on a time constraint of36 h for these ASTER data (and potentially 12 h forMODIS) in order to limit the temperature change to only theeffects of diurnal solar heating and cooling. In order to avertthis issue, spectral reflectance (MOD09) can be used if thesurface is assumed to be Lambertian for small viewing andsolar angles (S. Liang, personal communication, 2009).Therefore, the broadband albedo, a, was calculated directlyfrom the MOD09 reflectance values using the approach ofLiang [2001]:

a ¼ 0:160 x1ð Þ þ 0:291 x2ð Þ þ 0:243 x3ð Þ þ 0:116 x4ð Þþ 0:112 x5ð Þ � 0:081 x7ð Þ � 0:015 ðA1Þ

where xn is the reflectance value of spectral band n(Table A1).[50] For ASTER, the AST_07XT standard surface

reflectance product includes bands from the VNIR andcross‐talk corrected SWIR spectral region [Iwasaki et al.,2002; Tonooka and Iwasaki, 2003; Iwasaki and Tonooka,2005]. Assuming a Lambertian surface, the broadbandalbedo, a, was calculated using the methodology of Liang[2001] directly from the AST_07XT values:

a ¼ 0:484 x1ð Þ þ 0:335 x3ð Þ þ 0:324 x5ð Þ þ 0:551 x6ð Þþ 0:305 x8ð Þ � 0:367 x9ð Þ � 0:015 ðA2Þ

where xn is the reflectance value of spectral band n(Table A1).[51] This method of calculating broadband albedo would

not be possible for ASTER data after 23 April 2008.

Beginning in May 2007, ASTER SWIR data was observedto have anomalous saturation of values in bands 5 through 9due to increasing SWIR detector temperature which reducesthe dynamic range of the data. This is a potential problemfor bright desert scenes such as the White Sands Dune Field.The data acquired for this study were reprocessed with thecurrent radiometric correction, and no saturated pixels oranomalous striping were observed. The last date of the dataused in this study was 10 March 2008, where data after23 April 2008 exhibited saturated values in the SWIR bandsdue to the rapid degradation of the detectors. Data afterApril 2008 were determined to be unusable; ASTER SWIRdetectors were no longer functioning after 13 January 2009.[52] Preliminary analysis revealed differences between the

broadband albedo calculated from MODIS and ASTER,even though these values should be similar because the dataare acquired at approximately the same time from the samesatellite. The maximum and average values of broadbandalbedo of the White Sands Dune Field calculated fromASTER data were consistently higher than those calculatedfrom MODIS data. In some isolated areas of the very brightplaya surface (less than 100 pixels), the albedo was greaterthan 1.0. These affects are due to the original data and notthe broadband calculations. Where similar bands werecompared from common areas at White Sands, theAST07XT data were higher than the MOD09 reflectancedata for the White Sands data. The mean differences werebetween 10% and 30% for TOA radiance and 30%–70% forcorrected reflectance values. Miura et al. [2008] found amean difference of ∼3% between ASTER and MODIS redand near‐infrared reflectance (NIR) bands at 5 km spatialresolution. Some of this variability was expected due to thedifferences between the ASTER and MODIS spatialresolutions and atmospheric correction techniques. Withlower spatial resolution, dynamic range was also expected tobe lower because less mixing of scene components in eachpixel occurs in data from an instrument of higher spatialresolution. The difference in broadband albedo between thetwo instruments was also likely due to the result of differentatmospheric correction schemes. ASTER broadband albedowas scaled to the MODIS broadband albedo for this study.This linear scaling shifted the peaks of the data (a bimodalfrequency distribution) into agreement (Figure A1, right).After scaling, the ASTER broadband albedo data had thesame mean but still had a larger dynamic range (20%) thanMODIS data, which was expected for an instrument ofhigher spatial resolution.

