SEARCH

SEARCH BY CITATION

Keywords:

  • soil moisture retrieval;
  • MM5-NOAH LSM;
  • water balance;
  • AMSR-E

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

Land surface models and satellite remote sensing are among the modern soil moisture retrieval techniques that can be used over large areas. However, the lack of ‘ground truth’ soil moisture measurements is still an obstacle in the evaluation and validation of soil moisture retrievals. In this study, a new scheme is used to assess soil moisture retrieval from both the NOAH Land surface Model (LSM) coupled with the fifth generation Mesoscale Model MM5 and the Advanced Microwave Scanning Radiometer AMSR-E. The proposed scheme is based on the strong correlation between changes in soil storage from rainfall runoff events and changes in the retrieved soil moisture either from the MM5-NOAH LSM or the AMSR-E. The aim of this study is to compare the application of the proposed scheme between these two different approaches for soil moisture estimation. The MM5-NOAH LSM provides soil moisture estimations at three different layers with depths 0–10 cm (surface layer 1), 10–40 cm (layer 2) and 40–100 cm (layer 3). In this study, the combined soil moisture over the top two layers (first and second) and the combined soil moisture over the first three layers (first, second and third) are used to account for the entire soil column for assessing the estimated soil moisture using changes in the storage from the water balance. The results have shown that the MM5 soil moisture from the combined top two layers has the better performance than either of the individual layers when compared to the catchment water storage. The results also show that the MM5-NOAH LSM soil moisture estimates have a slightly better performance than the AMSR-E surface soil moisture measurement. This preliminary assessment shows the benefits of using hydrological data in the validation of soil moisture retrieval methods. Copyright © 2013 Royal Meteorological Society


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

Soil moisture is one of the most important variables that integrates much of the land surface hydrology through the water and energy exchanges at the land–atmosphere interface. It affects surface temperature, the depth of the planetary boundary layer (PBL), circulation-wind patterns, and regional water energy budgets (Mahfouf et al., 1987; Lakshmi et al., 1997). Surface soil moisture plays a crucial role in hydrology and meteorology as it is important to estimate the ratio between evaporation and potential evaporation over the land surface, to estimate the distribution of precipitation between runoff and storage, and to computing several key variables of the land surface energy and water budget such as albedo and hydraulic conductivity (Wigneron et al., 2003). Regionally, soil moisture has a controlling function in the hydrological cycle in general through the interaction with the atmosphere and influences the climate. On the other hand, at the field scale soil moisture significantly affects the generation of runoff and erosion, plant growth and other important disciplines in agricultural and environmental fields (Lakshmi et al., 1997).

Soil moisture is a highly variable parameter both spatially and temporally due to the heterogeneity of soil properties, topography, land cover, evapotranspiration and precipitation. As a result, soil moisture is often somewhat difficult to measure accurately in both space and time, especially at large scales (Engman, 1991; Owe et al., 2001). Ground measurements, remote sensing and land surface models are the main sources that could provide soil moisture information. Conventional field measurement techniques have serious limitations in their ability to estimate the spatial distribution of soil moisture appropriately. This is because they are point-based measurements and cannot represent the spatial distribution of the soil moisture particularly over large areas where dense ground networks are needed. However, a significant correlation between the in situ soil moisture observation and the coarse resolution product of soil moisture has been observed in previous studies (Albergel et al., 2010, 2012; Brocca et al., 2010). Remote sensing techniques provide the most feasible capability to monitor soil moisture over a range of spatial and temporal scales (Jackson and Schmugge, 1989). Various remote sensing techniques have been evaluated and proven to be a valuable source of information for the measurement of soil moisture. For the past two decades, microwave remote sensing radiometry has been extensively used for soil moisture monitoring over different conditions of topography and vegetation. Moreover, passive microwave radiometric measurements in the 1–6 GHz range have been very valuable in estimating soil moisture (Calvet et al., 2011). However, remote sensing can observe the spatial distribution of soil moisture only in the top few centimetres of soil surface (Schmugge et al., 1983).

The use of land surface models (LSMs) assimilated with meteorological observations is another approach to estimate soil moisture at regional or global scales through the integration of the atmospheric forcing with the physical formulation of the LSM (Gao et al., 2004). Energy and water exchange occurs continuously at the interface between the land surface and the lower atmosphere. Thus, a strong connection exists between the land surface hydrological processes and the weather. In addition, the conversion of thermal and radiative energy to latent heat is responsible for the linkage between the energy and water balances. Hence, it has been widely accepted that land surface processes play a vital role in the mesoscale atmospheric models that resolve spatial scales from 1 to 100 km (Chen and Dudhia, 2001; Sridhar et al., 2002). As a result, a number of LSMs have been developed in recent years for their application in mesoscale meteorological models in which mesoscale weather is significantly affected by the local variation of temperature and moisture. Increasingly finer spatial and temporal resolutions and improved planetary boundary layer PBL parameterization were the key motivation behind that progress (Chen and Dudhia, 2001). The role of soil moisture in regional weather prediction and the interaction between the atmosphere and the land surface have been demonstrated in many studies (Hipps et al., 1994; Betts et al., 1996; Entekhabi et al., 1996; Chen and Dudhia, 2001; Sridhar et al., 2002; Sahoo et al., 2008; Chen et al., 2010; Patil et al., 2011). It has been shown that physically-based modelling is an important tool for studying the coupling between the LSM and the mesoscale atmospheric model in order to represent the temporal evolution of soil moisture (Reen et al., 2006). LSMs such as the one developed by the National Centre for Environmental Prediction (NCEP)/Oregon State University (OSU)/Air Force/Office of Hydrology (NOAH) (Pan and Mahrt, 1987; Chen et al., 1996) have been coupled with the fifth generation Mesoscale Model MM5.

Most of previous LSM research (either coupled or uncoupled with a meteorological weather model) has been focused on either improving or evaluating the model performance (Liu et al., 2005; Miao et al., 2007; Koster et al., 2009; Van der velde et al., 2009). However, the relevance of surface soil moisture to flood prediction and to a range of agricultural applications was the reason behind a published study by Kong et al. (2011) where the UK Met Office Surface Exchanges Scheme (MOSES) was used to represent land surface processes in the Met Office's Unified Model (MetUM) for soil moisture estimation over an agricultural site in Norfolk. The validation results of MOSES versus ground soil moisture measurements showed that MOSES performs well in soil moisture estimation in general. Another numerical investigation was conducted by Hossain et al. (2005) aimed at evaluating the uncertainty in the prediction of soil moisture from a one dimensional uncoupled land surface model forced by hydro-meteorological and radiation data. A study conducted by Venalainen et al. (2005) concluded that the potential evaporation calculated using the Penman–Monteith equation could be estimated accurately using data obtained from the output of a high resolution numerical atmospheric model (HIRLM, High Resolution Limited Area Model).

It is obvious that there is a lack of research in the evaluation of soil moisture retrievals from coupled LSMs with mesoscale models. Hence, the NOAH LSM coupled with the numerical weather measoscale model MM5 is adopted in this study to estimate and examine the estimation of soil moisture. The NOAH Land surface model was developed by several groups in the US (NCEP, Oregon State University, the Air Force Weather Agency and Air Force Research Lab and the Office of Hydrology Division: see http://www.emc.ncep.noaa.gov/mmb/gcp/noahlsm/Noah_LSM_USERGUIDE_2.7.1.htm).

However, evaluating and assessing the performance of such an LSM is not a trivial exercise due to the lack of high-resolution spatial and temporal ground measurements of soil moisture. Recently, the International Soil Moisture Network ISMN was initiated to overcome many of these limitations (see Dorigo et al., 2011), as the network participants share their soil moisture datasets with the ISMN on a voluntary and no-cost basis. Nevertheless, their data coverage is still very limited. In the UK, only limited data sets of soil moisture measurements exist while at the same time there is an abundance of rainfall and flow measurements. Therefore, from a hydrological point of view, a new scheme is proposed to assess the soil moisture retrieval from satellite (Al-Shrafany et al., 2012), which is further evaluated in this study. This scheme is used to assess and validate the estimated soil moisture values from the three soil layers of the MM5-NOAH LSM and moreover to compare the soil moisture estimations from the MM5-LSM and satellite. The proposed scheme is an empirical approach based on the correlation between changes in the catchment water storage calculated from the water balance equation and changes in the soil moisture retrieved using the NOAH LSM coupled with the MM5. The results from this approach (as compared with in situ soil moisture sensors) are more relevant to hydrological applications.

