A First Assessment of the SMOS Soil Moisture ... - Semantic Scholar
1612
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
A First Assessment of the SMOS Soil Moisture Product With In Situ and Modeled Data in Italy and Luxembourg Teodosio Lacava, Member, IEEE, Patrick Matgen, Luca Brocca, Marco Bittelli, Nicola Pergola, Member, IEEE, Tommaso Moramarco, and Valerio Tramutoli
Abstract—The European Space Agency Soil Moisture and Ocean Salinity (SMOS) mission was launched on November 2, 2009. Providing accurate soil moisture (SM) estimation is one of its main scientific objectives. Since the end of the commissioning phase, preliminary global SMOS SM data [Level 2 (L2) product] are distributed to users. In this paper, we carried out a first assessment of the reliability of this product through a comparison with in situ observed and modeled SM over three different sites: One is located in Luxemburg, and two are located in Italy. The period from August 1, 2010, to July 1, 2011, has been analyzed, giving us the opportunity to evaluate the satellite response to different SM states. The selected period is important for hydrological predictions as it is typically characterized by a sequence of transitions from dry to wet and from wet to dry conditions. In order to compare SMOS and ground SM measurements, a two-step approach has been applied. First, an exponential filter has been applied to approximate root-zone SM, and second, a cumulative distribution function matching has been employed to remove systematic differences between satellite and in situ observations and model simulations of SM. Our results indicate rather good reliability of the filtered and bias-corrected SM estimates derived from the first SMOS L2 products. Bearing in mind that an updated/advanced version of the SMOS SM product has been recently produced, our preliminary results already seem to confirm the potential of SMOS for monitoring of water in soils. Index Terms—Remote sensing, soil moisture (SM), Soil Moisture and Ocean Salinity (SMOS), validation.
I. I NTRODUCTION
T
HERE is ample evidence that hydrological processes are significantly conditioned by a river basin’s wetness state, e.g., [1] and [2]. Many field studies show that the distribution Manuscript received March 31, 2011; revised September 28, 2011, December 17, 2011, and January 18, 2012; accepted January 22, 2012. Date of publication March 12, 2012; date of current version April 18, 2012. T. Lacava and N. Pergola are with the Institute of Methodologies for Environmental Analysis (IMAA), National Research Council (CNR), 85050 Potenza, Italy (e-mail: [email protected]; [email protected]). P. Matgen is with the Centre de Recherche Public–Gabriel Lippmann, 4422 Belvaux, Luxembourg (e-mail: [email protected]). L. Brocca and T. Moramarco are with the Research Institute for GeoHydrological Protection (IRPI), National Research Council (CNR), 06128 Perugia, Italy (e-mail: [email protected]; [email protected]). M. Bittelli is with the Department of Agro-Environmental Science and Technology, University of Bologna, 40126 Bologna, Italy (e-mail: marco. [email protected]). V. Tramutoli is with the Department of Engineering and Physics of the Environment (DIFA), University of Basilicata, 85100 Potenza, Italy (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TGRS.2012.2186819
of a catchment’s water is controlled by soil water storage, with runoff yield abruptly rising when a certain storage threshold is exceeded, e.g., [3]–[5]. Penna et al. [6] argue that the soil water stored in hillslope areas is released only during wetter conditions, when flow paths between the hillslope and riparian zone become connected. Therefore, monitoring the closeness to thresholds is essential to accurately predict stream responses to rainfall events. One approach to do this is through the measurement of surface soil moisture (SM) (SSM), the latter being considered as a proxy of soil water storage that enables the periodical assessment of the readiness of a river basin to generate a fast discharge response. Apart from in situ measurements and hydrological models, SM can be estimated with sensor data obtained from satellite platforms. In the last years, the capability of Earth observation systems to provide reliable SM measurements has been widely investigated. In particular, the data acquired by microwave sensors, both active and passive, have been used, providing detailed information about SM variability in the space–time domain [7]–[9]. The launch of the Soil Moisture and Ocean Salinity (SMOS) mission, the European Space Agency (ESA) dedicated SM mission, in November 2009 [10]–[13], clearly indicates the need and the will of the international scientific community of having a better SM estimation from space. The main appeal of the Microwave Imaging Radiometer with Aperture Synthesis (MIRAS) on board SMOS over existing sensors is that the emitted microwave radiation is measured at L-band (1.4 GHz) at a reasonable spatial resolution (the average ground resolution is 43 km) whereas previous spaceborne radiometers operated at frequencies of 6.6 GHz (C-band) and higher. At L-band and depending on soil characteristics and wetness conditions, 5 cm of soil is probed on average [14]. As the thickness of the sampled layer by the radiometers increases with increasing wavelength and due to the fact that the vegetation optical depth and other perturbations less impact lower frequencies, it is expected that L-band radiometers such as MIRAS show the highest potential for SM retrieval studies [7], [15]. Moreover, the vegetation and soil contributions to the signal can be separated more easily as MIRAS images any point of the surface at several angles [11], [12]. One of the several possible retrieval models for inverting SM from dualpolarization multiangular brightness temperature observations of SMOS is the L-band Microwave Emission of the Biosphere (L-MEB) model [16]–[18]. The validity of the model has been verified in a number of tower-based case studies over mostly
0196-2892/$31.00 © 2012 IEEE
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
1613
TABLE I M AIN C HARACTERISTICS OF THE T EST S ITES U SED FOR T HIS S TUDY (LAT: L ATITUDE , LON: L ONGITUDE , E LEV: E LEVATION , MAR: M EAN A NNUAL R AINFALL , AND MAT: M EAN A NNUAL T EMPERATURE )