A2. Thermal Inertia Model

[53] Xue and Cracknell [1995] presented a simplified TImodel that can be used for areas of variable soil moisture.They also present a sensitivity analysis to input parametersand an example of its operational use to derive ATI and anapproximation of TI from AVHRR data, utilizing phaseangle information and the magnitude of the diurnal tem-perature change [Xue and Cracknell 1995]. A thoroughdescription of the model’s applications utilizing MODISdata can be found in the work of Cai et al. [2005, 2007a].The model is shown below for clarity of the method used inthis paper and to correct discrepancies between previouslypublished versions of the equations of Cai et al. [2005,


19 of 23

2007a]. From Xue and Cracknell [1995] thermal inertia (P)is defined as

P ¼ ð1� aÞS0Ct

�Tffiffiffi!

p A1 cos !t2 � �1ð Þ � cos !t1 � �1ð Þ½ �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ 1

bþ 1

2b2

r8>><>>:

þ A2 cos !t2 � �2ð Þ � cos !t1 � �2ð Þ½ �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2þ

ffiffiffi2

p

bþ 1

2b2

r9>>=>>;

ðA3Þ

where S0 is the solar constant (W m−2), Ct is the atmospherictransmittance (assumed for this study to be 0.75), w isEarth’s angular frequency, a is the broadband albedoderived from satellite image data, and b is dependent on thetime of maximum daytime temperature tmax estimated fromthe nearest ground based weather station at the White SandsDune Field:

b ¼ tan !tmaxð Þ1� tan !tmaxð Þ ðA4Þ

Phase difference terms are calculated from b as

�1 ¼ arctanb

1þ b

� �ðA5Þ

�2 ¼ arctanb

ffiffiffi2

p

1þ bffiffiffi2

p� �

ðA6Þ

The coefficient of Fourier series (An) is calculated as

An ¼ 2 sin � sin�

�sinðnyÞ þ 2 cos � cos�

� n2 � 1ð Þ� n sinðnyÞ cosy� cosðnyÞ siny½ � ðA7Þ

where d is the solar declination, a is the latitude, and

y ¼ arccos tan � tan�ð Þ ðA8Þ

t1 and t2 are the times of maximum and minimum tem-peratures, typically considered to be 1400 and 0200 localtime. However, the satellite overpass times obtained fromsatellite metadata are at ≈1100 and ≈2210 local time,respectively. The general diurnal temperature cycle of theland surface is driven by insolation, and an accurate esti-mation of the LST temperature wave was needed in order topredict the maximum and minimum temperatures at othertimes. The shape of the LST wave has been described ac-cording to a single cosine function, followed by a period ofcooling where the shape of the temperature curve is modi-fied by an exponential decay in temperature:

TðtÞ ¼T0 þ Ta cos

�

!t � tmaxð Þ

� �; t < ts

T0 þ Ta cos�

!t � tmaxð Þ

� �e�

t�tsk ; t � ts

8><>: ðA9Þ

where T0 is the initial morning temperature, Ta is the diurnaltemperature amplitude, ts is the start time of the attenuationfunction, and k is the attenuation constant [Schädlich et al.,2001]. Improvements have been made in fitting remotesensing data to this type of model [van den Bergh et al.,2007], but only the simple cosine model was used todetermine the actual maximum (Tmax) and minimum (Tmin)temperatures of Cai et al. [2005]. Therefore, the diurnaltemperature difference (DT) was determined from the day(T1) and night (T2) time temperature ASTER images usingTmax − Tmin, where

Tmax ¼ T1 þ T1 � T2ð Þ cos !tmaxð Þ � cos !t1ð Þ½ �cos !t1ð Þ � cos !t2ð Þ

Tmin ¼ T2 þ T1 � T2ð Þ cos !tminð Þ � cos !t2ð Þ½ �cos !t1ð Þ � cos !t2ð Þ

ðA10Þ

T1 − T2 is the temperature difference between the 1100 and2210 satellite overpass times. T1 − T2 is always less thanTmax − Tmin in the case of these ASTER data, and wouldresult in an underestimation of TI if not scaled usingequation (A10).[54] Applying this model to remote sensing image data

results in TI maps in thermal inertia units (TIU) and a spatialresolution equal to that of the input data used. The Xue andCracknell [1995] model was modified by Zhenhua andYingshi [2006]. They describe the model as more accurate,needing only the daily maximum temperature instead of theday‐night temperature difference and the model accounts forsensible and latent heat [Zhenhua and Yingshi, 2006].Comparison of the soil moisture retrievals for both modelsusing MODIS data showed similar results to actual soilmoisture measurements of Zhenhua and Yingshi [2006]. Thedata produced in this paper are reported and discussed asapproximated TI data, which represents ATI at the scale andunits of real thermal inertia (P).

