Soil Hydraulic Property Estimation Using Remote Sensing: a Review

Home , Remote sensing (geology)

Special Section: VZJ Soil Hydraulic Property Estimation Anniversary Issue Using Remote Sensing: A Review A review of recent developments related to soil hydraulic property estimation using remote Binayak P. Mohanty* sensing is presented. Several soil hydraulic parameter estimation techniques using proximal-, air-, or satellite-based remotely sensed soil moisture, land surface temperature, and/ or evapotranspiration time series have evolved over the past decades. In particular, microwave remote sensing of near-surface soil moisture has played a key role in this respect. Inverse modeling, regression techniques, data assimilation methods by utilizing soil–vegetation–atmosphere transfer models, genetic algorithm based optimization, and uncertainty quantification using ensemble based approaches under synthetic and field conditions have been developed and adapted over these years. In this context multiple approaches for estimating effective soil hydraulic parameters at different spatial scales (e.g., point to remote sensing footprints) by various scaling methods are also summarized. Methods reviewed include traditional bottom-up approaches, such as homogenization of local scale properties up to footprint scale, as well as newer top-down approaches, such as use of inverse model- A review of recent developments ing of remotely sensed soil moisture or temperature using genetic algorithms, Markov Chain Monte Carlo simulations with data assimilation, and multi-scale Bayesian neural networks. related to soil hydraulic property While new remote sensing platforms provide powerful tools and techniques for noninvasive estimation using remote sensing soil hydraulic property estimation of the vadose zone at multiple scales, current limitations is presented. including vision for future developments are also summarized.

B.P. Mohanty, Dep. of Biological and Agricul- Abbreviations: AMSR-E, Advanced Microwave Scanning Radiometer; ANN, artificial neural network; BNN, tural Engineering, Texas A&M Univ., College Bayesian framework expanded multi-scale ANN technique; ESA, European Space Agency; ESTAR, Electroni- Station, TX 77843. *Corresponding author cally Scanned Thinned Array Radiometer; ET, evapotranspiration; GPR, ground penetrating radar; MCMC, ([email protected]). Markov Chain Monte Carlo; NMCGA, Noisy Monte Carlo Genetic Algorithm; PALS, Passive and Active L- band Sensor; PSR, Polarimetric Scanning Radiometer; PTF, pedotransfer function; RS, remote sensing; SMAP, Soil Moisture Active and Passive; SMOS, Soil Moisture and Ocean Salinity; SVAT, soil–vegetation– Vadose Zone J. atmosphere transfer. doi:10.2136/vzj2013.06.0100 Received 12 June 2013.

Soil hydraulic properties (water retention, hydraulic conductivity) are by far the most important land surface parameters to govern the partitioning of soil moisture between infiltration and evaporation fluxes at a range of spatial scales. However, an obstacle to their practical application at the field, catchment, watershed, or regional scale is the difficulty of quantifying the “effective” soil hydraulic functions q(h) and K(h), where q is the soil water or moisture content, h (or y) is the pressure head, and K is unsaturated hydraulic conductivity. Proper evaluation of the water balance near the land–atmosphere boundary depends strongly on appropriate characterization of soil hydraulic parameters under field conditions and at the appropriate process scale. In recent years a number of approaches have been adopted to tackle this problem including: (i) bottom-up approaches, where larger-scale effective parameters are calculated by aggregating point-scaleinsitu hydraulic property measurements; (ii) top-down approaches, where effective soil hydraulic parameters are estimated by inverse modeling using remotely sensed soil moisture or soil temperature measurements; and (iii) artificial neural network approaches, where effective soil hydraulic parameters are estimated by exploiting the correlations with soil texture, topographic attributes, and vegetation characteristics at multiple spatial resolutions. In this review we summarize these recent developments.

66Soil Hydraulic Functions For meso- and regional-scale soil–vegetation–atmosphere transfer (SVAT) schemes in hydroclimatic models, pixel dimensions may range from several hundred square meters to © Soil Science Society of America several square kilometers. Pixel-scale soil hydraulic parameters and their accuracy are criti- 5585 Guilford Rd., Madison, WI 53711 USA. All rights reserved. No part of this periodical may cal for the success of hydroclimatic and soil hydrologic models. Unsaturated flow typically be reproduced or transmitted in any form or by any means, electronic or mechanical, including pho- uses closed-form functions to represent water retention characteristics and unsaturated tocopying, recording, or any information storage and retrieval system, without permission in writing hydraulic conductivities. Leij et al. (1997) provided a review of popular closed-form from the publisher.