homogeneous land surface conditions [15], [19]. In a similar experiment, Panciera et al. [20] applied the L-MEB model with the proposed set of model parameters over grassland and cropland cover types in Australia. They showed that the targeted accuracy level of 0.04-m3 /m3 volumetric SM could only be reached over grassland-covered sites while significant underestimations of SSM were found over crop-covered areas. Initial results with actual SMOS data seem to confirm the good sensitivity of MIRAS to SM variations [21]–[24]. A “worrying problem” [25] for passive microwave remote sensing, particularly radiometry with wavelengths longer than C-band, is the radio frequency interference (RFI). As a result, brightness temperature may be degraded over densely populated areas at the risk of rendering SM retrievals impossible [26]. Such a problem becomes even more relevant when it happens in what should be a protected band reserved for the Earth Exploration Satellite Service, space research, and radio astronomy [27]. The first results show that L-band SMOS data, in some places, are also contaminated by RFI [28]–[32]. Kerr [12] qualifies the pixel heterogeneity and the significant differences in behavior from different targets that go with it as the most stringent limitation currently impacting the accuracy of SMOS-derived SM estimates. Other possible limitations that are currently not well mastered are related to topography, frozen soils, snow cover, urban areas, permanent water bodies, bare dry soil, litter, and forest emission. It is well known that an important issue when validating coarse-resolution satellite SM products with ground measurements relates to the disparity of spatial scales between the two data sets [33], [34]. However, several studies have shown that point measurements of SM can be representative of larger areas (e.g., [35]–[37]), thereby reducing the commensurability problem. However, systematic differences between remote-sensingderived and in situ observations are usually detected even though the temporal dynamics are very similar, e.g., [38]. To address this problem, several techniques have been developed for matching the variability of satellite data with in situ ones, e.g., [8], [9], [33], [39], and [40]. In this context, the objective of this paper is to carry out a field evaluation of the SMOS Level 2 (L2) SM product with ground measurements carried out in three experimental research basins: One is located in Luxembourg, and two are located in Italy. Such an approach gives the opportunity to assess SMOS SM data sensitivity under different observation conditions (i.e., latitude, topography, snow presence, land cover uses, vegetation presence, etc.). In particular, the study covers the period August 1, 2010–July 1, 2011, thereby allowing us to evaluate the capability of the sensor to capture the transitions
from dry to wet as well as from wet to dry conditions and to verify possible limitations related to the presence of snow. The SMOS SM product is compared with both in situ and modeled SM data to provide a robust assessment of its reliability. II. M ETHODS Due to the difference in sensing depth between satellite (e.g., ∼5 cm for MIRAS) and in situ sensors (5–10 cm for this study; see Table I), the semiempirical approach proposed by Wagner et al. [41], also named “exponential filter,” is used to obtain the root-zone SM (RZSM) product from the SSM directly sensed by satellite sensors. In addition, in order to match satellite time series (both SSM and RZSM) to in situ ones, the cumulative distribution function (CDF) matching approach (e.g., [39]) is employed. This two-step approach is briefly described in the following along with the soil water balance model that has been used for the computation of the modeled SM data. A. Exponential Filter The profile SM values are often reasonably well correlated with SSM values (e.g., [42]) because both are an expression of meteorological conditions during the preceding days and weeks. The exponential filter, introduced by Wagner et al. [41], is a simple and effective method to retrieve profile SM values from surface ones. It relies on the analytical solution of a differential equation, assuming that the variation in time of the average profile SM is linearly related to the difference between the surface and the profile values. The method is dependent on only one parameter, the characteristic time length T , that characterizes the temporal variation of SM within the root-zone profile. Despite its simplicity, the algorithm was found reliable in predicting profile SM values based on SSM information using both in situ observed [43]–[45] and modeled [46] data. Because of the higher sensing depths of L-band radiometers, it can be expected that there is a reduced need for implementing a transformation routine. However, it is also worth mentioning that a low-pass filter also tends to reduce random noise affecting the investigated raw time series. For a detailed description of the method, the reader may refer to [41] and [47]. In this study, the recursive formulation of the method is used [48]. B. CDF Matching Approach The systematic differences between remote-sensing-derived and in situ observations of SM prevent the absolute agreement between the two time series. Consequently, the comparison
1614
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
of these two time series is often aided by normalizing the remotely sensed data to better match the distribution of ground measurements. In this study, the CDF matching approach (e.g., [39]) that can be considered as an enhanced nonlinear technique for removing systematic differences between two data sets is implemented. Through this method, the satellite data are rescaled in such a way that its CDF matches the CDF of in situ measurements. The CDF matching is performed on the SSM (and RZSM) time series, yielding SSM-CDF (and RZSM-CDF).
C. Soil Water Balance Model In addition to in situ observations, modeled SM data are used for the validation of the SMOS SM products. The structure of the soil water balance model used in this study was derived by using SM observations carried out in an experimental catchment located in central Italy [2]. The model has been applied in different test sites located across Europe with good performances obtained from all sites [9], [40]. The model, which requires routinely measured meteorological variables (rainfall and air temperature) as input data, incorporates only five parameters, and the default time step is 1 h. Moreover, the model was found reliable even when it was calibrated only with a limited number of observations [2], thus allowing to confidently apply the model over large areas and for periods different from those employed for parameter calibration. It is also worth mentioning that the water balance model is forced with rainfall and temperature data averaged over an area of 15 km by 15 km surrounding each test site. As the spatial resolution of the hydrological model is similar to the footprint of the SMOS satellite, the modeled SM data are representative of an area larger than the one covered by in situ measurements, which arguably reduces the problem of the spatial mismatch between ground and satellite data.