A3. Estimation of Erosion Threshold Velocity RatioFrom Soil Moisture

[55] Soil moisture of bare or sparsely vegetated surfacesas a function of thermal inertia is given by Ma and Xue[1990]:

P ¼ 2:1d1:2�0:2dsd�

�

� �s e

�0:007 wdsd��20

� �2� �

þ d0:8þ0:2dsd�

�

� �s

264

375

1=2

�0:2�

d�d2s

0:001

ffiffiffiffiffiffiffiffi100

p2664

3775 ðA11Þ

where ds is the soil density, d� is the water density, and � isthe volumetric soil moisture. Using this relationship, lookuptables were generated for soil moisture as a function ofdensity and thermal inertia. Because this model is used in thecontext of remote sensing, density must be assumed. Thevariation of density in the model between 2.30 and 2.65 g cm−3

results in slightly different but similar relationships between


20 of 23

soil moisture and TI (see main text, Figure 2a). For arid landswhere the actual soil moisture is low (0∼15%), the density isassumed not to significantly affect soil moisture retrievals.For the data analysis here, a density of 2.65 g cm−3 wasassumed for the calculation of the TI.[56] An operational parameterization of the wind erosion

threshold as a function of soil moisture for semiarid soils isgiven by Fécan et al. [1999]:

u*t�

u*td

¼ 1; for � < �0

u*t�

u*td

¼ 1þ 1:21 �� 0ð Þ0:68h i0:5

; for � > �0 ðA12Þ

where � is the percentage of volumetric soil moisture, �′ isthe absorbed soil water percentage dependent on clay con-tent, and u*t�/u*td is the ratio of the wet to dry thresholdwind velocity or WTR. With this relationship between soilmoisture and WTR, the lookup table defined by equation(A11) was refined to determine WTR directly as a func-tion of TI (see main text, Figure 2b).

Notation

An Fourier series coefficient.a latitude of pixel.a albedo, unitless.

ATI apparent thermal inertia.Ct atmospheric transmittance.c specific heat, J kg−1 K−1.ds soil density, kg m−3.d� water density, kg m−3.r density, kg m−3.K thermal conductivity, W m−2 K−1.k attenuation constant.P thermal inertia, J m−2 s−1/2.TI modeled thermal inertia, J m−2 s−1/2 (or TIU).p precipitation, mm.S0 solar constant, W m−2.T0 initial temperature, K.Ta diurnal temperature amplitude, K.T1 day temperature at t1, K.T2 night temperature at t2, K.Tn night atmospheric temperature, K.Td day atmospheric temperature, K.

Tmax estimated maximum day temperature, K.Tmin estimated minimum night temperature, K.DT brightness temperature difference, K.t1 day satellite overpass time.t2 night satellite overpass time.Dt time difference between satellite overpasses.tmax time of maximum daytime temperature.tmin time of minimum nighttime temperature.u*t� wet soil threshold wind velocity (m s−1).u*td dry soil threshold wind velocity (m s−1).dn phase difference terms.d solar declination of pixel.� volumetric soil moisture, %.

�′ absorbed soil water, %.xn reflectance or emissivity value of spectral band n.w angular frequency of Earth.

[57] Acknowledgments. The authors wish to express their gratitudeto Crystal Schaaf, Shunlin Liang, and Guoyin Cai for their helpful com-ments on this work, and the assistance of staff at the White Sands NationalMonument, especially David Bustos. The quality of this manuscript wasgreatly improved by the thorough and helpful reviews by anonymous re-viewers. Funding for this project was provided through the NASA SolidEarth and Natural Hazards Program (NAG5–13730), the ASTER scienceteam (NNG04‐GO69G), as well as the Earth and Space Science Fellowship(NESSF) Program (NNX06‐AF92H). Aerial photography data were madeavailable and downloaded from http://seamless.usgs.gov; MODIS andASTER data were obtained through https://lpdaac.usgs.gov/; climate andsoil moisture data were acquired from http://www.cdc.noaa.gov andhttp://www.wrcc.dri.edu/.

ReferencesAbrams, M. (2000), The Advanced Spaceborne Thermal Emission andReflectance Radiometer (ASTER): Data products for the high spatial res-olution imager on NASA’s Terra platform, Int. J. Remote Sens., 21(5),847–859, doi:10.1080/014311600210326.

Allmendinger, R. J. (1972), Hydrologic control over the origin of gypsum atLake Lucero, White Sands National Monument, New Mexico, MS thesis,182 pp., N. M. Inst. of Min. and Technol., Socorro.