www.VadoseZoneJournal.org expressions for soil water retention and hydraulic conductivity. infiltration in unsaturated soil have been conducted (Mohanty and Gardner’s exponential model of hydraulic conductivity, Brooks Zhu, 2007; Zhu and Mohanty 2002a,b, 2003a,b, 2004, 2006; Zhu and Corey, and van Genuchten soil water retention functions rep- et al., 2004, 2006). They also investigated how effective parameters resent some of the most widely used models. under various flow scenarios and landscape heterogeneity (example scenarios of landscape heterogeneity are shown in Fig. 1a), such as Gardner–Russo Model: the dryness of the fields, mosaics of soil textures across the landscape, organization of subsurface soil layers, presence of plants or 1/(1+b ) éùæ11 öæ ö =êúçç +a÷÷ -a roots, water management practices as irrigation and drainage, and She( )çç 1 h÷÷ exp h [1a] ëûêúè22 øè ø landscape configuration with various slope, aspect, and elevation may or may not differ. -ah K= Kes [1b] 66Inverse Modeling and Effective Brooks–Corey Model: Soil Hydraulic Parameterization To address the land surface heterogeneity issue in the context of −l Se(y) = (ay) when ay > 1 [2a] developing effective soil properties at a remote sensing footprint scale, Camillo et al. (1986) was among some of the early ones who Se(y) = 1 when ay £ 1 [2b] envisioned that soil hydraulic properties may be estimated using −b K(y) = Ks(ay) when ay > 1 [3a] microwave based soil moisture and soil temperature that drive water and energy fluxes near the land surface (Fig. 1b). Feddes et K(y) = Ks when ay £ 1 [3b] al. (1993) further asserted that the inverse modeling on the basis where b=l( + 2) + 2 of soil moisture status at the top and bottom of the soil profile, cumulative areal evapotranspiration, and cumulative percolation van Genuchten–Mualem Model: to groundwater for estimating effective soil hydraulic parameters (based on small scale soil physics) at large scale is feasible under m q()h -q éù1 different land cover conditions. In addition, local scale studies ==res êú [4a] Se êún qsat -q res ëûêú1|+ ay | by Ahuja et al. (1993) and Chen et al. (1993) showed successes in describing average soil hydraulic conductivity of the soil profile 2 -m ml (y) =11 -( ay)mn éùéù +( ay)n 1 +( ay)n [4b] by using the changes in surface soil moisture. Using different soil KKs { ëûëûêúêú} textures, they observed that the strongest relationship between surface soil moisture and profile soil hydraulic conductivity exists 1 m=-1 for sandy soils and the relationship weakens with higher clay con- n tent. In addition, they discovered hydroclimatic conditions with enhanced evaporation may also affect the parameters estimation. where Se is degree of saturation; qs is saturated water content; qres Burke et al. (1997, 1998) provided a calibration technique for soil is residual water content, Ks is the saturated hydraulic conductiv- water and energy balance model for estimating soil water reten- ity; a, b, and l are related to pore-size distributions (note that tion and hydraulic conductivity parameters. They investigated the significance of these parameters may vary among different saturated hydraulic conductivity, bulk density, air entry pressure, models); m and n are empirical parameters; and l is a parameter and other soil hydraulic parameters. Hollenbeck et al. (1996) and that accounts for the dependence of the tortuosity and the correla- Mattikalli et al. (1998) developed regression based tools for relat- tion factors on the water content. When these models are used in ing airborne-based multi-temporal passive microwave (L band) large heterogeneous scale processes, major questions remain about brightness temperature values (which relates to soil moisture how to represent hydraulic properties over a heterogeneous soil status) with saturated hydraulic conductivity and soil texture for volume and what averages of hydraulic property shape parameters remote sensing footprints across large regions. Chang and Islam to use for these models. Studies addressing the impact of averaging (2000) developed an artificial neural network scheme for estimat- methods of shape parameters, parameter correlation, and correla- ing soil textural class from remotely sensed brightness temperature tion length on ensemble flow behavior in heterogeneous soils have data during Washita 92 field campaign in Oklahoma. Uri et al. been undertaken by various researchers in the past (e.g., Yeh et (1991) utilized soil moisture based on passive microwave radiom- al., 1985; Mantoglou and Gelhar, 1987). Other studies develop- etry (1.46 GHz) in the evaluation of drainage patterns at different ing effective parameters that will predict ensemble behavior of locations of a large watershed leading to shallow subsurface soil the heterogeneous soils and investigating the effectiveness of the hydraulic property estimation. With the availability of near sur- “effective parameters” in terms of the degree of correlation between face soil moisture using passive and active microwave sensing from parameters for the steady-state and transient evaporation and air-based sensors (e.g., Passive and Active L-band Sensor [PALS],

www.VadoseZoneJournal.org p. 2 of 9 Fig. 1. (a) Heterogeneous landscapes with diff erent vegetation, soils, and topography and (b) conceptual framework for inverse estimation of soil hydraulic properties using near-surface soil moisture time series data from remote sensing platform.

Electronically Scanned Th inned Array Radiometer [ESTAR], 2008) have shown success in calibrating soil hydraulic properties and Polarimetric Scanning Radiometer [PSR]) and space-based at large scale using surface soil moisture from remote sensing. It sensors (e.g., Advanced Microwave Scanning Radiometer [AMSR- is worth mentioning that there have been mixed results for esti- E], Soil Moisture and Ocean Salinity [SMOS], AQUARIUS) in mating surface soil moisture using active microwave (radar-based) the past decade, estimating soil hydraulic parameters under dif- remote sensing, due to sensitivity of high frequency backscatter ferent hydrologic and climatic conditions has been studied by to the nature and degree of surface interaction. In contrast, pas- several research groups. Among others, Santanello et al. (2007) sive microwave sensing has been found to be most successful to parameterized soil hydraulic parameters and soil texture using a measure land surface soil moisture and holds the most promise to combination of passive microwave remote sensing soil moisture estimate profi le soil hydraulic properties. of NASA’s push-broom microwave radiometer (PBMR, L-band, 21 cm) and active microwave RADARSAT-1 (C-band; 5.6 cm) Besides volumetric soil moisture content (related to land surface imagery, a parameter estimation algorithm (PEST; Doherty, 2004) emissivity), land surface temperature measurements have also been and Noah Land Surface Model in a semiarid watershed and com- investigated to produce surface soil hydraulic properties. Using pared the estimation to pedotransfer function (PTF) based soil latent heat fl ux (evapotranspiration) as the matching variable, parameters. Th ey suggested that satellite based remotely sensed Gutmann and Small (2007) concluded that inverse modeling soil moisture data becoming widely available on a regular basis with soil skin temperature measured by remotely sensed infrared in the recent years at the global scale, the prospect of quantify- measurements provide more accurate soil hydraulic parameters for ing such eff ective parameters would be more a reality at multiple land surface models than soil texture based PTF schemes. In yet scales. Other studies (e.g., Hogue et al., 2005; Peters-Lidard et al., another follow-up study, Gutmann and Small (2010) evaluated the