The key points presented in this paper concern research firstly to retrieve soil moisture using the NOAH LSM coupled with the MM5 over three different depths of soil profile, and secondly to assess the performance of the soil moisture estimations over the different soil depth combinations. An intercomparison between the LSM soil moisture and satellite soil moisture is also shown at the end of this paper.

2. Study area and data

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

The Brue catchment was selected to be the study area due to the availability of hydrological data over a continuous long period and because it was considered to be a representative rural UK catchment used for hydrological modelling. It is located in Somerset, South West England (51.075°N, 2.58°W) with a drainage area of 135 km2 and its terrain elevation ranges between 115 and 190 m (Figure 1). The Brue catchment is predominantly a rural catchment of modest relief as most of the catchment is agricultural land (95.22%) with a few patches of forest (3.12%) and residential areas (1.66%). In terms of soil texture, 49% of the catchment is clay, 29% is fine loam and 21% is silt (Mellor et al., 2000). In this study, the UK Environment Agency provided hourly catchment-averaged rainfall and river flow data for the Brue catchment for the years 2004–2006.

image

Figure 1. Brue catchment site map and digital terrain elevation in UK National Grid Reference (the black dot symbol represents the catchment)

Download figure to PowerPoint

The European Centre for Medium-Range Weather Forecasts (ECMWF) has in the past produced three major re-analyses data sets: FGGE, ERA-15 and ERA-40. The last of these consisted of a set of global analyses describing the state of the atmosphere and land and ocean-wave conditions from mid-1957 to mid-2002. The ERA-Interim as initially described by the ECMWF (Dee et al., 2011) is an ‘interim’ re-analysis for the period 1979-present in preparation for the next-generation extended re-analysis to replace ERA-40 (see ECMWF Newsletters No. 111 and 115). The ERA-Interim products are publicly available on the ECMWF Data Server, at a 1.5° resolution and can be accessed by the ECMWF's Meteorological Archive and Retrieval System (MARS). The ERA-Interim archive is more extensive than that for ERA-40, e.g. the number of pressure levels is increased from ERA-40 23–37 levels and additional cloud parameters are included. In this study, the years 2004 and 2005 of the ERA-Interim re-analysis data with 6 h temporal resolution were used as initial and lateral boundary conditions to drive the Numerical Weather Prediction (NWP) model MM5.

Microwave brightness temperature measurements at 6.9 GHz for the same years (2004–2006) from the AMSR-E on board the Aqua satellite were used for satellite soil moisture estimation. AMSR-E measures brightness temperature at six frequencies (6.9, 10.7, 18.7, 23.8, 36.5, and 89 GHz) with both horizontal (H) and vertical (V) polarizations. More detailed information about the AMSR-E data used in this study is available in Al-Shrafany et al. (2012) and other related general information can be found in the technical paper (Knowles et al., 2006) published by the National Snow and Ice Data Centre (NSIDC). The years used for calibration of the physical models were 2004 and 2005, whereas the validation was performed in the year 2006.

3. Methodologies

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

3.1. MM5-NOAH land surface model configuration

The Fifth-Generation NCAR/Penn State Mesoscale Model MM5 is the latest in a series developed from a mesoscale model used at Penn State in the early 1970s that was later documented by Anthes and Warner (1978). Particular attention has been given to the non-hydrostatic dynamics, multi nesting capability and four dimensional data assimilation. Although mesoscale model users are being encouraged to move on to the next generation mesoscale model, the Weather Research and Forecasting system (WRF) (Skamarock et al., 2005), the MM5 system is still a popular choice due in part to its performance, availability and stability (see for instance Liguori et al., 2009, 2011). Mesoscale models in general and MM5 in particular are often used for three key applications: (1) regional climate simulations, (2) numerical weather prediction, and, (3) air quality prediction. However, the MM5 can be coupled with a Land surface model, and one of the main prognostic parameters is the volumetric soil moisture vsm. The MM5 is a grid based model, using the finite differencing method to resolve the model dynamics at different pressure levels. Detailed information of the MM5 dynamics and its integration can be found in the MM5 online tutorial by Dudhia et al. (2005).

The original LSM was developed at the Oregon State University (OSU) by Pan and Mahrt (1987). It is based on the coupling of the diurnally dependent Penman potential evaporation approach of Mahrt and Ek (1984), the multilayer soil model of Mahrt and Pan (1984), and the primitive canopy model of Pan and Mahrt (1987). It has been modified by Chen et al. (1996) to include an explicit canopy resistance formulation used by Jacquemin and Noilhan (1990) and a surface runoff scheme of Schaake et al. (1996). The NOAH LSM has benefited from a series of improvements particularly in increasing the soil layers from two to four. Hence, it is widely adopted by the National Centres for Environmental Prediction (NCEP) and showed an adequate performance in the NCEP coupled Eta Model. This is one of the reasons why the NOAH LSM was selected to be implemented in the MM5 model besides its moderate complexity (Chen and Dudhia, 2001). The coupled MM5-NOAH LSM model has a vertical soil profile with a total depth of 2 m below the surface and it is partitioned into four soil layers with lower boundaries at 10, 40, 100 and 200 cm below the surface (Figure 2). The root zone is fixed at 100 cm (i.e. including the top three soil layers). Thus, the lower 100 cm of soil layer acts as a reservoir with gravity drainage at the bottom. The MM5-NOAH LSM has one canopy layer and one snow layer and has the following prognostic variables: soil moisture and soil temperature in the soil layers, canopy moisture, snow height, and surface and ground runoff accumulation. Evapotranspiration is handled by using soil and vegetation types. The vegetation characteristics of each grid of the model are represented by the dominant vegetation type of that grid because the model horizontal grid resolution is larger than 1 km × 1 km.

image

Figure 2. Vertical soil profile layers of the MM5-NOAH LSM

Download figure to PowerPoint

The soil thermal properties depend on the soil type. The soil water movement and flow between the soil layers is governed by the mass conservation law and the diffusivity form of Richards' equation (Chen and Dudhia, 2001) as follows:

  • equation image(1)

where θ is the volumetric soil water content, D is the soil water diffusivity (m2 s−1) and K is the hydraulic conductivity (m s−1) and both are functions of θ; t is time (s) and z is the soil layer depth (m); and Fθ is represent sources and sinks for soil water (i.e., precipitation, evaporation and runoff). D and K are highly nonlinearly dependent on the soil moisture (Chen and Dudhia, 2001) and in particular when the soil is dry, they can change several orders of magnitude even for a small variation in soil moisture. By expanding and integrating Equation ((1)) (Chen and Dudhia, 2001) over four soil layers, the following layers are produced:

  • equation image(2)
  • equation image(3)
  • equation image(4)
  • equation image(5)

where, equation image is the soil layer thickness (for layers 1–4 respectively); Pd is the precipitation not intercepted by the canopy; equation image is the canopy transpiration taken by the canopy root within the root zone layers (the root zone is up to three layers in the coupled MM5-LSM); and Edir is the direct evaporation from the top surface soil layer. A conceptual parameterization for the sub-grid treatment of precipitation and soil moisture is governed by the infiltration. The heat transfer through the soil vertical profile is governed by the thermal diffusion equation. A single linearized surface energy budget equation is applied to determine the surface temperature to reflect a linearly combined ground-vegetation surface (Chen and Dudhia, 2001; Chen et al., 2010). The surface runoff is calculated using the Simple Water Balance (SWB) technique. A detailed description of the MM5-NOAH LSM can be found in (Chen and Dudhia, 2001).

For the purposes of this study, the MM5 model was set up to have three domains (D1, D2 and D3) with grid resolutions of 108 km for the outside domain, 36 km for the middle domain and 12 km for the inner domain. The innermost domain D3 is able to capture the local scale features of the study area. The MM5 domains are nested with two-way interaction, in which the boundary conditions for the finer grid are generated from the coarse grid model results while the fine grid model results update the variables on the coarse grid (Dudhia et al., 2005/MM5 tutorial). In addition, the standard nesting ratio used by MM5 in every time step is 3:1, in which each domain takes information from the mother domain. For each mother domain time step, the domain runs three time steps before feeding back information to the mother domain. Hence, nested domains feeding back to each other can lead to improved model behaviour at the boundaries. In order to mitigate the spatial distortion associated with the map projection applied, the domains are positioned in such a way that the Brue catchment is located at the centre of all three domains (see Figure 3).

image

Figure 3. Configuration of the MM5-NOAH LSM domains D2 and D3 to produce a 12 km resolution (D3) soil moisture over the Brue

Download figure to PowerPoint

The model is configured to have 23 vertical levels with thinner layers near the ground surface (the model top is at 100 hPa) and a dimensionless quantity σ is used to define the model's vertical levels as it is shown in the following equation:

  • equation image(6)

where p0 is the reference-state pressure, pt is a specified constant top pressure, and ps0 is the reference-state surface pressure.