III. S TUDY A REAS AND SM M EASUREMENTS The three test sites used for the validation of the SMOS SM products are located in Italy (two sites) and Luxembourg (one site). These sites have been used in previous studies [11], [49] for the validation of coarse-resolution satellite SM products retrieved by active and passive microwave sensors. The calibration of the sensors against gravimetric SM measurements was carried out for each site, obtaining a good accuracy (less than 0.02 m3 /m3 ). Therefore, the in situ SM data can be assumed to be sufficiently accurate and representative of the SM variability in the investigated areas. Table I synthesizes the main characteristics of each site in terms of location, soil texture, vegetation cover, and measurement depth, while Fig. 1 shows the location of the study sites with respect to the SMOS satellite footprint. For all the investigated sites, in situ SM observations are available for the period from August 2010 to July 2011 with a temporal resolution of 1 h. The SMOS SM evaluation is carried out with both in situ observations and modeled data as reference data. In the following, the main characteristics of the three test sites are described. Note that, in all test sites, rainfall and temperature data are collected as forcing data for driving the soil water balance model. A. Italy—CER The Cerbara (CER) site is located in central Italy, and it was installed by the Hydrology Service of Umbria region for civil protection activities. The area is characterized by a Mediterranean semihumid climate with an average precipitation of 900 mm/year and a mean annual temperature of 13 ◦ C. The SM probe, based on time-domain reflectometry (TDR) technique (ThetaProbe ML2X, Delta-T Device), was set up in November 2009 and continuously measures volumetric SM in the soil column at 10-, 20-, and 40-cm depths; the data at 10-cm depth are considered for this study.
D. Performance Index
B. Italy—SPC
For each comparison, the two following statistical indexes are used to evaluate the SMOS SM product accuracy: (1) RM SD = (SMSAT − SMINSITU )2 (SMSAT − SMINSITU )2 (2) R= 1− (SMINSITU − SMINSITU )2
The San Pietro Capofiume (SPC) site is located in northern Italy (Emilia–Romagna region), and it was installed by the Service of Hydrology, Meteorology and Climate of the Regional Agency for Environmental Protection in Emilia–Romagna (ARPA-SIMC), mainly for agricultural purposes. The area is characterized by a Mediterranean climate with an average precipitation of 750 mm/year and a mean annual temperature of 16 ◦ C. For this site, TDR volumetric SM measurements (TDR100, Campbell Scientific Inc.) have been collected at seven depths at an hourly time step; the data at 10-cm depth are used for this study.
BIAS = (SMSAT − SMINSITU )
(3)
where RM SD is the root mean squared difference between the in situ (observed and modeled) SMINSITU and the different satellite SM products SMSAT , the overbar is the mean operator, R is the correlation coefficient, and BIAS is the mean difference between in situ and satellite products. It has to be noted that, instead of RM SE (where E stands for “error”), we used RM SD to underline that also ground measurements contain errors (instrumental and representativeness) and, hence, they cannot be considered as the “true” SM.
C. Luxembourg In the southern part of Luxembourg, the experimental Bibeschbach (BIB) catchment (10.8 km2 ) is considered as the third test site. It is characterized by a gently rolling landscape and a typical humid temperate climate with an average precipitation of about 740 mm/year and a mean annual temperature
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
Fig. 1.
1615
Location and land cover of the three test sites in Luxembourg and Italy with the boundaries of the SMOS pixels for those areas.
of 9 ◦ C. The site is mainly covered by forest and agricultural land on loamy soils. Since 2005, the basin has been equipped with a set of 40 sensors (ECH2O, DecagonTM) measuring the volumetric SM of the topsoil layer at a depth of 4–7 cm using the capacitance principle. The measurement sites are located on the edge of grasslands or agricultural plots and are distributed throughout the catchment. The SM sensors are connected to data loggers that record an hourly dielectric constant of the medium. A calibration with regular TDR point measurements is performed to convert the signal into volumetric SM content. The calibration equation is established for each measurement site individually. All measurements exhibit a similar temporal pattern of SM fluctuations. For this study, the area-averaged volumetric SM of the near-surface SM probes (∼5 cm depth) is considered. IV. SMOS S ATELLITE P RODUCTS The SMOS Soil Moisture User Data Product 2 (SMUDP2), provided by ESA in the framework of a Cat-1 project, has been used in this study. In particular, such data, delivered by ESA [50], have been generated by the SMOS L2 SM processor, taking Level 1c (L1c) and other auxiliary data (i.e., brightness temperature, polarization, radiometric uncertainty, incidence
angle, azimuth angle, elliptical footprint semi-axes) as input data. Using recalibrated L1c data, an updated version of the L2 data has been produced recently by the Centre d’Etudes Spatiales de la Biosphère [51]. SMUDP2 data have a spatial resolution of 43 km on average and a geolocation error of 400 m; the data are mapped on an icosahedral Snyder equal area hexagonal grid of aperture 4 and resolution 9 (ISEA 4H9) with a spatial resolution of 15 km (see also Fig. 1). In the observation period (i.e., August 1, 2010–July 1, 2011), SMUDP2 has been generated regularly, except in the period December 27, 2010–January 1, 2011, probably due to some SMOS anomalies [52]. V. R ESULTS First, the modeled SM data are computed for each test site by optimizing the soil water balance model parameters against in situ observations. Fig. 2 shows the comparison between modeled and observed data for the three sites, highlighting the good model performance. In particular, the correlation coefficient R values between the observed and modeled data equal 0.97, 0.93, and 0.84 for the CER, SPC, and BIB sites, respectively (for BIB, no modeled data were achievable for August 2010); the root mean squared difference RM SD is always lower than 0.025 m3 /m3 . The agreement between modeled and observed
1616
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
Fig. 2. Time series of in situ observed and modeled SM for (a) CER, (b) SPC, and (c) BIB sites. The hourly rainfall pattern is also shown.