www.VadoseZoneJournal.org p. 3 of 9 use of MODIS based surface temperature for estimating landscape hydraulic properties, based on indirect relationship among soil temperature, soil moisture, and evapotranspiration. Th ey argued that under complex soil, topographic and vegetation heterogeneity in large landscapes, dominant water fl ow processes including infi ltration, evaporation, transpiration, and surface runoff may vary across space and time scales and thus can be eff ectively represented by landscape hydraulic properties instead of small-scale soil hydraulic properties. Th ey also showed that using the inversely estimated soil hydraulic properties decreases error in modeled latent and sensible heat fl uxes by more than 60% compared to standard texture based classifi cation. Using infrared-measured soil surface Fig. 2. (a) NASA’s upcoming Soil Moisture Active Passive (SMAP) satellite mission to be launched in 2014. (b) NASA’s Twin Otter temperature and time domain refl ectometry–based soil water con- aircraft and ground sensor deployment during multi-scale cal/val fi eld tents in a framework coupling HYDRUS_1D model with global campaigns of satellite-based microwave sensors. optimizer (DREAM), Steenpass et al. (2010) inversely estimated van Genuchten–Mualem parameters for synthetic and real fi eld conditions. Th ey also compared the parameters obtained by using domain in an eff ective manner. Scope of their study also evaluated single objective (soil surface temperature) and multi-objective (soil if the method is robust to handle various nonlinearities and uncer- surface temperature and water content) in their study. tainties under diff erent hydrologic and climatic conditions. While theoretically the method can be adapted for any soil hydraulic Currently, soil moisture data from remote sensing (RS) are lim- conductivity function and land surface model, van Genuchten– ited to the near-surface soil layers. Under minimal vegetation Mualem model was used within SWAP model framework (van cover, the maximum penetration depth of a microwave L-band Dam et al., 1997) in their studies. Near-surface soil moisture data sensor is about 5 cm. Many studies have been done to retrieve was used to derive eff ective q(h) and K(h) for the vadose zone for soil moisture profi les using these passive and active microwave each remote sensing pixel assuming the constitutive functions brightness temperature data. Currently deployed SMOS mission of Mualem–van Genuchten and vertical soil water movement in by European Space Agency (ESA), AQUARIUS a NASA mis- the unsaturated zone defi ned by the Richards equation. A fl ow sion, and the upcoming Soil Moisture Active and Passive (SMAP) chart representation of the assimilation process is shown in Fig. 3. mission of NASA in +2014 (Fig. 2a) are designed to populate soil Sample results from this approach are shown in Fig. 4. Extending moisture time series data across the globe for this decade (2010– their original work, Ines and Mohanty (2008c) developed the 2020). Soil moisture retrieved from these satellite platforms are concept of Noisy Monte Carlo Genetic Algorithm (NMCGA) further calibrated and validated at multiple scales using airborne to formulate an ensemble-based general method for estimating microwave sensors and handheld or ground-based sensors (Fig. the eff ective soil hydraulic parameters and their uncertainties at 2b) and pilot studies in various watersheds and hydroclimatic the remote sensing footprint scale. Th e main assumption of the regions across the globe. domain-dependent parameter estimation concept is based on the idea that the eff ective forms of the soil hydraulic functions (at the Researchers have taken this unique opportunity to develop and RS footprint) can be inferred by the derived eff ective soil hydraulic evaluate advanced inverse estimation methods for soil hydraulic parameters from large-scale RS soil moisture data inversion with properties using remote sensing data and their scaling behavior no lateral exchange of fl ow across adjacent RS footprints. In the recently. Using time series of passive microwave (L-band) based synthetic case studies under pure (one soil texture) and mixed-pixel soil moisture for airborne (e.g., ESTAR, PSR) and satellite (multiple soil textures) conditions, NMCGA performed well in (e.g., AMSR-E) footprints, Ines and Mohanty (2008a, b, 2009) estimating the eff ective soil hydraulic parameters even with pixel developed Genetic Algorithm based inverse modeling and data complexities contributed by various soil types and land manage- assimilation methods for quantifying effective soil hydraulic ment practices (rainfed/irrigated). With the airborne and satellite parameters at the remote sensing footprint scale. Th ey conducted remote sensing cases, NMCGA also performed well for estimating several numerical and fi eld studies for various possible landscape eff ective soil hydraulic properties so that when applied in forward scenarios including diff erent soil textures (coarse to fi ne), ground- stochastic simulation modeling it can mimic large-scale soil mois- water table depths (shallow to deep), land covers (bare to vegetated), ture dynamics. Th eir results also suggest a possible scaling down and climatic (arid to humid) conditions. Th e basic premises of their eff ect of the eff ective soil water retention curve q(h) at the larger work include if surface soil moisture temporal dynamics hold the satellite remote sensing pixel compared to airborne remote sensing memory of hydrologic fl uxes (infi ltration, redistribution, evapo- pixel. Expanding on the inverse modeling and data assimilation transpiration) occurring across the entire vadose zone and thus can idea, Shin et al. (2012) developed a layer-specifi c soil moisture provide a representation of soil hydraulic properties of the entire assimilation scheme by interjecting subsurface soil moisture

www.VadoseZoneJournal.org p. 4 of 9 Fig. 3. Implementation of near-surface soil moisture assimilation scheme (aft er Ines and Mohanty, 2008a).