Four Dimensional Data Assimilation (FDDA) is used in which a full-physics of the model is run while incorporating observations/re-analysis. In this way a dynamical consistency would be assured while the observations keep the model close to the true conditions (Stauffer and Seaman, 1990). Newtonian-relaxation or nudging is the technique used by the MM5 model in such data assimilation scheme. In this technique, an artificial tendency term based on the difference between the model state and the observed state is added to the prognostic equations for a particular variable such as wind, temperature and water vapour in order to relax the model state towards the observed state or a given analysis. For this study, the grid nudging is implemented through feeding the model in its standard input format with the given re-analysis on the model grid over the data assimilation period. Thus, the model relaxes its solution towards the re-analysis data. The Nudging Factor Gα (where α represent a particular variable) determines the relative magnitude of the term that is added into the variable prognostic equation. In this study, three dimensional analysis nudging is implemented for the wind and temperatures fields and their nudging factor values are selected according to the study done by Stauffer and Seaman (1990).

The cumulus parameterization scheme used in this study for the MM5 model was the Kain–Fritsch (KF), which is based on relaxation to a profile and predicts both updraft and downdraft properties and also detrains cloud and precipitation. This cumulus parameterization scheme uses a sophisticated cloud-mixing scheme in order to determine entrainment/detrainment. More details can be found in Kain and Fritsch (1993). For the Planetary Boundary Layer PBL, the MRF or Hong-Pan scheme was the selected option due to its suitability for high resolution in PBL. This scheme was implemented in the NCEP MRF model (see Hong and Pan, 1996 for more details). Mixed-Phase was the selected scheme that dealt with the microphysics of the model. This scheme adds supercooled water to cloud and rain water field that predicted explicitly the microphysical processes. The NOAH LSM was used to retrieve the surface parameters and in particular soil moisture (see Chen and Dudhia, 2001).

3.2. MM5-NOAH LSM soil moisture estimation

Numerical experiments were conducted in this study for several events in 2004–2006 to predict a continuous time series of the soil moisture for three soil layers. The ECMWF/Era-interim/reanalysis data with a spatial resolution of 1.5° × 1.5° and a temporal resolution of 6 h was used in this study as initial and lateral boundary conditions. The boundary conditions are fed on to the coarsest domain which has a comparable resolution (108 km) and it is then dynamically downscaled all the way down, from 108 km for domain 1–36 km and 12 km for domains 2 and 3 respectively. The smallest resolution of 12 km is consistent with the area covered by the Brue catchment area. The classification of vegetation by the U.S Geological Survey (USGS) is adopted in the MM5-NOAH LSM to define the vegetation types that cover the study area, whereas the soil types are defined by the Food and Agriculture Organisation (FAO) database. Default values of the model parameters such as the soil and vegetation parameters were selected in this study. All simulations are conducted from 0000 UTC 1 January 2004 to 0000 UTC 31 December 2004 for the first simulation and similarly for the follow-on simulations using 2005 and 2006 data. The model output was retrieved at hourly intervals. As a result, 3 year hourly time series of three soil layers (with layer thicknesses of 10, 30, 60 cm) of soil moisture values are estimated from the three MM5-NOAH LSM domains. The innermost domain (domain 3) soil moisture was adopted in this study (see Figure 4(c)), as domain 3 has the highest and required spatial resolution of 12 km, which has a similar area to the Brue catchment.

image

Figure 4. Brue catchment (2004–2005): (a) evapotranspiration (ET), (b) precipitation (c) volumetric soil moisture (VSM) from the MM5 LSM for three single layers with soil depths: surface layer (0–10 cm), second layer (10–40 cm) and third layer (40–100 cm); and (d) VSM from the MM5 LSM combining several layers; (e) AMSRE satellite soil moisture time series

Download figure to PowerPoint

A three-level soil layer configuration was adopted in the coupled MM5-NOAH LSM in this study in order to capture the daily, weekly and seasonal evolution of soil moisture and mitigating the possible truncation error in discretization (Sridhar et al., 2002). The combination of soil moisture over several layers was adopted here to take into account the entire depth of the soil column when using the water balance changing in the storage to assess the estimated soil moisture from the NOAH LSM. It is assumed that the soil is homogenous and has no significant variations in its characteristics. Hence, two cases of layer combinations were produced. Firstly, the soil moisture was combined over the first two layers (first and second) of the NOAH LSM where the soil column depth would be 40 cm. Secondly, the soil moisture over the three first layers (first, second and third) was combined with a total soil column depth of 100 cm. The soil moisture in the combined layers is computed by:

  • equation image(7)
  • equation image(8)

where θ is the original soil moisture estimation from the LSM, z is the soil layer depth, and the subscripts (1, 2 or 3) indicate the soil layer indices.

For comparison purposes of the MM5-NOAH LSM soil moisture assessment against the retrieved soil moisture from satellite, a third set of numerical experiments was conducted for some selected events in 2006. The same simulation settings were used for the 2004, 2005 and 2006 simulation experiments. A statistical t-test was conducted to examine the significance of the linear correlation between the changing in the water storage and the changes in the MM5-NOAH LSM soil moisture. In such a test, the p-value, which measures the strength of evidence, is computed. A value of 0.05 was adopted as the significance level. If the computed p-value is below the significance level then there is a significant linear relationship between the change in the storage and change in the estimated soil moisture from the MM5-NOAH LSM, otherwise there is no statistical significance. The t-test results are presented later in this paper.

3.3. AMSR-E soil moisture estimation

The explanation of estimation of near surface soil moisture from the AMSR-E satellite is summarized in this section. It is important to mention that a detailed description of the developed methodology to estimate soil moisture from satellite can be found in Al-Shrafany et al. (2012). The AMSR-E instrument is a passive microwave radiometer that makes measurements of thermal radiation from the land surface in the centimetre wave band at ascending (1300) and descending (0130) passes (Owe et al., 2001; Njoku et al., 2003). The AMSR-E measurements from the descending pass were considered in this study. This is due to the reasonable uniform temperature and moisture profiles at night times, while in the daytime soil moisture estimates may reflect the effects of diurnal surface layer drying. The techniques adopted for soil moisture retrieval provided spatially averaged soil moisture data, which is ideal for environmental and hydrological modelling and monitoring. Such spatially averaged area data sets are logistically and economically difficult to obtain through traditional in situ measurement techniques. The technique only uses the horizontal and vertical polarization brightness temperatures Tb at one frequency (6.9 GHz) observed by the AMSR-E in a descending mode. The use of lower frequencies (e.g. 6.9 GHz) allows a greater penetration depth and the measurements are less affected by the vegetation (Schmugge et al., 2002).

The AMSR-E satellite has a footprint size of 25 km at which all retrieval calculations are based on. With respect to the average soil and vegetation biophysical characteristics, a uniform footprint is assumed. Therefore, the surface soil moisture and vegetation optical depth are subsequently extracted as averaged footprint values. The soil and vegetation temperatures are assumed to be approximately equal in the use of the AMSR-E descending measurements as the temperature and soil profiles are reasonably uniform. Moreover, the effects of the atmospheric moisture and the multiple scattering in the vegetation layer are negligible due to the AMSR-E low frequency measurements (up to X-band, i.e. ∼10 GHz). The radiative transfer equation explains the relationship between surface parameters such as surface soil moisture, vegetation water content, surface temperature, and microwave brightness temperature (Tb) (Jackson et al., 1982; Njoku et al., 2003). It includes contributions from the soil, vegetation and atmosphere in the upwelling radiation from the land surface as observed by the instrument. The brightness temperatures at H and V polarizations are given by:

  • equation image(9)
  • equation image(10)

where the subscripts H and V refer to the horizontal and vertical polarizations respectively; Ts is the single surface temperature; es is the soil emissivity at H (esH) and V (esV) polarizations; ω is the vegetation single scattering albedo and Γ is the transmissivity. The contributions from soil and vegetation represented by the surface roughness and density of vegetation canopy respectively have significant effects on the soil reflectivity. Surface roughness reduces the sensitivity of emissivity to soil moisture variations, and thus reduces the range in measurable emissivity from dry to wet soil conditions (Wang, 1983). The statistical parameters that characterize the scale of roughness of a randomly rough surface are known as the h and Q parameters. The h-Q model developed by Wang and Choudhury (1980) was considered in Al-Shrafany et al. (2012) to account for the roughness effects when the soil moisture was retrieved from the AMSR-E.