data indicates that the ground observations are not much affected by measurement errors or any other type of error, thus representing an appropriate benchmark for SMOS data except for their spatial representativeness. It is worth noting that, because of the calibration of model parameters with ground measurements, the time series of in situ and modeled SM cannot be considered as two independent validation data sets. However, the integration of in situ measurements and model results enables a more comprehensive and efficient assessment of the reliability of the SMOS SM product as it reduces the impact of ground data uncertainties related to measurement errors in the case of in situ data as well as forcing data and model parameters in the case of modeled data. For the comparison between ground and satellite data, the following procedure is adopted. For each test site, the satellite pixel whose centroid is the nearest to the in situ location is identified and considered for the subsequent analysis. Then, the in situ and modeled SM dates corresponding to the acquisition time of the SMOS passes are retrieved, and the corresponding data values are used for the SMOS evaluation. Next, the satellite RZSM product is computed starting from the SSM time series by optimizing the value of the characteristic time length parameter T obtained by maximizing the correlation between the RZSM and the in situ time series. Finally, the CDF matching approach is applied to satellite data (SSM and RZSM) to match
their CDF to the CDF of the in situ data, thereby yielding biascorrected SSM-CDF and RZSM-CDF time series. The results of all the comparisons between SMOS SM products (SSM and RZSM) and in situ data in terms of RM SD, BIAS, and R values are reported in Table II both with respect to the original values and the rescaled data using the CDF matching approach. The sample size and the optimized value of the characteristic time length parameter T , used for RZSM computation, are also indicated. Therefore, the field evaluation is carried out separately with three soil wetness products (i.e., with and without the application of the exponential filter and the bias removal). As all intermediate results are provided in the manuscript, the utility and the effect of each processing step can be investigated separately. Fig. 3 shows the time series of the three SMOS SM products (SSM, SSM-CDF, and RZSMCDF) against in situ observations for the three sites while Fig. 4 shows the corresponding scatter plots. In Figs. 3 and 4, it can be seen that the SSM-CDF and RZSM-CDF SM products are able to reproduce the temporal pattern of ground observations with sufficient accuracy, even at a fine temporal scale. On the other hand, when the original SSM product is considered, a tendency to underestimate the ground observations of SM is apparent for all the sites (see also the BIAS values reported in Table II). This observation is well in line with other studies dealing with SMOS SM validation [53], [54].
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
1617
TABLE II S UMMARY OF THE R ESULTS OF THE C OMPARISON B ETWEEN IN SITU AND SMOS SM P RODUCTS (SSM: S URFACE SM, RZSM: ROOT-Z ONE SM, R: C ORRELATION C OEFFICIENT, RMSD: ROOT M EAN S QUARE D IFFERENCE , CDF: C UMULATIVE D ISTRIBUTION F UNCTION M ATCHING , T: C HARACTERISTIC T IME L ENGTH , AND N: S AMPLE S IZE ). T HE RMSD AND THE BIAS A RE E XPRESSED IN m3 /m−3
Fig. 3. Time series of in situ SM, OBS, and SMOS SM products for (a) CER, (b) SPC, and (c) BIB sites (SSM: Surface SM, RZSM: Root-zone SM, and CDF: Cumulative distribution function matching approach).
For the investigated 11-month period, while 209 SMOSderived SM estimates are available for the CER site, the sample size is close to 300 for the other two study areas (equal to 309 for the SPC site and 292 for the BIB site). For the BIB site, no data are available for the month of December due to
a continuous presence of snow, which did not allow any SM retrieval from SMOS data. The difference in the number of acquisitions can be explained also by the occasional occurrence of RFI that has hampered SM retrieval, mainly for the CER site.
1618
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
Fig. 4. Comparison of in situ SM, OBS, and SMOS SM products for (a) CER, (b) SPC, and (c) BIB sites (SSM: Surface SM, RZSM: Root-zone SM, CDF: Cumulative distribution function matching approach).
Despite the existing scale mismatch between the satellite footprint and the area covered by ground measurements, the results in terms of correlation are rather satisfactory with R-values ranging between 0.314 and 0.637 (on average, equal to ∼0.518) for the original SSM and the SSM-CDF products. On the other hand, the absolute SM values are characterized by high RM SD values that are equal, on average, to 0.108 m3 /m3 , mainly due to the large underestimation that is visible in Fig. 3. After removing systematic differences through the CDF matching approach, RM SD values are significantly decreased with an average value that is now equal to 0.054 m3 /m3 . On the other hand, it can be seen that the agreement is higher when modeled data are used rather than in situ measurements (both in terms of R and RM SD), arguably because the spatial scale covered by the model is similar to the SMOS pixel size. If the exponential filter is applied, thus obtaining the RZSM product, R values increase significantly (0.617–0.810), but the RM SD is still characterized by values (0.059–0.123 m3 /m3 ) that are fairly similar to those that were found for the unprocessed SSM product. On the other hand, the filtered and rescaled SMOS SM product, i.e., RZSM-CDF, provides the best results with R values ranging between 0.591 and 0.877 and RM SD values between 0.024 and 0.046 m3 /m3 . With respect to the obtained T values, for the comparison with in situ observations, T is equal to 15 days for all the sites while lower values were obtained with modeled data (on average, equal to nine days). The obtained T values are higher than the ones obtained by Kerr et al. [11] and Matgen et al. [49] who compared different SM products, obtained by the Advanced Scatterometer and Advanced Microwave Scanning Radiometer sensors, with in situ observations over the same test sites. This unexpected result supports the assumption that the main effect of applying an exponential filter is to reduce random errors affecting the time series. This result is arguably due to the high temporal variability of the first delivery of the SMOS SM product (see Fig. 3). As a consequence, the exponential filter tends to smoothen the temporal fluctuations, and by reducing the noise level in the raw data provides a better agreement with ground observations. This would explain why higher values for T were obtained than in previous studies dealing with sensors characterized by lower sensing depths. The result seems to indicate that the exponential filter leads to a removal of noise in the SMOS data set rather
Fig. 5. (a) Correlation coefficient R and (b) root mean squared difference RM SD between in situ observed and modeled and SMOS SM data for all the investigated sites and by separating (asc) ascending and (desc) descending passes (SSM: Surface SM, RZSM: Root-zone SM, and CDF: Cumulative distribution function matching approach).