Fig. 4. Eff ects of (a, b, e, and f ) initial and bottom boundary conditions, (b, c, g, and f ) rooting depth, and (b, d, f, and h) root density on eff ective water retention q(h) and hydraulic conductivity K(h) estimation for the Oklahoma Southern Great Plains 1997 (SGP97) Site ARS-135 for a loam soil (BBC is bottom boundary condition, RZmax is maximum rooting depth, RZdensity is root density, and Dassim denotes predicted by the inverse modeling based near-surface soil moisture assimilation scheme) (aft er Ines and Mohanty, 2008a).

information. Th ey demonstrated that soil layers and vertical het- with MODIS data) to the objective function in the soil hydraulic erogeneity (layering sequences) could impact the uncertainty of property estimation scheme shown in Fig. 3, Shin et al. (2013) quantifying eff ective soil hydraulic parameters. Using layer-specifi c presented the improvement of eff ective soil hydraulic parameter assimilation, they found further improvement of soil hydraulic estimation in comparison to values based on single objective func- parameter estimation and their accuracy. tion (soil moisture content only) in diff erent hydroclimates. Note, however, that technical issues relating to ET measurements using Evaporation and transpiration plays a complementary role to root remote sensing tools persist to date. Further development on ET zone soil moisture for soil water balance. Depending on root water estimation using remote sensing techniques in the future will refi ne uptake and soil evaporation, soil water may redistribute across soil the eff ective soil hydraulic property estimation. Uncertainties of profi le maintaining certain functional thresholds. Adding evapo- water fl uxes in SVAT model by inverting soil moisture and evapotranspiration (ET) (estimated by S-SEBI and SEBAL algorithms transpiration in 22 hydroclimates were analyzed by Pollacco and

www.VadoseZoneJournal.org p. 5 of 9 Mohanty (2012). The primary conclusions from their study include technique also has the potential to derive upscaled parameters of uncertainties in simulated water fluxes increases as (i) the climate other geophysical properties used in remote sensing of land surface gets drier, (ii) the texture gets coarser, or (iii) the roots grow deeper. states. The developed MCMC algorithm with SVAT model can They attributed the uncertainty in recharge to soil moisture and be very useful for land–atmosphere interaction studies and further transpiration decoupling. Soil moisture decoupling occurs when understanding of the physical controls responsible for soil mois- the information provided by surface flux is no longer representative ture dynamics at different scales. Subsequently, Das et al. (2010) of root zone flux. Transpiration decoupling occurs when there is proposed an algorithm that uses the Karhunen–Loève expansion substantially more water storage at depth. These new root zone in conjunction with the MCMC technique, which employs mea- ecohydrologic process understanding will enhance accuracy of sured soil moisture values to characterize the saturated hydraulic effective soil hydraulic property estimation at the landscape scale conductivity distributions of an agricultural field at a 30-m resolu- in the future. tion. In a similar line, Montzka et al. (2011) developed a particle filter data assimilation scheme using ALOS and SMOS satellite Besides microwave remote sensing, Vereecken et al. (2007) also based soil moisture data and HYDRUS-1D model for deriving suggested that combination of field-scale proximal-sensing van Genuchten soil hydraulic parameters. techniques and hydrologic models provides the platform for accounting the upscaled behavior of the soil hydraulic prop- Indirect estimation techniques using PTFs (e.g., Sharma et al., erties and flow processes in heterogeneous soils. In that spirit, 2006) provide an effective alternative to direct measurements Lambot et al. (2009) developed a technique using full-waveform and/or to inverse modeling and data assimilation tools for large integrated hydro-geophysical inversion of time-lapse off-ground scale soil hydraulic parameters. This approach examines the effect ground penetrating radar (GPR). One of the advantages of this of including topographic and vegetation attributes, besides pedo- technique over the microwave techniques described earlier is logic attributes, on the prediction of soil hydraulic properties deeper penetration depth. using PTFs. With the increasing availability of remote sensing products from air- and spaceborne sensors at different spatial 66 scales, topographic and vegetation attributes are easily available Scaling Up or Down Soil from digital elevation models and normalized difference vegeta- Hydraulic Parameters tion index. Traditionally, PTFs have been used to estimate the An alternative to the inverse estimation technique described required soil hydraulic parameters from other available or easily above, Das et al. (2008) developed a Markov Chain Monte Carlo measurable soil properties. While most previous studies derive and (MCMC) based data assimilation algorithm to derive upscaled adopt these parameters at matching spatial scales (1:1) of input land surface hydraulic parameters (from apriori local-scale param- and output data, Jana et al. (2007) developed a methodology to eter distributions) for a SVAT model using time series data of derive soil water retention functions at the point or local scale satellite-measured atmospheric forcings (e.g., precipitation), and using the PTFs trained with coarser scale input data. This study land surface states (e.g., soil moisture and vegetation). Their study was a novel application of an artificial neural network (ANN)- focused especially on the evaluation of soil moisture measurements based PTF scheme across two spatial support scales within the Rio of the Aqua satellite based AMSR-E instrument using the MCMC- Grande basin in New Mexico. The ANN was trained using soil based scaling algorithm. Soil moisture evolution was modeled at texture and bulk density data from the SSURGO database (scale a spatial scale comparable to the AMSR-E soil moisture product, 1:24,000) and then used for predicting soil water contents at differ- with the hypothesis that the characterization of soil microwave ent pressure heads with point-scale data (1:1) inputs. The resulting emissions and their variations with space and time on soil surface outputs were corrected for bias before constructing the soil water within the AMSR-E footprint can be represented by an ensemble characteristic curve using the van Genuchten equation. A hierar- of upscaled soil hydraulic parameters. They demonstrated the fea- chical approach with training data derived from multiple clustered tures of the MCMC-based parameter upscaling algorithm (from subwatersheds (with varying spatial extent) was used to study the field to satellite footprint scale) within a SVAT model framework effect of the increase in spatial extent. The results show good agree- to evaluate the satellite-based brightness temperature/soil moisture ment between the soil water retention curves constructed from the measurements for different hydro-climatic regions, and identified ANN-based PTFs and field observations at the local scale near Las the temporal effects of vegetation (leaf area index) and other envi- Cruces, NM. The robustness of the multiscale PTF methodology ronmental factors on AMSR-E based remotely sensed soil moisture was further tested with a separate data set from the Little Washita data. The SVAT modeling applied for different hydroclimatic watershed region in Oklahoma. Overall, ANN coupled with bias regions revealed the limitation of AMSR-E measurements in high- correction was found to be a suitable approach for deriving soil vegetation regions. The study also suggested that inclusion of soil hydraulic parameters at a finer scale from soil physical properties moisture evolution from the proposed upscaled SVAT model with at coarser scales and across different spatial extents. Later, using AMSR-E measurements in data assimilation routine will improve nonlinear bias correction and cumulative distribution function the quality of soil moisture assessment in a footprint scale. The matching, Jana et al. (2008) developed a Bayesian framework to