Allmendinger, R. J., and F. B. Titus (1973), Regional hydrology and evap-orative discharge as a present‐day source of gypsum at White SandsNational Monument, New Mexico, N. M. Bur. Mines and Miner. Resour.Open File Rep. OF‐55, 53 pp., Socorro.

Belly, P. Y. (1964), Sand movement by wind, Tech. Memo. 1, 24 pp., U.S.Army Corps of Eng., Coastal Eng. Res. Cent., Washington, D. C.

Bindlish, R., T. J. Jackson, E. Wood, H. Gao, P. Starks, D. Bosch, andV. Lakshmi (2003), Soil moisture estimates from TRMM MicrowaveImager observations over the southern United States, Remote Sens.Environ., 85, 507–515, doi:10.1016/S0034-4257(03)00052-X.

Bindlish, R., T. J. Jackson, A. J. Gasiewski, M. Klein, and E. Njoku (2006),Soil moisture mapping and AMSR‐E validation using the PSR inSMEX02, Remote Sens. Environ., 103, 127–139, doi:10.1016/j.rse.2005.02.003.

Cai, G., J. Wu, Y. Xue, Y. Hu, J. Guo, and J. Tang (2005), Soil moistureretrieval from MODIS data in northern china plain using thermal inertiamodel (SoA‐TI), in 2005 IEEE Geoscience and Remote Sensing Sym-posium Proceedings, pp. 4501–4504, Inst. of Electr. and Electr. Eng.,Piscataway, N. J.

Cai, G., Y. Cue, Y. Hu, Y.Wang, J. Guo, Y. Luo, C.Wu, S. Zhong, and S. Qi(2007a), Soil moisture retrieval fromMODIS data in northern China Plainusing thermal inertia model, Int. J. Remote Sens., 28(16), 3567–3581,doi:10.1080/01431160601034886.

Cai, G., J. Wu, Y. Xue, W. Wan, and X. Huang (2007b), Oil spill detectionfrom thermal anomaly using ASTER data in Yinggehai of Hainan, China,in 2007 IEEE Geoscience and Remote Sensing Symposium, 23–27 July,2007, Barcelona, Spain, pp. 898–900, Inst. of Electr. and Electr. Eng.,Piscataway, N. J.

Chen, W., Z. Dong, Z. Li, and Z. Yang (1996), Wind tunnel test of theinfluence of moisture on the erodibility of loessial sandy loam soils bywind, J. Arid Environ., 34, 391–402.

Chepil, W. S. (1956), Influence of moisture on erodibility of soil by wind,Soil Sci. Soc. Am. Proc., 20, 288–292.

Coolbaugh, M. F., C. Kratt, A. Fallacaro, W. M. Calvin, and J. V. Taranik(2007), Detection of geothermal anomalies using Advanced SpaceborneThermal Emission and Reflection Radiometer (ASTER) thermal infraredimages at Bradys Hot Springs, Nevada, USA, Remote Sens. Environ.,106, 350–359, doi:10.1016/j.rse.2006.09.001.

Cornelis, W. M., and D. Gabriels (2003), The effect of surface moisture onthe entrainment of dune sand by wind: An evaluation of selected models,Sedimentology, 50, 771–790, doi:1046/j.1365-3091.2003.00577.x.

Crabaugh, M. M. (1994), Controls on accumulation in modern and ancientwet eolian systems, Ph.D. dissertation, 135 pp., Univ. of Tex., Austin.

Cracknell, A. P., and Y. Xue (1996), Thermal inertia determination fromspace‐a tutorial review, Int. J. Remote Sens., 17(3), 431–461, doi:10.1080/01431169608949020.

Entekhabi, D., T. J. Jackson, E. Njoku, P. O’Neill, and J. Entin (2008), SoilMoisture Active/Passive (SMAP) mission concept, Proc. SPIE, 7085,70850H, doi:10.1117/12.795910.

Fan, Y., C. J. Duffy, and D. S. Oliver (1997), Density‐driven groundwaterflow in closed desert basins: Field investigations and numerical experiments,J. Hydrol., 196(1–4), 139–184, doi:10.1016/S0022-1694(96)03292-1.


21 of 23

Fécan, F., B. Marticorena, and G. Bergametti (1999), Parameterization ofthe increase of the aeolian erosion threshold wind friction velocity dueto soil moisture for arid and semi arid areas, Ann. Geophys., 17(1),149–157.

Federal, M. D. A. (2004), Landsat GeoCover ETM+ 2000 Edition MosaicsTile N‐03–05.ETM‐EarthSat‐MrSID, 1.0, USGS, Sioux Falls, SouthDakota.