than to a conversion of SSM into estimates of SM in deeper layers. To further investigate the SMOS reliability, the ascending and the descending passes are identified, and the same comparisons are carried out separately for the two orbits. Fig. 5 summarizes and compares the results in terms of R and RM SD values for all the sites. The selection of data corresponding to the ascending passes between 03:00 and 05:00 Greenwich Mean Time (GMT) slightly improves the results with, for SSM, an average increase of correlation that is equal to 0.054 and a reduction of 0.003 m3 /m3 of the RM SD values. Only considering the descending passes, between 17:00 and 19:00 GMT, clearly reduces the performance. This behavior can be ascribed to several factors: 1) ascending data, particularly for the period investigated, are more stable than descending ones in terms of temperature variations [55], and 2) residual RFI effects, if
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
present, are less important for the ascending passes for which human activity is comparatively low. However, a more in-depth investigation is clearly needed to corroborate these findings. One motivation for the SMOS mission is to improve the effectiveness of flood forecasting systems [12]. Our study provides some promising first results in this respect. The transition from dry to wet and saturated conditions, which is well visible in the in situ and modeled SM time series (Fig. 2) between October 1, 2010, and December 1, 2010, is quite well captured by the RZSM-CDF signal (see Fig. 3). Similarly, in early spring, between March 1, 2011, and May 1, 2011, the RZSMCDF signal over the BIB and SPC sites closely follows the gradual transition from nearly saturated to very dry conditions, while there is a slightly delayed response in the SMOS data for the CER site. This means that, in threshold-mediated catchments where fast storm flow is initiated once a certain SM threshold is approached, there is clearly a potential for filtered and bias-corrected SMOS-derived SM data to serve as a proxy of soil water storage that may help to verify and eventually improve hydrological model predictions. However, it is also obvious from our results that the detection of SM threshold changes is difficult to achieve with the unprocessed SMOS SSM data. Moreover, it is important to note that, in all test sites, the dynamic range of SM fluctuations is rather high, with SM varying by more than 0.25 m3 /m3 between the beginning and end of the investigation period. While our results clearly illustrate the capability of SMOS data for monitoring season-dependent SM variations, one could argue that, even with targeted 0.04 m3 /m3 accuracy, it still remains difficult to effectively detect event-specific SM fluctuations (see, for example, the sudden SM fluctuation observed in June 2011 in both CER and SPC sites that is not captured by the SMOS signal). However, the distinction between periods potentially exhibiting high runoff yields and periods with limited runoff production appears to be feasible with the accuracy levels expected from the SMOS L2 SM data, which is confirmed by the results demonstrated in this paper.
1619
of the SMOS L2 data is significantly larger than the ranges associated with in situ measured and modeled data. Second, the application of the exponential filter and the CDF matching operator yields RZSM estimates that are highly correlated with in situ measured and modeled SM (R between 0.591 and 0.877), indicating a satisfactory capability of SMOS for monitoring SSM variability. The RM SD values between the postprocessed SMOS L2 SM data and the in situ measurements are 0.042 m3 /m3 for the CER site, 0.037 m3 /m3 for the SPC site, and 0.029 m3 /m3 for the BIB site. Moreover, the postprocessed SMOS L2 SM captures the transition from dry to wet as well as from wet to dry conditions that is characteristic of the passage from autumn to winter and from winter to spring, respectively. This result highlights an interesting potential of SMOS SM data for improved hydrological predictions. In fact, by reliably distinguishing between the two main hydrological states, SMOS may help to monitor a hydrological system’s closeness to critical thresholds. Due to the limited sample size and the fact that this is the first delivery of the SMOS SM product, the results of this study need to be considered as preliminary and, clearly, need to be both confirmed and deepened in future investigations. ACKNOWLEDGMENT
VI. C ONCLUSION
The authors would like to thank the Umbria Region (Italy), the Regional Agency for Environmental Protection in Emilia–Romagna (Italy), and the Centre de Recherche Public–Gabriel Lippmann (Luxembourg) for providing the in situ soil moisture (SM) and hydrometeorological data. This research was carried out in the framework of the Category 1 Project 4699: “Assessment of EO-based soil wetness variation maps for geohazards monitoring and mitigation” funded by the European Space Agency (ESA) in the framework of First SMOS Data Announcement of Opportunity. Additional data over Luxembourg were acquired in the framework of the ESA Category 1 Project 8696: “On the value of SMOS MIRASderived volumetric SM for hydrological monitoring and prediction applications: A field evaluation over Luxembourg.”