www.VadoseZoneJournal.org p. 6 of 9 expand their multi-scale ANN technique (also called BNN) for a certain level of accuracy, microwave sensing for soil moisture estimating soil hydraulic parameters at multiple scales including is only limited to the regions with vegetation biomass less than their uncertainties. Further, Jana and Mohanty (2011) and Jana et a prescribed threshold value (e.g., L-band radiometer retrieval al. (2012) evaluated their BNN technique using several remotely up to 4 Kg m−2) and does not work for densely vegetated areas. sensed and in situ land surface data in different hydroclimates. Future efforts should be geared at developing improved soil Their findings suggested that the BNN-based approach could be moisture retrieval algorithms categorized by remote sensing used for up or downscaling soil hydraulic parameters effectively. platforms by explicitly accounting for type of sensor, footprint size, and within-footprint heterogeneity, and by hydroclimatic More recently, using remote sensing data, Jana and Mohanty (2012) regions. In this context, using other remote sensing spectrum evaluated various soil hydraulic properties estimation and scaling (e.g., optical, thermal, hyper-spectral, full range of spectra), and scheme in the context of watershed hydrology. The equivalence of proximal sensing (e.g., in situ sampling, spectroscopy, electro- the upscaled parameters was tested by simulating water flow for the magnetic induction, electrical resistivity tomography, GPR) watershed pixels in HYDRUS-3D, and comparing the resultant tools for different soil attributes (moisture, texture, organic soil moisture states with data from the ESTAR airborne sensor carbon, mineralogy, iron content, salinity, and other proxies), during the SGP97 hydrology experiment. as described by Ben-Dor et al. (2009), Mulder et al. (2011), and Casa et al. (2013), can possibly be fused and translated to 66 more complex soil hydraulic properties by developing PTFs. Current Challenges and These values can subsequently be upscaled or downscaled Future Venues for Development matching to the application. or Improvement Using inverse estimation and/or spatial scaling based on micro- 3. Decoupling of surface and subsurface moisture in various wave remote sensing of soil moisture proved to be one of the most hydroclimates, such as done by Pollacco and Mohanty (2012), potent venues for soil hydraulic property estimation from local Pollacco et al. (2013), and Shin et al. (2012), under various to regional scales. While these methods provide the initial break- soil textures, plant rooting depths, and precipitation condi- through for estimating some of the most challenging properties tions, reduces accuracy of estimated soil hydraulic parameters of soil (effective hydraulic properties with large support volume) using only surface soil moisture content observed by remote needed for many hydrologic, land–atmosphere interaction, and sensing in the objective function for parameter optimization. environmental applications, it is far from optimal. Among others, By adding weighted ET as a second criterion in the optimi- primary limitations of these approaches and possible ways forward zation formulation improves the efficacy of the soil hydraulic to develop and improve them further include: parameter estimation method. However, measurement/estimation of ET at the corresponding (matching to soil moisture 1. Penetration depth of microwave remote sensing is limited to measurement) spatiotemporal scale using remote sensing is far approximately top few centimeters (?0–5 cm for L-Band radi- from being perfect. Issues such as boundary layer turbulence ometers) of soil depth, where many land surface disturbances effect, energy balance closure, and others remain to be further and soil structures play a significant role resulting in errors developed. Concerted efforts of measuring soil water, vegeta- in the dynamics of soil hydrologic processes. In the future, tion water, and ET using co-located remote sensing platforms microwave sensors with different wavelengths (e.g., p-band) with matching resolution will provide maximum impetus for penetrating to deeper depths may able to minimize such errors. the progress soil hydrologic science and its applications at dif- Deployment of such a sensor antenna aboard a satellite plat- ferent scales. form, however, is an engineering challenge in the near future. Alternatively, soil hydrologic models with data assimilation 4. Spatial and temporal resolution of currently available remote approach (e.g., ensemble Kalman Filter) by interjecting various sensing data may not be optimum for capturing all the hydro- local, proximal, or airborne soil water measurements at deeper logic processes (e.g., overland flow across remote sensing depths and other surrogate variables could extend the accuracy footprints) including the extreme scenarios (e.g., complete of effective soil hydraulic parameterization schemes. saturation and preferential flow during precipitation) within a remote sensing footprint resulting in (nonoptimal) soil hydrau- 2. Accuracy of soil moisture retrieval using brightness tempera- lic parameters, which may fail to describe the entire spectrum ture (based on soil emissivity) is currently being affected by of hydrologic events. Upcoming satellite sensors as SMAP a number of factors such as type of sensor (i.e., bandwidth/ will generate multiresolution (3, 9, and 36 km) soil moisture frequency, polarization, scanning pattern, and footprint size), products, providing the opportunity for generating multiscale within-footprint heterogeneity of soil texture, vegetation, and effective soil hydraulic properties. In this context, however, it landscape (topographic) features, as well as limited observation should be pointed out that computational cost for using fine of vegetation water content and temperature (affected by both spatial and temporal resolution data in inverse modeling and vegetation and soil temperature). Furthermore, for maintaining data assimilation schemes may possibly be far greater with only