The vegetation will absorb or scatter the radiation emanating from the soil, and it will also emit its own radiation. Generally speaking, the integral contribution of the surface roughness and vegetation canopy is more difficult to separate unless one of them is known a priori. An analytical approach developed by Meesters et al. (2005) is considered in Al-Shrafany et al. (2012) for calculating vegetation optical depth from the Microwave Polarization Difference Index (MPDI) and the dielectric constant of the soil. The MPDI effectively normalizes out the effects of the surface temperature, resulting in a quantity that is highly dependent on the soil moisture and vegetation. The MPDI is defined as:

  • equation image(11)

Hence, the brightness temperatures are converted to volumetric soil moisture values with the Land Parameter Retrieval Model LPRM (Owe et al., 2008; Wang et al., 2009). The volumetric soil moisture vsm is retrieved from Equation ((11)) (after substituting in Equations ((9)) and ((10)) and taking into account of the roughness effect) as:

  • equation image(12)

where esH and esV is the soil emissivity at H and V polarization respectively, τ is the vegetation optical depth. The h and Q parameters are empirically calibrated since the lowest frequency of the AMSR-E instrument is at 6.9 GHz, and its footprint scale is large which results in no data available to quantify the regional variability of the those parameters. Therefore, in order to estimate the optimal (h and Q) values for a particular catchment area, a new approach is proposed in Al-Shrafany et al. (2012) for this purpose. That approach used the event-based water balance approach in the context of catchment storage calculation. Hence, for a range of h and Q values, the volumetric soil moisture was retrieved from the AMSR-E using Equation ((12)) and taking into account the best correlation between changes in the water storage Δs (from the water balance using rainfall, runoff and evapotranspiration) and changes in the satellite soil moisture Δθ. The optimal values for the h and Q parameters were obtained when the best correlation between Δs and Δθ was achieved. Al-Shrafany et al. (2012) also showed that changes in the volumetric soil moisture were very sensitive to the selection of the h parameter, but less sensitive to the selection of the Q parameter. This method was considered as a potential technique to assess the retrieved soil moisture from the AMSR-E satellite for hydrological applications. In this current study, the same approach is also adopted to assess the estimated soil moisture from the three layers of the MM5-NOAH LSM. Therefore, a brief summary of the event-based water balance approach is presented in the next section.

3.4. Water balance as a proposed scheme

Water balance is a modelling framework for simplifying, describing and quantifying the hydrological water budget. It can be applied to a catchment area within a time interval (annual, monthly, weekly). Water balance is hydrologically driven by the variation in precipitation and temperature, besides some other local factors such as vegetation, soil and land use. A water balance can be used to help manage water supply and predict where there may be water shortages. It is also used in irrigation, runoff assessment (e.g. through the Rainfall-Runoff model), flood control and pollution control. In this study, the water balance is the basis of the proposed scheme and offers a preliminary validation tool in the hydrological and meteorological community for the retrieved soil moisture from the MM5-NOAH LSM. This scheme was developed due to the lack of ground in situ measurements of soil moisture in the UK and most other places around the world. The scheme can be used to calibrate and validate soil moisture estimations in large areas due to the abundance of hydrological data (rainfall and flow) such as the Brue catchment. The performance of this scheme is based on the correlation between the change in catchment water storage Δs and the corresponding change in soil moisture Δθ retrieved from the MM5-NOAH LSM approach calculated on an event basis.

The water balance equation is given by:

  • equation image(13)

where P is the precipitation in mm, R is the runoff volume in mm, ET0 is the evapotranspiration in mm and Δs is the change in soil water storage in mm. The water balance takes into account the main hydrological processes taking place within the catchment. The evapotranspiration (ET0) is calculated with a method called the Penman Monteith equation recommended by the Food and Agriculture Organisation (FAO) (Allen et al., 1998) and it is used in this study to incorporate the impacts of the evapotranspiration in the water storage calculations. The FAO Penman–Monteith method requires the following meteorological data: solar radiation, air temperature, air humidity and wind speed, to derive the parameters for calculating ET0. The observed meteorological data for the Brue catchment were obtained from the British Atmospheric Data Centre BADC. The mathematical formulation of the Penman-Monteith equation and all the related calculation procedures can be found in the FAO report published by Allen et al. (1998).

The water balance application in this study is an event-based approach. The rainfall-runoff events have been chosen over a 3 year period (2004–2006) and for each selected event, the total runoff volume (both direct runoff and base flow) can be calculated using the measured flows. Several selected rainfall-flow events were used to assess soil moisture estimation from the MM5-NOAH LSM (see Tables 1 and 2). The change in the soil moisture Δθ is given by:

  • equation image(14)

where Δθ is the change in the vsm, θ1 is the vsm before the event and θ2 is the vsm after the event. Assuming a sufficient number of rainfall-runoff events, the correlation between Δs and Δθ can be calculated. The following section explains in detail the result of this analysis.

Table 1. Results summary of Δs (change in the storage) and Δθ (changes in the MM5-NOAH LSM soil moisture estimation) over three soil layer depths for the Brue catchment in (2004 and 2005)
Flow eventDurationTotal rain (mm)Run off volume (mm)Total ET (mm)ΔS (mm)Δθ (%) at single LSM layersΔθ (%) at combined LSM layers
2004Start date, timeEnd date, time (0–10) cm(10–40) cm(40–100) cm(0–40) cm(0–100) cm
January_16 January 020011 January 000027.816.64.27.00.750.751.501.301.65
January_211 January 010013 January 010028.416.14.28.11.001.000.750.670.93
February_11 February 02003 February 010014.711.23.50.030.000.000.000.130.13
February_26 February 02008 February 020031.920.54.47.00.500.751.750.570.98
March_111 March 010018 March 020022.24.54.213.51.252.001.751.651.60
March_218 March 030023 March 020014.64.97.42.30.751.251.750.961.15
April21 April 020024 August 020013.85.27.41.20.250.250.750.210.60
May4 May 01008 May 020034.811.810.312.71.501.250.501.010.85
June22 June 020026 June 000033.41.415.716.31.502.000.751.521.38
July7 July 010013 July 020037.43.719.414.32.252.002.251.722.48
August2 August 01006 August 000025.50.914.89.80.250.500.000.620.27
September17 September 230021 September 00004.80.810.1− 6.10.000.000.75− 0.450.82
October_12 October 02006 October 230031.24.48.118.71.002.001.251.681.52
October_213 October 010016 October 000011.54.94.71.90.500.751.250.611.02
November.20 November 020023 November 230012.57.23.51.80.250.251.500.210.80
December18 December 030021 December 000036.118.92.814.41.201.001.700.690.48
2005           
January_110 January 010013 January 230011.07.23.50.30.750.500.750.480.85
January_222 January 000025 January 020010.06.72.70.60.250.250.000.220.10
February_15 February 01008 February 230022.88.93.710.20.750.750.250.800.10
February_210 February 020014 February 010024.410.74.88.91.001.251.751.221.60
March29 March 01001 April 010033.58.34.720.51.752.502.752.232.65
April_117 April 020020 April 020023.46.36.710.41.501.501.501.271.50
April_226 April 000030 April 230014.57.37.7− 0.50.25− 0.25− 1.00− 0.45− 0.70
May19 May 230023 May 000036.54.89.7222.002.751.501.891.78
June_15 June 01008 June 010014.14.27.62.30.750.750.500.640.60
June_224 June 020027 June 020019.35.511.32.50.25− 0.500.00− 0.18− 0.35
July24 July 010026 June 020019.20.911.96.40.751.750.501.120.93
August13 August 000015 August 230015.80.610.54.70.500.250.000.260.15
September10 September 030013 September 00008.80.49.6− 1.20.500.250.250.620.43
October_112 October 020015 October 010019.01.65.711.71.251.500.751.250.73
October_229 October 00001 November 230015.65.07.92.70.751.000.500.660.63
November6 November 01008 November 230024.915.64.64.71.000.501.250.471.00
December1 December 01006 December 020065.642.04.519.11.251.252.001.531.70
Table 2. Results summary of Δs (change in the storage) and Δθ (changes in the MM5-NOAH LSM soil moisture estimation) over three soil layer depths for the Brue catchment in (2006)
Flow eventDurationTotal rain (mm)Run off volume (mm)Total ET (mm)ΔS (mm)Δθ (%) at single LSM layersΔθ (%) at combined LSM layers
2006Start date, timeEnd date, time (0–10) cm(10–40) cm(40–100) cm(0–40) cm(0–100) cm
January30 December 2005 00005 January 2006 020033.616.83.813.00.380.410.560.390.49
February14 February 010019 February 020026.211.83.411.01.000.311.340.480.99
March_17 March 000013 March 020030.814.44.611.80.850.76− 1.470.78− 0.57
March_226 March 010028 March 23008.44.14.00.3− 0.50− 0.150.71− 0.240.33
April29 March 23004 April 020021.47.76.96.8− 0.250.04− 0.03− 0.03− 0.03
May21 May 000031 May 020065.821.410.234.21.251.582.111.491.86
June25 June 020029 June 010019.21.712.15.4− 0.50− 0.41− 0.29− 0.43− 0.35
July5 July 010010 July 000039.01.814.622.60.350.88− 1.420.75− 0.55
August28 August 010031 August 230020.61.313.36.0− 1.00− 0.33− 0.08− 0.49− 0.25
September28 September 23005 October 020027.62.59.915.21.001.801.601.601.60
October19 October 000028 October 020077.828.16.443.33.254.346.574.075.57
November23 October 000029 October 230051.033.74.213.11.250.993.871.052.74
December_13 December 01006 December 000015.29.33.32.60.750.470.890.540.75
December_210 December 020015 December 020018.314.32.91.10.000.510.000.380.15