Fujisada, H., G. B. Bailey, G. G. Kelly, S. Hara, M. J. Abrams (2005),ASTER geometric performance, IEEE Trans. Geosci. Remote Sens.,43, 2707–2714, doi:10.1109/TGRS.2005.847924.

Gillespie, A. R., and A. B. Kahle (1977), Construction and interpretation ofa digital thermal inertia image, Photogramm. Eng. Remote Sens., 43(8),983–1000.

Greeley, R., and J. D. Iversen (1985),Wind as a Geologic Process: On Earth,Mars, Venus and Titan, 348 pp., Cambridge Univ. Press, Cambridge, U. K.

Hotta, S., S. Kubota, S. Katori, and K. Horikawa (1984), Sand transport bywind on a wet sand surface, in Coastal Engineering ‐ Nineteenth CoastalEngineering Conference: Proceedings of the International Conference,Sept. 3–7, 1984, Houston, TX, edited by B. L. Edge, pp. 1265–1281,Am. Soc. Civ. Eng., New York.

Huang, J., H. van den Dool, and K. P. Georgarakos (1996), Analysis ofmodel‐calculated soil moisture over the United States (1931–1993) andapplications to long‐range temperature forecasts, J. Clim., 9(6), 1350–1362, doi:10.1175/1520-0442(1996)009<1350:AOMCSM>2.0.CO;2.

Huang, M.‐F., J. Yu, S. Hu, and M. Jing (2006), Study on evapotranspira-tion estimation of small drainage based on ASTER data, in 2006 IEEEGeoscience and Remote Sensing Symposium Proceedings, pp. 696–3699, Inst. of Electr. and Electr. Eng., Piscataway, N. J.

Iwasaki, A., and H. Tonooka (2005), Validation of a crosstalk correctionalgorithm for ASTER/SWIR, IEEE Trans. Geosci. Remote Sens., 43,2747–2751, doi:10.1109/TGRS.2005.855066.

Iwasaki, A., H. Fujisada, H. Akao, O. Shindou, and S. Akagi (2002),Enhancement of spectral separation performance for ASTER/SWIR,Proc. SPIE, 4486, 42–50, doi:10.1117/12.455140.

Jackson, T. J. (1993), Measuring surface soil moisture using passive micro-wave remote sensing, Hydrol. Processes, 7, 139–152, doi:10.1002/hyp.3360070205.

Jackson, T. J. (1997), Soil moisture estimation using SSM/I satellite data aover grassland region, Water Resour. Res., 33, 1475–1484, doi:10.1029/97WR00661.

Jacob, F., F. Petitcolin, T. Schmugge, E. Vermote, A. French, and K. Ogawa(2004), Comparison of land surface emissivity and radiometric tempera-ture derived from MODIS and ASTER sensors, Remote Sens. Environ.,90, 137–152, doi:10.1016/j.rse.2003.11.015.

Kahle, A. B. (1977), A simple thermal model of the Earth’s surface for geo-logic mapping by remote sensing, J. Geophys. Res., 82, 1673–1680,doi:10.1029/JB082i011p01673.

Kahle, A. B., and R. E. Alley (1992), Separation of temperature and emit-tance in remotely sensed radiance measurements, Remote Sens. Environ.,42, 107–111, doi:10.1016/0034-4257(92)90093-Y.

Kahle, A. B., A. R. Gillespie, and A. F. H. Goetz (1976), Thermal inertiaimaging: A new geologic mapping tool, Geophys. Res. Lett., 3(1), 26–28,doi:10.1029/GL003i001p00026.

Katra, I., and N. Lancaster (2008), Surface‐sediment dynamics in a dustsource from spaceborne multispectral thermal infrared data, Remote Sens.Environ., 112, 3212–3221, doi:10.1016/j.rse.2008.03.016.

Katra, I., S. Scheidt, and N. Lancaster (2009), Changes in active eolian sandat northern Coachella Valley, California, Geomorphology, 105(3–4),277–290, doi:10.1016/j.geomorph.2008.10.004.

Kocurek, G., and K. G. Havholm (1994), Eolian sequence stratigraphy ‐ aconceptual framework, in Siliclastic Sequence Stratigraphy, edited byP. Weimer and H. W. Posamentier, AAPG Mem. 58, 393–409.