www.VadoseZoneJournal.org p. 7 of 9 a small gain in accuracy of effective soil hydraulic parameters. Chen, C., D.M. Thomas, R.E. Green, and R.J. Wagenet. 1993. Two-domain estimation of hydraulic properties in macropore soils. Soil Sci. Soc. Am.J. Users are advised to make informed decisions on the accuracy 57:680–686. doi:10.2136/sssaj1993.03615995005700030008x required, and thus grid resolution (for downscaling or upscal- Das, N.N., B.P. Mohanty, and Y. Efendiev. 2010. Characterization of effective -sat urated hydraulic conductivity in an agricultural field using Karhunen‐Loève ing) based on specific application, extent of hydrologic domain, expansion with the Markov chain Monte Carlo technique. Water Resour. landscape complexity, and computational burden. Res. 46:W06521. doi:10.1029/2008WR007100 Das, N.N., B.P. Mohanty, and E.G. Njoku. 2008. An Markov chain Monte Carlo algorithm for upscaled soil-vegetation-atmosphere-transfer modeling to 5. Discrepancy in soil hydrologic process scale based on evaluate satellite-based soil moisture measurements. Water Resour. Res. 44:W05416. doi:10.1029/2007WR006472 established continuum-scale capillary bundle theory leading Doherty, J. 2004. PEST: Model independent parameter estimation. User manu- to Darcy and Richards type model formulations as opposed to al. 5th ed. Watermark Numerical Computing, Brisbane, Australia. Feddes, R.A., M. Menenti, P. Kabat, and W.G.M. Bastiaanssen. 1993. Is large- the underlying dominant hydrologic processes at the support scale inverse modeling of unsaturated flow with areal average evaporation and surface soil moisture as estimated from remote sensing feasible? J. Hy- scale of remote sensing footprint. Although the formalized drol. 143:125–152. doi:10.1016/0022-1694(93)90092-N capillary-based governing flow equation may be best suited Gutmann, E.D., and E.E. Small. 2010. A method for the determination of the hydraulic properties of soil from MODIS surface temperature for use in land‐sur- for local flow processes and provide the basis of currently face models. Water Resour. Res. 46:W06520. doi:10.1029/2009WR008203 available parameter estimation schemes, the physical basis of Gutmann, E.D., and E.E. Small. 2007. A comparison of land surface model soil hydraulic properties estimated by inverse modeling and pedotransfer func- their adoption at the scale of the remote sensing footprint is tions. Water Resour. Res. 43:W05418.doi:10.1029/2006WR005135 debated. Alternative formulations based on thermodynamic Hogue, T.S., L. Bastidas, H. Gupta, S. Sorooshian, K. Mitchell, and W. Emmerich. 2005. Evaluation and transferability of the Noah land surface model in semi- principles, hydrologic functionality rules, or scale-appropriate arid environments. J. Hydrometeorol. 6:68–84. doi:10.1175/JHM-402.1 water balance models across different irregular hydrologic Hollenbeck, K.J., T.J. Schmugge, G.M. Hornberger, and J.R. Wang. 1996. Identi- fying soil hydraulic heterogeneity by detection of relative change in passive units (e.g., catchments, watersheds, river basins) using existing microwave remote sensing observations. Water Resour. Res. 32:139–148. or novel remote sensing attributes or indices may prove to be doi:10.1029/95WR02916 Ines, A.V.M., and B.P. Mohanty. 2008a. Near-surface soil moisture assimilation more effective for describing large scale soil hydrology. for quantifying effective soil hydraulic properties under different hydroclimatic conditions. Vadose Zone J. 7:39–52. doi:10.2136/vzj2007.0048 Ines, A.V.M., and B.P. Mohanty. 2008b. Near-surface soil moisture assimilation to quantify effective soil hydraulic properties using genetic Although limitations for remote sensing technology persist in algorithm. 1. Conceptual modeling. Water Resour. Res. 44:W06422. terms of penetration depth, spatial and temporal resolutions, doi:10.1029/2007WR005990. Ines, A.V.M., and B.P. Mohanty. 2008c. Parameter conditioning with a noisy accuracy, and associated soil hydrologic process description at Monte Carlo genetic algorithm to estimate effective soil hydraulic properties the footprint scale, the above studies show its great promise for from space. Water Resour. Res. 44:W08441. doi:10.1029/2007WR006125. Ines, A.V.M., and B.P. Mohanty. 2009. Near-surface soil moisture assimilation to estimating effective soil hydraulic properties and its application quantify effective soil hydraulic properties using genetic algorithm. 