4. Results and discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

4.1. MM5-NOAH LSM soil moisture results

Numerical experiments were conducted for several events in 2004–2006 (2 years of data were used to assess the MM5-NOAH LSM scheme and 1 year in 2006 was used to compare with the satellite soil moisture retrieval). As a result, a 3 year time series of soil moisture values were estimated from the three MM5 domains. The soil moisture results produced from domain 3 were adopted in this study as presented in Figure 4. It can be seen from this figure that for all the three layers, the soil moisture predictions are consistent with the observed rainfall and evapotranspiration dynamics. Furthermore, the surface layer (0–10 cm) soil moisture is more variable in comparison to the other two deeper layers as it rapidly responds to changes in rainfall and evapotranspiration. The results show that 2004 had the ‘wetter summer’ compared with 2005 as it had a total rainfall over the summer (June, July and August) of around 311.8 mm, while for 2005 the total rainfall for the same period was 167.2 mm. There is no significant increase in the evapotranspiration measurements over that year except for some sporadic cases. Thus, relatively high soil moisture values are shown in late summer and autumn. In addition, the results for the year 2004 also show the typical climatological annual cycle of soil moisture.

Spatial, temporal and vertical resolutions are important issues that need careful consideration when comparing soil moisture estimations from different sources for assessment and validation purposes. Field measurements are usually obtained at point scales whereas the MM5-NOAH LSM simulates the land surface processes averaged over 12 km spatial grids. From the aforementioned water balance scheme, event-based Δs calculation was adopted here, where 33 rainfall-runoff events were selected over a period of 2 years 2004–2005 (see Table 1). For each selected event, Δθ was calculated for all three NOAH LSM soil layers by applying a simple subtraction process (see Equation ((14))) between the soil moisture values before and after an event. Hence, the correlation between Δs and Δθ for all 33 selected events was computed for the three soil layers and the results are shown in Figures 5(a)–(c). It can be seen from this figure that there is a good correlation in terms of R2 between Δs and Δθ at the surface and second soil layer about (0.67) and (0.70) respectively (Figure 5(a) and (b)). However, the best correlation is achieved at the second soil layer depth (30 cm) as shown in Figure (5b). This is because soil depth has a larger contribution at the second layer in the soil moisture calculations, which in turn offers a better consistency between changes in the water balance storage and the changes in the calculated soil moisture. Nevertheless, the correlation R2 result over the third soil layer was poor (R2 = 0.3) (see Figure 5(c)). One reason for that is the vegetation root zone over the Brue catchment is up to 30 cm as it is dominated by grass vegetation type and this in turn controls the evapotranspiration rate which is one of the main components in the water balance calculation. One key function of plant roots within the soil–plant–atmosphere system is to connect the soil environment to the atmosphere by providing a link in the pathway for water fluxes from the soil through the plant to the atmosphere. Fluxes along the soil–plant–atmosphere interaction are regulated by above-ground plant properties such as the leaf stomata, which can regulate plant transpiration when interacting with the atmosphere. Therefore, it can be said that Δθ is calculated over a soil profile depth up to 40 cm. A combination of the model estimated soil moisture over different layers was conducted in order to account for the entire depth of the soil column which in turn provided a proper representation of the correlation between the change in the water balance storage and the change in the NOAH LSM soil moisture estimation. Therefore, two additional cases were considered by combining soil moisture estimations from different layers. In the first case the estimated soil moisture from the first two layers (first and second) of the NOAH LSM was computed following Equation ((7)), so that the combined soil moisture represents the volumetric water content in the soil column depth up to 40 cm. In the second case, the estimated soil moisture from the first three layers (first, second and third) of the NOAH LSM was computed following Equation ((8)), so that the combined value of soil moisture represents the volumetric water content in the soil depth up to 100 cm. The same aforementioned process used to assess the estimated soil moisture from NOAH LSM as single layers was applied to assess the combined soil moisture produced from these two new cases and their results are presented in Figures 5(d) and (e).

image

Figure 5. Correlation between the change in the storage and the change in the MM5/LSM soil moisture in 2004 and 2005 at (a) surface layer (0–10 cm); (b) second layer (10–40 cm); (c) third layer (40–100 cm); (d) combined layer (first and second) and (e) combined layer (first, second and third) layers of NOAH LSM

Download figure to PowerPoint

It is obvious from Figure 5(d) that R2 is higher around 0.75 if we combine soil moisture layers 1 and 2 rather than using individual soil moisture layers (see Figure 5(a) and (b)). In this case, both the catchment water storage Δs calculated from the water balance equation and the combined soil moisture Δθ from the first and second soil layers of the NOAH LSM better represent the change in the water content over a soil column depth up to 40 cm This result clearly indicates the vital role of taking into account several soil layers for assessing and evaluating soil moisture estimations. On the other hand, if we combine the three top soil layers with a total soil depth of 100 cm with the catchment water storage, the results are poorer (R2 = 0.41). This is mainly because of the vegetation root zone limitation as discussed above.

A similar discussion can also explain the results produced for the validation year 2006 (see Figures 6 and 7, and Table 2). Figure 7 clearly shows that the combination of the first two layers of the NOAH LSM in 2006 produced the best correlation (R2 = 0.74) between Δs and Δθ (see Figure 7(d)). As the simulation has been run for 1 year, it is important to examine the significance of the linear correlation between the event-based calculations between the changing in the water storage and the changing in the estimated volumetric soil moisture from the MM5-NOAH LSM. Therefore, a statistical t-test was conducted to examine the significance of the resulted linear correlation between Δs and Δθ. A usual significance level of 0.05 was adopted in this study and the p-values are computed for every case. All the estimated p-values were shown to be below the significance level, which in turn indicates that there is a significant linear correlation between Δs and Δθ although the results from layer 3 as a single soil layer do not show a good result (R2 = 0.37) due to the vegetation root zone limitation (see Figure 7(a)–(e)). It is important to mention that the correlation was used as a performance indicator given the fact that the units of Δs and Δθ are in mm and m3 m−3 respectively and the use of other performance measures such as the root mean square error would not be appropriate.

image

Figure 6. Volumetric soil moisture (VSM) time series from the MM5-NOAH LSM in 2006 for: (a) three single layers with soil depths: (surface layer (0–10 cm), second layer (10–40 cm) and third layer (40–100 cm)); (b) two cases of layers combinations (first and second) layers and (first, second and third) layers; (c) AMSR-E satellite soil moisture time series