Kocurek, G., and N. Lancaster (1999), Aeolian system sediment state:Theory and Mojave Desert Kelso dune field example, Sedimentology,46, 505–515, doi:10.1046/j.1365-3091.1999.00227.x.

Kocurek, G., M. Carr, R. Ewing, K. G. Havholm, Y. C. Nagar, and A. K.Singhvi (2007), White Sands Dune Field, New Mexico: Age, dunedynamics and recent accumulations, Sediment. Geol., 197(3–4), 313–331, doi:10.1016/j.sedgeo.2006.10.006.

Langer, A. M., and P. F. Kerr (1966), Mojave playa crusts: Physical prop-erties and mineral content, J. Sediment. Petrol., 36(2), 377–396.

Langford, R. P. (2003), The Holocene history of the White Sands DuneField and influences on eolian deflation and playa lakes, QuaternaryInt., 104(1), 31–39, doi:10.1016/S1040-6182(02)00133-7.

Langford, R. P., J. M. Rose, and D. E. White (2009), Groundwater salinityas a control on development of eolian landscape: An example from theWhite Sands of New Mexico, Geomorphology , 105 , 39–49,doi:10.1016/j.geomorph.2008.01.020.

Liang, S. (2001), Narrowband to broadband conversions of land surfacealbedo I: Algorithms, Remote Sens. Environ., 76, 213–238, doi:10.1016/S0034-4257(00)00205-4.

Liu, Y., Y. Yamaguchi, and C. Ke (2007), Reducing the discrepancybetween ASTER and MODIS land surface temperature products, Sensors,7(12), 3043–3057, doi:10.3390/s7123043.

Ma, A. N., and Y. Xue (1990), A study of remote sensing informationmodel of soil moisture, paper presented at 11th Asian Conference onRemote Sensing, State Sci. and Technol. Com. of China, Beijing,15–21 Nov.

Marticorena, B., and G. Bergametti (1995), Modeling the atmospheric dustcycle: 1. Design of a soil‐derived dust emission scheme, J. Geophys.Res., 100, 16,415–16,430, doi:10.1029/95JD00690.

McKee, E. D. (1966), Structures of dunes at White Sands National Monu-ment, New Mexico (and a comparison with structures of dunes fromother selected areas), Sedimentology, 7, 3–69, doi:10.1111/j.1365-3091.1966.tb01579.x.

McKenna‐Neuman, C., and W. G. Nickling (1989), A theoretical and windtunnel investigation of the effect of capillarity water on the entrainmentof sediment by wind, Can. J. Soil Sci., 69, 79–96.

Miura, T., H. Yoshioka, K. Fujiwara, and H. Yamamoto (2008), Inter‐comparison of ASTER and MODIS surface reflectance and vegetationindex products for synergistic applications to natural resource monitor-ing, Sensors, 8(4), 2480–2499, doi:10.3390/s8042480.

Nasipuri, P., T. J. Majumdar, and D. S. Mitra (2006), Study of high‐resolution thermal inertia over western India oil fields using ASTER data,Acta Astronaut., 58(5), 270–278, doi:10.1016/j.actaastro.2005.11.002.

Nickling, W. G., and C. McKenna‐Neuman (1994), Aeolian sedimenttransport, in Geomorphology of Desert Environments, edited by A. J.Parsons and A. D. Abrahams, pp. 517–556, Springer, London.

Nickovich, S., A. Papdopoulos, O. Kakaliagou, and G. Kallos (2001),Model for prediction of desert dust cycles in the atmosphere, J. Geophys.Res., 106, 18,103–18,130, doi:10.1029/2000JD900794.

Njoku, E. G., T. J. Jackson, V. Lakshmi, T. K. Chan, and S. V. Nghiem(2003), Soil moisture retrieval from AMSR‐E, IEEE T, IEEE Trans.Geosci. Remote Sens., 41, 215–229, doi:10.1109/TGRS.2002.808243.

Peŕez, C., S. Nickovic, J. M. Baldasano, M. Sicard, F. Rocadenbosch, andV. E. Cachorro (2006a), A long Saharan dust event over the westMediterranean: Lidar, Sun photometer observations, and regional dustmodeling, J. Geophys. Res., 111, D15214, doi:10.1029/2005JD006579.

Peŕez, C., S. Nickovic, G. Pejanovic, J. M. Baldasano, and E. Ōzsoy(2006b), Interactive dust‐radiation modeling: A step to improve weatherforecasts, J. Geophys. Res., 111, D16206, doi:10.1029/2005JD006717.