This paper has investigated the potential of SMOS for capturing the SM variability in three study areas characterized by different hydrometeorological regimes. We are aware that, because of the differences in scale and coverage between the point measurements and the SMOS observations, a justified evaluation with in situ SM measurements remains difficult. However, and in spite of this study’s focus on specific areas (and limited time periods), we feel that there are elements of this investigation that are of more general significance. Also, it is important to take into account that the data set that we used refers to the very first SMOS SM product release. First, the combination of SMOS MIRAS brightness temperature observations with the SMOS L2 SM processor provides SSM estimates that, on all sites, are correlated reasonably well (R between 0.345 and 0.637) with in situ and modeled SM. More specifically, on all sites, the SMOS L2 SM data have a mean value that is significantly lower than the mean of the in situ measurements. Moreover, the dynamical range
[1] A. W. Western and R. B. Grayson, “The Tarrawarra data set: Soil moisture patterns, soil characteristics and hydrological flux measurements,” Water Resour. Res., vol. 34, no. 10, pp. 2765–2768, 1998. [2] L. Brocca, F. Melone, and T. Moramarco, “On the estimation of antecedent wetness condition in rainfall-runoff modelling,” Hydrol. Process., vol. 22, no. 5, pp. 629–642, Feb. 2008. [3] H. J. Tromp-van Meerveld and J. J. McDonnell, “Threshold relations in subsurface stormflow 1. A 147 storm analysis of the Panola hillslope,” Water Resour. Res., vol. 42, p. W02 410, Feb. 2006. doi:10.1029/ 2004WR003778. [4] J. Latron and F. Gallart, “Runoff generation processes in a small Mediterranean research catchment (Vallcebre, Eastern Pyrenees),” J. Hydrol., vol. 358, no. 3/4, pp. 206–220, Sep. 2008. [5] C. Spence, “A paradigm shift in hydrology: Storage thresholds across scales influence catchment runoff generation,” Geography Compass, vol. 4, no. 7, pp. 819–833, Jul. 2010. [6] D. Penna, M. Borga, D. Norbiato, and G. Dalla Fontana, “Hillslope scale soil moisture variability in a steep alpine terrain,” J. Hydrol., vol. 364, no. 3/4, pp. 311–327, Jan. 2009. [7] J. C. Calvet, J. P. Wigneron, J. Walker, F. Karbou, A. Chanzy, and C. Albergel, “Sensitivity of passive microwave observations to soil
R EFERENCES
1620
[8]
[9]
[10]
[11]
[12] [13] [14]
[15]
[16]
[17]
[18]
[19] [20]
[21]
[22]
[23]
[24]
[25]
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
moisture and vegetation water content: L-band to W-band,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 4, pp. 1190–1199, Apr. 2011. T. Lacava, L. Brocca, G. Calice, F. Melone, T. Moramarco, N. Pergola, and V. Tramutoli, “Soil moisture variations monitoring by AMSU-based soil wetness indices: A long-term inter-comparison with ground measurements,” Remote Sens. Environ., vol. 114, no. 10, pp. 2317–2325, Oct. 2010. L. Brocca, S. Hasenauer, T. Lacava, F. Melone, T. Moramarco, W. Wagner, W. Dorigo, P. Matgen, J. Martínez-Fernández, P. Llorens, J. Latron, C. Martin, and M. Bittelli, “Soil moisture estimation through ASCAT and AMSR-E sensors: An intercomparison and validation study across Europe,” Remote Sens. Environ., vol. 115, no. 12, pp. 3390–3408, Dec. 2011. doi:10.1016/j.rse.2011.08.003. Y. Kerr, P. Waldteufel, J. P. Wigneron, J. M. Martinuzzi, J. Font, and M. Berger, “Soil moisture retrieval from space: The Soil Moisture and Ocean Salinity (SMOS) mission,” IEEE Trans. Geosci. Remote Sens., vol. 39, no. 8, pp. 1729–1735, Aug. 2001. Y. Kerr, P. Waldteufel, J. P. Wigneron, S. Delwart, F. Cabot, J. Boutin, M. J. Escorihuela, J. Font, N. Reul, C. Gruhier, S. E. Juglea, M. R. Drinkwater, A. Hahne, M. Martín-Neira, and S. Mecklenburg, “The SMOS mission: New tool for monitoring key,” Proc. IEEE, vol. 98, no. 5, pp. 666–687, May 2010. Y. Kerr, “Soil moisture from space: Where are we?,” Hydrogeol. J., vol. 15, no. 1, pp. 117–120, Feb. 2007. H. Barré, B. Duesmann, and Y. Kerr, “SMOS: The mission and the system,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 587–593, Mar. 2008. S. Delwart, C. Bouzinac, P. Wursteisen, M. Berger, M. Drinkwater, M. Martín-Neira, and Y. H. Kerr, “SMOS validation and the COSMOS campaigns,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 695– 704, Mar. 2008. K. Saleh, J. P. Wigneron, P. Waldteufel, P. de Rosnay, M. Schwank, J. C. Calvet, and Y. Kerr, “Estimates of surface soil moisture under grass covers using L-band radiometry,” Remote Sens. Environ., vol. 109, no. 1, pp. 42–53, Jul. 2007. J. P. Wigneron, Y. Kerr, P. Waldteufel, K. Saleh, M. J. Escorihuela, P. Richaume, P. Ferrazzoli, M.-J. Escorihuela, J. P. Grant, B. Hornbuckle, P. deRosnay, J.