Download figure to PowerPoint

image

Figure 7. Correlation between change in the storage and change in the MM5/LSM soil moisture in 2006 at (a) surface layer (0–10 cm); (b) second layer (10–40 cm); (c) third layer (40–100 cm); (d) combined (first and second) and (e) combined (first, second and third) layers of NOAH LSM

Download figure to PowerPoint

Soil moisture content usually increases when there is more free water produced from high rainfall rates. As a result, Δs increases in the winter season which in turn cause increases in Δθ while in the summer season Δs decreases causing a decrease in Δθ. However, the evapotranspiration plays an important role in controlling the mean flow and in turn the change in water storage Δs. Overall, Δθ increases with Δs over the three MM5-NOAH LSM layers whether in single or combined mode. Table 1 shows that at the surface soil layer (0–10 cm), the highest values for Δs and Δθ were found during the winter and spring due to the high rainfall rate and low evapotranspiration rate which in turn lead to an increase in soil water storage (see events December 2005 and March 2004 and 2005). A similar condition was also observed over the second soil layer (10–40 cm). However, a significant increase in the Δs values was observed during the summer for some particular events such as in June and July 2004 due to the rainfall rates exceeding the evapotranspiration rate and consequently increasing the value of Δs over the catchment. In regards to the third soil layer (40–100 cm), Table 1 shows that the Δθ values did not change over some particular events despite of the significant change in the corresponding Δs values (see events February, March and October 2004). This indicates that Δθ is not representative of the soil moisture within the third soil layer. One of the potential reasons is that the evapotranspiration rate is controlled by the catchment vegetation root zone which is extended in the soil up to 30 cm due to the type of the dominant vegetation over the Brue catchment.

In terms of soil layer depth, Table 1 shows that the deeper the soil layer is the higher the water content is available in the winter season while the inverse occurs during the summer season as the water content level becomes lower when the soil layer becomes deeper. Consequently, Δθ has the highest values in winter and the lowest values in the summer particularly at the third layer of LSM (40–100 cm). When a combination of different soil moisture layers is used, similar results are shown (see Table 1), with higher values of Δθ in winter and lower values of Δθ in summer. The results of combining layers 1–2 with a depth of 40 cm were better than those combining layers 1–2–3 with a depth of 100 cm due to the vegetation root zone limitation that affects the evapotranspiration rates and in turn the estimated soil moisture over the soil layer depth. Similarly, the results from 2006 using single and combined NOAH LSM soil layers (see Table 2) corroborate the results explained above.

In this study, the estimated soil moisture from the coupled NOAH LSM with the numerical weather model MM5 over three single soil layers depths agrees with the findings of Kong et al. (2011) in terms of the consistency of the predicted soil moisture for all the three LSM soil layers with the observed rainfall and evapotranspiration as shown in Figure 4, although the study by Kong et al. (2011) was conducted using an uncoupled land surface model known as MOSES.

4.2. Intercomparisons against AMSR-E soil moisture retrieval

Three years of descending dual polarized AMSR-E brightness temperatures at 6.9 GHz were converted into the volumetric soil moisture vsm using the Land Parameter Retrieval Model LPRM. As a result, the simulated MPDI time series was iteratively computed through the LPRM for the range of vsm (0–0.80) and the retrieved soil moisture over the Brue catchment can be seen in Figure 4(e) in the context of time series. This figure shows that on average, the mean annual volumetric soil moisture is around 10.5%. The lowest values were found during the summer (May, June and July), which were around 5%. Larger values (around 47%) were found during the winter (December, January and February). The two surface roughness parameters h and Q were calibrated using the water storage change. Two years (2004–2005) of hydrological data were used for calibration of these parameters and 1 year (2006) data was used for validation purposes. The optimal h and Q parameters were found by maximizing the correlation between Δs and Δθ. The correlation results during the calibration and validation phase are shown in Figure 8(a) and (b) respectively. It can be seen from those figures that there is a good correlation (R2 = 0.74) and (R2 = 0.71) between changes in the storage and changes in the AMSR-E soil moisture for calibration and validation datasets respectively.

image

Figure 8. Correlation relationship between change in the storage Δs and changes in AMSR-E soil moisture Δθ for all events included within: (a) calibration and (b) validation data sets

Download figure to PowerPoint

In terms of comparing the MM5-NOAH LSM and the AMSR-E soil moisture estimations for hydrological applications, the results from the 2006 simulations are adopted (see Figures 6 and 7(d)). The combination of the first two layers of the NOAH LSM (see Figure 6(b)) was used due to its best performance in terms of correlation between Δs and Δθ. Therefore, those results were compared with the validation results of the satellite surface soil moisture estimation (see Figure 8(b)). The comparison result shows that the correlation result between Δs and Δθ in the MM5-NOAH LSM was slightly better than that predicted from the AMSR-E satellite. However, both methods agreed that the proposed scheme of water balance application can potentially be a useful tool in the assessment and validation process of the soil moisture estimation from both satellite and land surface model that coupled to a numerical weather model.

Four important issues have to be taken into account when comparing the MM5-NOAH LSM soil moisture estimation against the AMSR-E satellite soil moisture retrieval, which are as follows.

  1. Soil depth. One significant advantage of conducting MM5-NOAH LSM for estimating long term soil moisture time series in comparison with satellite soil moisture estimation is the use of deep soil depths in the former technique (i.e. applicability of estimation soil moisture over three layers with different depths 10, 30 and 60 cm) and the ability to combine these layers. The AMSR-E soil measurement reflects only the moisture content of the microwave soil moisture sampling depth which is at most within a range of 0–2 cm due to the microwave wavelength limitations. Therefore, only the surface soil layer can be detected by the satellite techniques which in turn pose a significant limitation on the use of these soil moisture observations despite the importance of the surface soil moisture in some particular application such as hydrology and agriculture.

  2. Horizontal resolution. Soil moisture estimations represent average values over the horizontal retrieval area (i.e. area-averaged values) in both approaches for estimating soil moisture. The coupled MM5-NOAH LSM used several high resolution fields that characterize land surface conditions (Chen and Dudhia, 2001) such as vegetation, water and soil characteristics at fine scales and capture the dynamics of the associated land surface forcing in order to meet the increasing demand for employing high resolution mesoscale models up to 1 km horizontal resolution and in turn offers high resolution soil moisture estimations. In comparison, the AMSR-E satellite data provided surface soil moisture result at 25 km horizontal resolution which is good for global retrievals. However, soil moisture results retrieved from the AMSR-E over the Brue catchment showed a good consistency with the observed rainfall and flow.

  3. Range of soil moisture values. Figures 4 and 6 have shown that the range of soil moisture values was large and significant in the retrieved soil moisture from the AMSR-E satellite between (5%) as a minimum value and (48%) as a maximum value in comparison with those estimated from the MM5-NOAH LSM. One potential reason is the time drift problem (i.e., error accumulation) that is not present in the satellite measurements as the satellite provides an instant measurement of brightness temperatures of the soil surface. Hence, the satellite measurement represents the instant state of the surface soil moisture at a particular time. It is found that despite of the existence of some high rainfall rate during the summer season, surface soil moisture observed low values as would be expected in such warm weather. In contrast, the calibration and validation results of the MM5-NOAH LSM combined layer estimation in Figures 4(d) and 6(b) showed that the soil moisture range is small between (23.8%) as a minimum value during summer and (42.5%) as maximum value during winter. As a result, it is obvious that the MM5-NOAH LSM soil moisture values are larger than the corresponding values retrieved from the AMSR-E satellite which in turn stresses the existence of the time drift problem in the hydrological and land surface models application.

  4. Assessment and validation scheme. It was mentioned at an early stage of this study that there is a lack of validation datasets that are spatially representative of soil moisture observations. This was the motivation to propose hydrological-based schemes for assessing the soil moisture estimation from the AMSR-E satellite and the MM5-NOAH LSM. Event-based water balance serves well in assessing the estimated soil moisture from the AMSR-E satellite as it was shown by the good correlation between Δs and Δθ (see Figures 8(a) and (b)). In contrast, the combined product (40 cm depth) of the MM5-NOAH LSM soil moisture showed a slightly better correlation between Δs and Δθ (see Figures 5(d) and 7(d)). Therefore, the combined product of soil moisture estimations from the MM5-NOAH LSM outperforms the soil moisture estimations from the AMSR-E satellite.