Pieri, D. C., andM. J. Abrams (2004), ASTERwatches the world’s volcanoes:A new paradigm for volcanological observations from orbit, J. Volcanol.Geotherm. Res., 135(1–2), 13–28, doi:10.1016/j.jvolgeores.2003.12.018.

Pratt, D. A., and C. D. Ellyett (1979), The thermal inertia approach to map-ping soil moisture and geology, Remote Sens. Environ., 8, 151–168,doi:10.1016/0034-4257(79)90014-2.

Price, J. C. (1977), Thermal inertia mapping: A new view of Earth, J. Geo-phys. Res., 82, 2582–2590, doi:10.1029/JC082i018p02582.

Price, J. C. (1980), The potential of remotely sensed thermal infrared datato infer surface soil moisture and evapotranspiration, Water Resour. Res.,16, 787–795, doi:10.1029/WR016i004p00787.

Price, J. C. (1985), On the analysis of thermal infrared imagery: The limitedutility of apparent thermal inertia, Remote Sens. Environ., 18, 59–73,doi:10.1016/0034-4257(85)90038-0.

Realmuto, V. (1990), Separating the effects of temperature and emissivity:Emissivity spectrum normalization, in Proceedings of the Second AnnualAirborne Earth Science Workshop, vol. 2, edited by E. A. Abbott, JPLPubl. 90–55, 31–35.

Reheis,M. C. (2006), A 16‐year record of eolian dust in southernNevada andCalifornia, USA: Controls on dust generation and accumulation, J. AridEnviron., 67(3), 487–520, doi:10.1016/j.jaridenv.2006.03.006.

Reynolds, R. L., J. C. Yount, M. Reheis, H Goldstein, P. Chavez Jr.,R. Fulton, J. Whitney, C. Fuller, and R. M. Forester (2007), Dust emis-sion from wet and dry playas in the Mojave Desert, USA, Earth Surf.Processes Landforms, 32(12), 1811–1827, doi:10.1002/esp.1515.

Rosen, M. R. (1994), The importance of groundwater in playas: A reviewof playa classifications and the sedimentology and hydrology of playas,in Paleoclimate and Basin Evolution of Playa Systems, edited by M. R.Rosen, Spec. Pap. Geol. Soc. Am., 289, 1–18.

Sabol, D. E., A. R. Gillespie, E. McDonald, and I. Danillina (2006), Differ-ential thermal inertia of geological surfaces, paper presented at 2ndAnnual International Symposium of Recent Advances in QuantitativeRemote Sensing, Univ. of Valencia, Torrent, Spain, 25–29 Jul.

Saleh, A., and D. W. Fyrear (1995), Threshold wind velocities of wet soilsas affected by windblown sand, Soil Sci., 160(4), 304–309.


22 of 23

Schaaf, C. B., et al. (2002), First operational BRDF, albedo nadir reflec-tance products from MODIS, Remote Sens. Environ., 83, 135–146,doi:10.1016/S0034-4257(02)00091-3.

Schädlich, S., F. M. Göttsche, and F.‐S. Olesen (2001), Influence of landsurface parameters and atmosphere on METEOSAT brightness tempera-tures and generation of land surface temperature maps by temporally andspatially interpolation atmospheric correction, Remote Sens. Environ., 75,39–46, doi:10.1016/S0034-4257(00)00154-1.

Scheidt, S. P., M. S. Ramsey, and N. Lancaster (2008a), Image mosaicgeneration of ASTER thermal infrared data: An application to extensivesand sheets and dune fields, Remote Sens. Environ., 112, 920–933,doi:10:1016/j.rse.2007.06.020.

Scheidt, S. P., M. S. Ramsey, and N. Lancaster (2008b), Thermal remotesensing of sand transport systems, paper presented at Planetary DunesWorkshop: A Record of Climate Change, Abstract 1403, Lunar Planet.Inst., Alamogordo, N. M.

Shao, Y., M. R. Raupach, and P. A. Findlater (1993), Effect of saltationbombardment on the entrainment of dust and wind, J. Geophys. Res.,98, 12,719–12,736, doi:10.1029/93JD00396.

Shao, Y., M. R. Raupach, and J. F. Leys (1996), A model for predictingaeolian sand drift and dust entrainment on scales from paddock to region,Aust. J. Soil Res., 34, 309–342, doi:10.1071/SR9960309.