-C. Calvet, T. Pellarin, R. Gurney, and C. Mätzler, “L-band Microwave Emission of the Biosphere (L-MEB) model: Description and calibration against experimental data sets over crop fields,” Remote Sens. Environ., vol. 107, no. 4, pp. 639–655, 2007. J. Wigneron, A. Chanzy, Y. H. Kerr, H. Lawrence, S. Jiancheng, M. J. Escorihuela, V. Mironov, A. Mialon, F. Demontoux, P. de Rosnay, and K. Saleh-Contell, “Evaluating an improved parameterization of the soil emission in L-MEB,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 4, pp. 1177–1189, Apr. 2011. Array Systems Computing Inc., Algorithm Theoretical Basis Document (ATBD) for the SMOS Level 2 Soil Moisture Processor Development Continuation Project, 2011. [Online]. Available: http://www.cesbio.upstlse.fr/data_all/SMOS-doc/atbd/ATBD_v3_4_20110124.pdf M. J. Escorihuela, A. Chanzy, J. P. Wigneron, and Y. Kerr, “Effective soil moisture sampling depth of L-band radiometry: A case study,” Remote Sens. Environ., vol. 114, no. 5, pp. 995–1001, May 2010. R. Panciera, J. P. Walker, J. D. Kalma, E. J. Kim, K. Saleh, and J. P. Wigneron, “Evaluation of the SMOS L-MEB passive microwave soil moisture retrieval algorithm,” Remote Sens. Environ., vol. 113, no. 2, pp. 435–444, Feb. 2009. I. Corbella, F. Torres, N. Duffo, V. Gonzalez-Gambau, I. Duran, M. Pablos, and M. Martin-Neira, “Some results on SMOS-MIRAS calibration and imaging,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 3768–3771. I. Corbella, F. Torres, N. Duffo, V. Gonzalez-Gambau, M. Pablos, I. Duran, and M. Martin-Neira, “MIRAS calibration and performance: Results from the SMOS in-orbit commissioning phase,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 9, pp. 3147–3155, Sep. 2011. J. T. Dall’Amico, F. Schlenz, A. Loew, and W. Mauser, “SMOS soil moisture validation: Status at the upper Danube Cal/Val site eight months after launch,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 3801–3804. C. Gruhier, Y. Kerr, F. Cabot, A. Mialon, T. Pellarin, P. De Rosnay, and M. Grippa, “SMOS L-band brightness temperatures compared to other passive microwaves sensors over West Africa area,” in Proc. Living Planet Symp., Bergen, Norway, Jun. 28–Jul. 2, 2010. W. Wagner, V. Naeimi, K. Scipal, R. A. M. De Jeu, and J. MartinezFernandez, “Soil moisture from operational meteorological satellites,” Hydrogeol. J., vol. 15, no. 1, pp. 121–131, Feb. 2007.
[26] C. S. Ruf, S. M. Gross, and S. Misra, “RFI detection and mitigation for microwave radiometry with an agile digital detector,” IEEE Trans. Geosci. Remote Sens, vol. 44, no. 3, pp. 694–706, Mar. 2006. [27] European Space Agency, SMOS Water Mission Winning Battle With Interference, 2010. [Online]. Available: http://www.esa.int/SPECIALS/ smos/SEMB5USOREG_0.html [28] A. Camps, J. Gourrion, J. M. Tarongi, A. Gutierrez, J. Barbosa, and R. Castro, “RFI analysis in SMOS imagery,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 2007–2010. [29] M. K. Hallikainen, J. Seppäanen, J. Hakkarainen, A. Rautiainen, and K. Aalto, “Studies of radio frequency interference at L-band using an airborne 2-D interferometric radiometer,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 2490–2491. [30] N. Skou, J. E. Balling, S. S. Sobjarg, and S. S. Kristensen, “Surveys and analysis of RFI in the SMOS context,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 2011–2014. [31] N. Skou, S. Misra, J. E. Balling, S. S. Kristensen, and S. S. Sobjaerg, “L-band RFI as experienced during airborne campaigns in preparation for SMOS,” IEEE Trans. Geosci. Remote Sens., vol. 48, no. 3, pp. 1398–1407, Mar. 2010. [32] E. Anterrieu, “On the detection and quantification of RFI in L1a signals provided by SMOS,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 10, pp. 3986–3992, Oct. 2011. [33] T. J. Jackson, M. H. Cosh, R. Bindlish, P. J. Starks, D. D. Bosch, M. Seyfried, D. C. Goodrich, M. S. Moran, and J. Du, “Validation of advanced microwave scanning radiometer soil moisture products,” IEEE Trans. Geosci. Remote Sens., vol. 48, no. 12, pp. 4256–4272, Dec. 2010. [34] C. Rüdiger, J. P. Walker, and Y. H. Kerr, “On the airborne spatial coverage requirement for microwave satellite validation,” IEEE Trans. Geosci. Remote Sens. Lett., vol. 8, no. 4, pp. 824–828, Jul. 2011. [35] P. de Rosnay, C. Gruhier, F. Timouk, F. Baup, E. Mougin, P. Hiernaux, L. Kergoat, and V. LeDantec, “Multi-scale soil moisture measurements at the Gourma meso-scale site in Mali,” J. Hydrol., vol. 375, no. 1/2, pp. 241–252, 2009. [36] L. Brocca, F. Melone, T. Moramarco, and R. Morbidelli, “Spatial– temporal variability of soil moisture and its estimation across scales,” Water Resour. Res., vol. 46, p. W02 516, Feb. 2010. doi:10.1029/ 2009WR008016. [37] D. G. Miralles, W. T. Crow, and