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

Soil moisture estimation is a key variable in many disciplines especially in the hydrological and meteorological fields. In this study, a coupled NOAH LSM with the mesoscale model MM5 was adopted to estimate soil moisture at three soil layer depths (10, 30 and 60 cm) as a first attempt to estimate soil moisture from a coupled land surface model for hydrological purposes. As a case study, the coupled NOAH LSM was applied to a vegetated site (the Brue catchment) located in the South West of England. The results revealed that in general the three layer soil moisture estimation of the MM5-NOAH LSM had a good agreement with the observed rainfall and evapotranspiration dynamics. However, comparing the MM5-NOAH LSM soil moisture simulations with the available observed rainfall and evapotranspiration is not adequate to assess the estimated soil moisture performance. Soil moisture estimation assessment and validation is a challenge in the hydrological community as it is difficult to measure accurately in both time and space. From the fact that hydrologists are primarily interested in runoff and water budgets, a new empirical scheme based on the water balance change that has been firstly proposed by Al-Shrafany et al. (2012) was applied in this study to assess soil moisture retrieved from the MM5-NOAH LSM. Two sets of numerical experiments were conducted for several cases for 2004 and 2005. Three layers of soil moisture values at different depths (10, 30, and 60 cm) were retrieved from the configured simulations. The results from the inner MM5 domain have a 12 km spatial resolution, which is considered representative of the study catchment area (135 km2). The change in the soil water storage calculated for selected rainfall-flow events was used to assess the estimated soil moisture from the coupled model. Basically, the calculated Δs over the catchment considers the total depth of soil by describing the flow of water in and out of a column of soil. Therefore, it is concluded from this study that in order to have a proper representation between the calculated water storage and the estimated soil moisture from the MM5-NOAH LSM layers, it is important to account for the key soil depth of the soil column when calculating changes in the estimated soil moisture. Hence, it was found that combining the first two soil moisture layers of the LSM produces a better result in terms of the correlation between Δs and Δθ than either the first or second layer could produce. It is important to mention that the vegetation root zone, which is a function of the vegetation type, has an effective impact on the correlation result between Δs and Δθ over different depths of soil layers through its significant contribution into the ET calculation and consequently into the soil moisture calculations over the soil layer depth. Hence, if the vegetation root zone has a limited depth into the soil layer, a co-efficient of poor performance would be generated, while the inverse is also correct. The proposed scheme application offers a promising opportunity for further assessing soil moisture estimations from coupled LSMs and taking advantage of the abundance of hydrological data.

As a comparison study and according to the achieved correlation between Δs and Δθ used as a model performance indicator, it has been shown that the MM5-NOAH LSM outperforms the AMSR-E satellite when the proposed scheme is applied for soil moisture assessment. In fact, the MM5-NOAH LSM combined soil moisture with a depth of 40 cm was better than remote sensing techniques in retrieving ‘soil moisture’ although it suffers from a time drift problem which is a common problem occurred in hydrological models. However, accounting for soil layer depth is the key advantage in the performance of the MM5-NOAH LSM soil moisture estimations.

Overall, it is concluded from this study that the empirically proposed scheme offers a useful validation tool and could play a valuable role in the assessment and evaluation of soil moisture estimated from the MM5-NOAH LSM and AMSR-E satellite data for hydrological purposes as the presented results were positive and promising. However, there are still many knowledge gaps that need further research. For example, the strong synergy between satellite soil moisture sensing and numerical weather modelling may enable us to make full use of the available information and produce soil moisture with a further improved accuracy (i.e., a data fusion approach).