Sherman, D. J. (1990), Evaluation of aeolian sediment sand transportequations using intertidal‐zone measurements, Saunton Sands, England,Sedimentology, 37, 385–392, doi:10.1111/j.1365-3091.1990.tb00967.x.

Svoboda, M., et al. (2002), The drought monitor, Bull. Am. Meteorol. Soc.,83(8), 1181–1190, doi:10.1175/1520-0477(2002)083<1181:TDM>2.3.CO;2.

Thome, K., F. Palluconi, T. Takashima, and K. Masuda (1998), Atmo-spheric correction of ASTER, IEEE Trans. Geosci. Remote Sens., 36,1199–1211, doi:10.1109/36.701026.

Tonooka, H., and A. Iwasaki (2003), Improvement of ASTER/SWIR cross-talk correction, Proc. SPIE, 5234, 168–179, doi:10.1117/12.511811.

van den Bergh, F., A. van Wyk, B. J. van Wyk, and G. Udahemuka (2007),A comparison of data‐driven and model‐driven approaches to brightnesstemperature diurnal cycle interpolation, SAIEE Afr. Res. J., 98(3), 81–86.

van den Dool, H., J. Huang, and Y. Fan (2003), Performance and analysisof the constructed analogue method applied to U.S. soil moisture over1981–2001, J. Geophys. Res. , 108(D16), 8617, doi:10.1029/2002JD003114.

Wang, C. Y., S. H. Qi, and Z. Niu (2004), Evaluating soil moisture status inChina using the temperature‐vegetation dryness index (TVDI), Can. J.Remote Sens., 30(5), 671–679.

Watson, K., L. C. Rowen, and T. W. Offield (1971), Application of thermalmodeling in the geologic interpretation of IR images, Remote Sens. Envi-ron., 3, 2017–2041.

Wolfe, R. E.,M. Nishihama, A. J. Fleig, J. A. Kuyper, D. P. Roy, J. C. Storey,and F. S. Patt (2002), Achieving sub‐pixel Geolocation accuracy in sup-port of MODIS land science, Remote Sens. Environ., 83, 31–49,doi:10.1016/S0034-4257(02)00085-8.

Xue, Y., and A. P. Cracknell (1992), Thermal inertia mapping: Fromresearch to operation, in Proceedings of the 18th Annual Conference ofthe Remote Sensing Society held in Univ. of Dundee on 15–17 September1992, edited by A. P. Cracknell and R. A. Vaughan, pp. 471–480,Remote Sens. Soc., Nottingham, U. K.

Xue, Y., and A. P. Cracknell (1995), Advanced thermal inertia modeling,Int. J. Remote Sens., 16(3), 431–446, doi:10.1080/01431169508954411.

Yamaguchi, Y., A. Kahle, H. Tsu, T. Kawakami, and M. Pniel (1998),Overview of the Advanced Spaceborne Thermal Emission and Reflec-tance Radiometer (ASTER), IEEE Trans. Geosci. Remote Sens., 36,1062–1071, doi:10.1109/36.700991.

Zhang, R. H., X. M. Sun, and Z. L. Zhu (2002), Remote sensing informa-tion model in surface evaporation from differential thermal inertia andit’s validation in Gansu Province, Sci. China Ser. D., 32, 1041–1050.

Zhang, R., X. Sun, Z. Zhu, H. Su, and X. Tang (2003), A remote sensingmodel for monitoring soil evaporation based on differential thermal iner-tia and its validation, Sci. China Ser. D, 46, 342–355.

Zhenhua, L., and Z. Yingshi (2006), Research on the method for retrievingsoil moisture using thermal inertia model, Sci. China Ser. D., 49, 539–545, doi:10.1007/s11430-006-0539-6.

Zhou, L., R. E. Dickinson, K. Ogawa, Y. Tian, M. Jin, T. Schmugge, andE. Tsvetsinskaya (2003), Relations between albedos and emissivitiesfrom MODIS and ASTER data over North African Desert, Geophys.Res. Lett., 30(20), 2026, doi:10.1029/2003GL018069.

N. Lancaster, Desert Research Institute, 2215 Raggio Pkwy., Reno, NV89512, USA.M. Ramsey and S. Scheidt, Department of Geology and Planetary

Science, 200 SRCC Bldg., University of Pittsburgh, Pittsburgh, PA15260, USA. ([email protected])


23 of 23

Determining soil moisture and sediment availability at ...

Documents