2. With air-borne remote sensing during SGP97 and SMEX02. Water Resour. Res. beyond continuum scale. The outlook of earth observing systems 45. doi:10.1029/2008WR007022 of various space agencies (e.g., NASA, ESA, JAXA) using satel- Jana, R.B., B.P. Mohanty, and E.P. Springer. 2007. Multi-scale pedo-transfer functions for soil water retention. Vadose Zone J. 6:868–878. doi:10.2136/ lite platforms is promising in the coming decade, and hydrologists vzj2007.0055 and soil scientists should take the opportunity to use them more Jana, R., B.P. Mohanty, and E. Springer. 2008. Bayesian implementation of multi-scale pedo-transfer functions with non-linear bias corrections. Water regularly for mapping soil hydrologic and other properties at the Resour. Res. doi:10.1029/2008WR006879. watershed, regional, continental, and global scales, supporting vari- Jana, R., and B.P. Mohanty. 2011. Enhancing PTFs with remotely sensed data for multi-scale soil water retention estimation. J. Hydrol. doi:10.1016/i.jhy- ous Earth systems science applications. drol.2010.12.043. Jana, R., and B.P. Mohanty. 2012. A comparative study of multiple approaches to soil hydraulic parameter scaling applied at the hillslope scale. Water Re- sour. Res. 48:W02520. doi:10.1029/2010WR010185. Jana, R., B.P. Mohanty, and Z. Sheng. 2012. Upscaling soil hydraulic parameters References in the Picacho Mountain region using Bayesian neural networks. Trans. AS- Ahuja, L.R., O. Wendroth, and D. Nielsen. 1993. Relationship between initial ABE 55:463–473. drainage of surface soil and average profile saturated conductivity. Soil Sci. Soc. Am. J. 57:19–25. doi:10.2136/sssaj1993.03615995005700010005x Lambot, S., E. Slob, J. Rhebergen, O. Lopera, K.Z. Jadoon, and H. Vereecken. 2009. Remote estimation of the hydraulic properties of a sand using full- Ben-Dor, E., S. Chabrillat, J.A.M. Dematte, G.R. Taylor, J. Hill, M.L. Whiting, and waveform integrated hydrogeophysical inversion of time-lapse, off-ground S. Sommer. 2009. Using imaging spectroscopy to study soil properties. Re- GPR data. Vadose Zone J. 8:743–754. doi:10.2136/vzj2008.0058 mote Sens. Environ. 113:S38–S55. doi:10.1016/j.rse.2008.09.019 Leij, F.J., W.B. Russell, and S.M. Lesch. 1997. Closed-form expressions for Burke, E.J., R.J. Gruney, L.P. Simmonds, and P.E. O’Neill. 1998. Using a mod- water retention and conductivity data. Ground Water 35:848–858. eling approach to predict soil hydraulic properties from passive micro- doi:10.1111/j.1745-6584.1997.tb00153.x wave measurements. IEEE Trans. Geosci. Rem. Sens. 36(2):454–460. doi:10.1109/36.662729 Mantoglou, A., and L.W. Gelhar. 1987. Stochastic modeling of large-scale transient unsaturated flow systems. Water Resour. Res. 23:37–46.doi:10.1029/ Burke, E.J., R.J. Gruney, L.P. Simmonds, and T.J. Jackson. 1997. Calibrating a WR023i001p00037 soil water and energy budget model with remotely sensed data to obtain quantitative information about the soil. Water Resour. Res. 33:1689–1697. Mattikalli, N.M., E.T. Engman, T.J. Jackson, and L.R. Ahuja. 1998. Microwave doi:10.1029/97WR01000 remote sensing of temporal variations of brightness temperature and near- surface soil water content during a watershed-scale field experiment, and Camillo, P.J., P.E. O’Neill, and R.J. Gurney. 1986. Estimating soil hydraulic pa- its application to the estimation of soil physical properties. Water Resour. rameters using passive microwave data. IEEE Trans. Geosci. Rem. Sens. GE- Res. 34:2289–2299. doi:10.1029/98WR00553 24(6):930–936. doi:10.1109/TGRS.1986.289708 Mohanty, B.P., and J. Zhu. 2007. Effective hydraulic parameters in horizontally Casa, R.F. Castaldi, S. Pascucci, B. Basso, and S. Pignatti. 2013. Geophysical and and vertically heterogeneous soils for steady state land-atmosphere inter- hyperspectral data fusion techniques for in-field estimation of soil proper- action. J. Hydrometeorol. 8:715–729. doi:10.1175/JHM606.1 ties. Vadose Zone J. 12. doi:10.2136/zj2012.0201. Montzka, C., M. Moradkhani, L. Weihermuller, H.-J.H. Franssen, M. Canty, and Chang, D.-H., and S. Islam. 2000. Estimation of soil physical properties using re- H. Vereecken. 2011. Hydraulic parameter estimation by remotely-sensed mote sensing and artificial neural network. Remote Sens. Environ. 74:534– top soil moisture observations with the particle filter. J. Hydrol. 399:410– 544. doi:10.1016/S0034-4257(00)00144-9 421. doi:10.1016/j.jhydrol.2011.01.020