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

The authors would like to thank the Ministry of Higher Education and Scientific Research MOHER in Iraq for funding this research. We also thank the UK Met Office, the Environment Agency and the British Atmospheric Data Centre (BADC) for providing the hydrological and meteorological data. Also thanks go to the European Centre for Medium-Range Weather Forecasts ECMWF for providing the Era-interim reanalysis data. The two anonymous reviewers and editor have provided constructive comments which significantly improved this manuscript.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study area and data
  5. 3. Methodologies
  6. 4. Results and discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References
  • Albergel C, Calvet JC, De Rosnay P, Balsamo G, Wagner W, Hasenauer S, Naemi V, Martin E, Bazile E, Bouyssel F, Mahfouf JF. 2010. Cross-evaluation of modelled and remotely sensed surface soil moisture with in situ data in southwestern France. Hydrol. Earth Syst. Sci. 14: 21772191.
  • Albergel C, De Rosnay P, Gruhier C, Muñoz-Sabater J, Hasenauer S, Isaksen L, Kerr Y, Wagner W. 2012. Evaluation of remotely sensed and modelled soil moisture products using global ground-based in situ observations. Remote Sens. Environ. 118: 215226.
  • Allen RG, Pereira LS, Raes D, Smith M. 1998. Crop Evapotranspiration–Guidelines for Computing Crop Water Requirements. Natural Resources Management and Environment Department: Rome, Italy.
  • Al-Shrafany DM, Rico-Ramirez MA, Han D. 2012. Calibration of roughness parameters using rainfall runoff water balance for satellite soil moisture retrieval. ASCE J. Hydrol. Eng. 17(6): 704714. DOI:10.1061/(ASCE)HE.1943–5584.0000508.
  • Anthes RA, Warner TW. 1978. Development of hydrodynamic models suitable for air pollution and other mesometeorological studies. Mon. Weather Rev. 106: 10451078.
  • Betts AK, Ball JH, Beljaars ACM, Miller MJ, Viterbo P. 1996. The land-surface-atmosphere interaction: a review based on observational and global modelling perspectives. J. Geophys. Res. 101: 72097225.
  • Brocca L, Melone F, Moramarco T, Wagner W, Hasenauer S. 2010. ASCAT soil wetness index validation through in situ and modelled soil moisture data in central Italy. Remote Sens. Environ. 114(11): 27452755.
  • Calvet JC, Wigneron JP, Walker JP, Karbou F, Chanzy A, Albergel C. 2011. Sensitivity of passive microwave observations to soil moisture and vegetation water content: L-band to W band. IEEE Trans. Geosci. Remote Sens. 49(4): 11901199.
  • Chen F, Dudhia J. 2001. Coupling an advanced land surface-hydrology model with the Penn-State-NCAR MM5 modelling system: part I: model implementation and sensitivity. Mon. Weather Rev. 129: 569585.
  • Chen F, Mitchell K, Schaake J, Xue Y, Pan HL, Koren V, Duan Q, Betts A. 1996. Modelling land surface evaporation by four schemes and comparison with FIFE observations. J. Geophys. Res. 101: 72517268.
  • Chen Y, Yang K, Zhou D, Qin J, Guo X. 2010. Improving the Noah land surface model in arid regions with an appropriate parameterisation of the thermal roughness length. J. Hydrometeorol. 11: 9951006.
  • Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S, Andrae U, Balmaseda MA, Balsamo G, Bauer P, Bechtold P, Beljaars ACM, van de Berg L, Bidlot J, Bormann N, Delsol C, Dragani R, Fuentes M, Geer AJ, Haimberger L, Healy SB, Hersbach H, Holm EV, Isaksen L, Kallberg P, Kohler M, Matricardi M, McNally AP, Monge-Sanz BM, Morcrette J-J, Park B-K, Peubey C, de Rosnay P, Tavolato C, Thepaut J-N, Vitarta F. 2011. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137: 553597.
  • Dorigo WA, Wagner W, Hohensinn R, Hahn S, Paulik C, Xaver A, Gruber A, Drusch M, Mecklenburg S, van Oevelen P, Robock A, Jackson T Jackson T, 2011. The international soil moisture network: a data hosting facility for global in situ soil moisture measurements. Hydrol. Earth Syst. Sci. 15(5): 16751698.
  • Dudhia J, Gill D, Manning K, Wang W, Bruyere C. 2005. PSU/NCAR mesoscale modelling system: tutorial classnotes and users'guide. Online tutorial. http://www.mmm.ucar.edu/mm5/documents/tutorial-v3-notes.html (accessed on 5 December 2011).
  • Engman ET. 1991. Application of microwave remote sensing of soil moisture for water resources and agriculture. Remote Sens. Environ. 35: 213226.
  • Entekhabi D, Rodriguez-Iturbe I, Castelli F. 1996. Mutual of soil moisture state and atmospheric processes. J. Hydrol. 184: 317.
  • Gao H, Wood EF, Drusch M, Crow W, Jackson TJ. 2004. Using a microwave emission model to estimate soil moisture from ESTAR observation during SGP99. J. Hydrometeorol. 5: 4963.
  • Hipps LE, Swiatek E, Kustas WP. 1994. Interactions between regional fluxes and the atmospheric boundary layer over a heterogeneous watershed. Water Resour. Res. 30(5): 13871392.
  • Hong SY, Pan HL. 1996. Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Weather Rev. 124: 23222339.
  • Hossain F, Anagnostou EN. 2005. Numerical investigation of the impact of uncertainties in satellite rainfall estimation and land surface model parameters on simulation of soil moisture. Adv. Water Resour. 28: 13361350.
  • Jackson TJ, Schmugge TJ. 1989. Passive microwave remote sensing system for soil moisture: some supporting research. IEEE Trans. Geosci. Remote Sens. 27: 225235.
  • Jackson TJ, Schmugge TJ, Wang JR. 1982. Passive microwave sensing of soil moisture under vegetation canopies. Water Resour. Res. 18: 11371142.
  • Jacquemin B, Noilhan J. 1990. Sensitivity study and validation of a land surface parameterisation using the HAPEX-MOBILHY data set. Boundary Layer Meteorol. 52: 93134.
  • Kain JS, Fritsch JM, 1993. Convective parameterization for mesoscale models: the Kain-Fritsch scheme. In The Representation of Cumulus Convection in Numerical Models, Emanuel KA, Raymond DJ (eds). American Meteorology Society: Boston, MA.
  • Kong X, Dorling S, Smith R. 2011. Soil moisture modelling and validation at an agricultural site in Norfolk using the MetOffice surface exchange scheme (MOSES). Meteorol. Appl. 18: 1827.
  • Koster RD, Guo Z, Yang R, Dirmeyer PA, Mitchell K, Puma MJ. 2009. On the nature of soil moisture in land surface models. J. Clim. 22: 43224335.
  • Knowles KW, Savoie MH, Armstrong RL, Brodzik MJ. 2006. AMSR-E/Aqua Daily EASE-Grid Brightness Temperatures (2004–2005). National Snow and Ice Data Centre. Digital Media: Boulder, CO.
  • Lakshmi V, Wood E, Choudhury B. 1997. Evaluation of special sensor microwave/imager satellite data for regional soil moisture estimation over the Red River Basin. J. Appl. Meteorol. 36: 13091328.
  • Liguori S, Rico-Ramirez MA, Cluckie ID, 2009. Uncertainty propagation in hydrological forecasting using ensemble rainfall forecasts. In Hydroinformatics in Hydrology, Hydrogeology and Water Resources, Cluckie ID, Chen Y, Babovic V, Konikow L, Mynett A, Demuth S, Savic DA (eds), Vol. 331. International Association of Hydrological Sciences: Oxford, UK; 3040.
  • Liguori S, Rico-Ramirez MA, Schellart A, Saul A. 2011. Using probabilistic radar rainfall nowcasts and NWP forecasts for flow prediction in urban catchments. Atmos. Res. 103: 8095.
  • Liu Y, Chen F, Warner T, Basara J. 2005. Verification of a measoscale data-assimilation and forecasting system for the Oklahoma City area during the Joint Urban 2003 field project. J. Appl. Meteorol. Climatol. 45: 912929.
  • Mahfouf JF, Richard E, Mascart P. 1987. The influence of soil and vegetation on mesoscale circulations. J. Clim. Appl. Meteorol. 26: 14831495.
  • Mahrt L, Ek M. 1984. The influence of atmospheric stability on potential evaporation. J. Clim. Appl. Meteorol. 23: 222234.
  • Mahrt L, Pan L. 1984. A two-layer model of soil hydrology. BoundaryLayer Meteorol. 29: 120.
  • Meesters A, De Jeu RAM, Owe M. 2005. Analytical derivation of the vegetation optical depth from the microwave polarization difference index. IEEE Trans. Geosci. Remote Sens. Lett. 2: 121123.
  • Mellor D, Sheffield J, O'Connell PE, Metcalfe AV. 2000. Stochastic space-time rainfall forecasting system for real time flow forecasting II: application of SHETRAN and ARNO rainfall runoff models to the Brue catchment. Hydrol. Earth Syst. Sci. 4: 617626.
  • Miao JF, Chen D, Borne K. 2007. Evaluation and comparison of Noah and Pleim-Xiu land surface models in MM5 using GOTE2001 data: spatial and temporal variations in near surface air temperature. Journal of Applied Meteorology and Climatology 46: 15871605. [Correction added 13 June 2013 after original online publication: the reference Miao et al. 2007 has been corrected.]
  • Njoku E, Jackson TJ, Lakshmi V, Chan TK, Nghiem SV. 2003. Soil moisture retrieval from AMSR-E. IEEE Trans. Geosci. Remote Sens. 41: 215229.
  • Owe M, De Jeu RAM, Walker J, Branch HS, Greenbelt MD. 2001. A methodology for surface soil moisture and vegetation optical depth retrieval using the microwave polarization difference index. IEEE Trans. Geosci. Remote Sens. 39: 16431654.
  • Owe M, De Jeu RAM, Holmes T. 2008. Multi-sensor historical climatology of satellite-derived global land surface moisture. J. Geophys. Res. 113(F1).
  • Pan HL, Mahrt L. 1987. Interaction between soil hydrology and boundary-layer development. Boundary Layer Meteorol. 38: 185202.
  • Patil MN, Waghmare RT, Halder S, Dharmaraj T. 2011. Performance of Noah land surface model over the tropical semi-arid conditions in western India. Atmos. Res. 99: 8596.
  • Reen BP, Stauffer DR, Davis KJ, Desai AR. 2006. A case study on the effects of heterogeneous soil moisture on mesoscale boundary layer structure in the Southern Great Plains, U.S.A. Part II: mesoscale modelling. Boundary Layer Meteorol. 120: 275314.
  • Sahoo AK, Dirmeyer PA, Houser PR, Kafatos M. 2008. A study of land processes using land surface models over the Little River Experimental Watershed, Georgia. J. Geophys. Res. 113: D20121.
  • Schmugge TJ. 1983. Remote sensing of soil moisture: recent advances. IEEE Trans. Geosci. Remote Sens. GE21: 145146.
  • Schmugge TJ, Kustas WP, Ritchie JC, Jackson TJ, Rango A. 2002. Remote sensing in hydrology. Adv. Water Resour. 25: 13671385.
  • Schaake JC, Koren VI, Duan QY, Mitchell K, Chen F. 1996. A simple water balance model (SWB) for estimating runoff at different spatial and temporal scales. J. Geophys. Res. 101: 74617475.
  • Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Wang W, Powers JG. 2005. A description of the Advanced Research WRF version 2. NCAR Technical Note.
  • Sridhar V, Elliott RL, Chen F, Brotzge JA. 2002. Validation of the NOAH-OSU land surface model using surface flux measurements in Oklahoma. J. Geophys. Res. 107: D20.
  • Stauffer DR, Seaman NL. 1990. Use of four-dimensional data assimilation in a limited-area mesoscale model. Part I: experiments with synoptic-scale data. Mon. Weather Rev. 118: 12501277.
  • Van der Velde RV, Su Z, Ek M, Rodell M, Ma Y. 2009. Influence of thermodynamic soil and vegetation parameterisations on the simulation of soil temperature states and surface fluxes by the Noah LSM over a Tibetan plateau site. Hydrol. Earth Syst. Sci. 13: 759777.
  • Venalainen A, Salo T, Fortelius C. 2005. The use of numerical weather forecast model predictions as a source of data for irrigation modelling. Meteorol. Appl. 12: 307318.
  • Wang JR. 1983. Passive microwave sensing of soil moisture content: the effects of soil bulk density and surface roughness. Remote Sens. Environ. 13: 329344.
  • Wang JR, Choudhury BJ. 1980. Remote sensing of moisture content over bare field at 1.4 GHz frequency. J. Geophys. Res. 86: 52775282.
  • Wang L, Wen J, Zhang T, Zhao Y, Tian H, Shi X, Wang X, Liu R, Zhang J, Lu S. 2009. Surface soil moisture estimation from AMSR-E observations over an arid area, Northwest China. Hyd. Earth Sys. Sci. Disc. 6: 10551087.
  • Wigneron JP, Calvet JC, Pellarin T, Van de Griend AA, Berger M, Ferrazzoli P. 2003. Retrieving near-surface soil moisture from microwave radiometric observations: current status and future plans. Remote Sens. Environ. 85: 489506.