www.VadoseZoneJournal.org p. 8 of 9 Mulder, V.L., S. de Bruin, M.E. Schaepman, and T.R. Mayr. 2011. The use of Aktivony I Passivnony Radiolokatsii v Meterologii. Gidrometeoizdat, Lenin- remote sensing in soil and terrain mapping- A review. Geoderma 162:1–19. grad. p. 46–50. doi:10.1016/j.geoderma.2010.12.018 van Dam, J.C., J. Huygen, J.G. Wesseling, R.A. Feddes, P. Kabat, P.E.V. van Wal- Peters-Lidard, C.D., D.M. Mocko, M. Garcia, J.A. Santanello, M.A. Tischler, sum, et al. 1997. Theory of SWAP version 2.0. Rep. 71. Dep. Water Resour., M.S. Moran, and Y.H. Wu. 2008. Role of precipitation uncertainty in Wageningen Agric. Univ., Wageningen, the Netherlands. the estimation of hydrologic soil properties using remotely sensed soil Vereecken, H., R. Kasteel, J. Vanderborght, and T. Harter. 2007. Upscaling hy- moisture in a semiarid environment. Water Resour. Res. 44:W05S18. draulic properties and soil water flow processes in heterogeneous soils: A doi:10.1029/2007WR005884 review. Vadose Zone J. 6:1–28. doi:10.2136/vzj2006.0055 Pollacco, J.A., and B.P. Mohanty. 2012. Uncertainties of water fluxes in SVAT Yeh, T.-C.J., L.W. Gelhar, and A.L. Gutjahr. 1985. Stochastic analysis of unsatu- models: Inverting surface soil moisture and evapotranspiration retrieved rated flow in heterogeneous soils: 1. Statistically isotropic media. Water from remote sensing. Vadose Zone J. 11. doi:10.2136/vzj2011.0167. Resour. Res. 21:447–456. doi:10.1029/WR021i004p00447 Pollacco, J.A.P., B.P. Mohanty, and A. Efstratiadis. 2013. The weighted objec- Zhu, J., and B.P. Mohanty. 2002a. Spatial averaging of van genuchten hydrau- tive function selector algorithm and its application to invert surface soil lic parameters for steady state flow in heterogeneous soils. Vadose Zone moisture and evapotranspiration in SVAT models. Water Resour. Res. J. 1:261–272. doi:10.1002/wrcr.20554. Zhu, J., and B.P. Mohanty. 2002b. Upscaling of hydraulic properties for Santanello, J.A., C.D. Peters-Lidard, M.E. Garcia, D.M. Mocko, M.A. Tischler, steady state evaporation and infiltration. Water Resour. Res. 38:1178. M.S. Moran, and D.P. Thoma. 2007. Using remotely-sensed estimates doi:10.1029/2001WR000704. of soil moisture to infer soil texture and hydraulic properties across a semi-arid watershed. Remote Sens. Environ. 110:79–97. doi:10.1016/j. Zhu, J., and B.P. Mohanty. 2003a. Effective hydraulic parameters for steady rse.2007.02.007 state flows in heterogeneous soils. Water Resour. Res. 39(8):1227. doi:10.1029/2002WR001831 Sharma, S.K., B.P. Mohanty, and J. Zhu. 2006. Including topography and vegetation attributes for developing pedo transfer functions in Southern Zhu, J., and B.P. Mohanty. 2003b. Upscaling of hydraulic properties of hetero- Great Plains of USA. Soil Sci. Soc. Am. J. 70:1430–1440. doi:10.2136/ss- geneous soils In: Methods of scaling in soil physics. CRC Press, Boca Raton, saj2005.0087 FL. p. 97–117. Shin, Y., B.P. Mohanty, and A.V.M. Ines. 2012. Soil hydraulic properties in one- Zhu, J., and B.P. Mohanty. 2004. Soil hydraulic parameter upscaling for steady- dimensional layered soil profile using layer-specific soil moisture assimila- state flow with root water uptake. Vadose Zone J. 3:1464–1470. tion scheme. Water Resour. Res. 48:W06529.doi:10.1029/2010WR009581 Zhu, J., and B.P. Mohanty. 2006. Effective scaling factor for transient infiltra- Shin, Y., B.P. Mohanty, and A.V.M. Ines. 2013. Estimating effective soil hydraulic tion in heterogeneous soils. J. Hydrol. 319:96–108. doi:10.1016/j.jhy- properties using spatially distributed soil moisture and evapotranspiration. drol.2005.07.004 Vadose Zone J. 12. doi:10.2136/vzj2012.0094. Zhu, J., B.P. Mohanty, A.W. Warrick, and M.Th. van Genuchten. 2004. Corre- Steenpass, C., J. Vanderborght, M. Herbst, J. Simunek, and H. Vereecken. 2010. spondence and upscaling of hydraulic functions for heterogeneous soils. Estimating soil hydraulic properties from infrared measurements of soil -sur Vadose Zone J. 3:527–533. face temperatures and TDR data. Vadose Zone J. 9:910–924. doi:10.2136/ Zhu, J., B.P. Mohanty, and N.N. Das. 2006. On the effective averaging schemes vzj2009.0176 of hydraulic properties at the landscape scale. Vadose Zone J. 5:308–316. Uri, U., P. Tambet, G.G. Shukin, V.A. Kashevar, Y.V. Kybakov, and Y.M. Shulgina. doi:10.2136/vzj2005.0035 1991. Use of SHF-radiometry during the evaluation of the land-reclamation state of drained earth (based on experiments in Estonia). Trudy Metody