M. H. Cosh, “Estimating spatial sampling errors in coarse-scale soil moisture estimates derived from pointscale observations,” J. Hydrometeorol., vol. 11, no. 6, pp. 1423–1429, Dec. 2010. [38] A. Loew and F. Schlenz, “A dynamic approach for evaluating coarse scale satellite soil moisture products,” Hydrol. Earth Syst. Sci., vol. 15, no. 1, pp. 75–90, Jan. 2011. doi:0.5194/hess-15-75-2011. [39] M. Drusch, E. F. Wood, and H. Gao, “Observation operators for the direct assimilation of TRMM microwave imager retrieves soil moisture,” Geophys. Res. Lett., vol. 32, p. L15 403, Aug. 2005. doi:10.1029/2005GL023623. [40] L. Brocca, F. Melone, T. Moramarco, W. Wagner, and S. Hasenauer, “ASCAT soil wetness index validation through in-situ and modeled soil moisture data in central Italy,” Remote Sens. Environ., vol. 114, no. 11, pp. 2745–2755, Nov. 2010. [41] W. Wagner, G. Lemoine, and H. Rott, “A method for estimating soil moisture from ERS scatterometer and soil data,” Remote Sens. Environ., vol. 70, no. 2, pp. 191–207, Nov. 1999. [42] D. Penna, H. J. Tromp-van Meerveld, A. Gobbi, M. Borga, and G. Dalla Fontana, “The influence of soil moisture on threshold runoff generation processes in an alpine headwater catchment,” Hydrol. Earth Syst. Sci., vol. 15, no. 3, pp. 689–702, Mar. 2011. [43] C. Albergel, C. Rüdiger, T. Pellarin, J. C. Calvet, N. Fritz, F. Froissard, D. Suquia, A. Petitpa, B. Piguet, and E. Martin, “From near-surface to root-zone soil moisture using an exponential filter: An assessment of the method based on in-situ observations and model simulations,” Hydrol. Earth Syst. Sci., vol. 12, pp. 1323–1337, 2008. [44] C. Albergel, J. C. Calvet, J. F. Mahfouf, C. Rüdiger, A. L. Barbu, S. Lafont, J. L. Roujean, J. P. Walker, M. Crapeau, and J. P. Wigneron, “Monitoring of water and carbon fluxes using a land data assimilation system: A case study for southwestern France,” Hydrol. Earth Syst. Sci., vol. 4, no. 6, pp. 1109–1124, Jun. 2010. [45] A. L. Barbu, J. C. Calvet, J. F. Mahfouf, C. Albergel, and S. Lafont, “Assimilation of soil wetness index and leaf area index into the ISBAA-gs land surface model: Grassland case study,” Biogeosciences, vol. 8, no. 7, pp. 1971–1986, Jul. 2011. [46] B. Bisselink, E. Van Meijgaard, A. J. Dolman, and R. A. M. De Jeu, “Initializing a regional climate model with satellite-derived soil moisture,” J. Geophys. Res., vol. 116, p. D02 121, Jan. 2011.
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
[47] A. Ceballos, K. Scipal, W. Wagner, and J. Martınez-Fernandez, “Validation of ERS scatterometer derived soil moisture data in the central part of the Duero Basin, Spain,” Hydrol. Process., vol. 19, no. 8, pp. 1549–1566, May 2005. [48] C. Albergel, C. Rüdiger, D. Carrer, J. C. Calvet, N. Fritz, V. Naeimi, Z. Bartalis, and S. Hasenauer, “An evaluation of ASCAT surface soil moisture products with in-situ observations in southwestern France,” Hydrol. Earth Syst. Sci., vol. 13, pp. 115–124, 2009. [49] P. Matgen, S. Heitz, S. Hasenauer, C. Hissler, L. Brocca, L. Hoffmann, W. Wagner, and H. H. G. Savenije, “On the potential of METOP ASCATderived soil wetness indices as a new aperture for hydrological monitoring and prediction: A field evaluation over Luxembourg,” Hydrol. Process., 2011, to be published. [50] European Space Agency, SMOS Level 2 Data Release to General Science Community, 2010. [Online]. Available: http://earth.esa.int/ object/index.cfm?fobjectid=7246 [51] CESBIO (Centre d’Etudes Spatiales de la BIOsphère), SMOS L2 Data is Reprocessed!, 2011. [Online]. Available: http://www.cesbio.ups-tlse.fr/ SMOS_blog/?p=1898 [52] European Space Agency, SMOS Anomalies—MIRAS Unavailability, 2011. [Online]. Available: http://earth.esa.int/object/index.cfm? fobjectid=7408 [53] A. Al Bitar, Y. Kerr, O. Merlin, P. Richaume, A. K. Sahoo, and E. F. Wood, “Validation of SMOS soil moisture over SCAN/SNOTEL network,” Geophys. Res. Abstracts, vol. 13, p. 1, 2011. doi:EGU2011-8700-1. [54] C. Rüdiger, J. P. Walker, Y. H. Kerr, O. Merlin, A. Mialon, and E. J. Kim, “SMOS L1c and L2 product validation with airborne and in-situ observations over south-eastern Australia,” Geophys. Res. Abstracts, vol. 13, p. 1, 2011. doi:EGU2011-5183. [55] M. Piles, A. Camps, M. Vall-llossera, I. Corbella, R. Panciera, C. Rüdiger, Y. H. Kerr, and J. Walker, “Downscaling SMOS-derived soil moisture using MODIS visible/infrared data,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 9, pp. 3156–3166, Sep. 2011.
Teodosio Lacava (M’12) was born in Potenza, Italy, in 1972. He received the M.S. degree in geological sciences and the Ph.D. degree in methods and technologies for environmental monitoring from the University of Basilicata, Potenza, in 1999 and 2004, respectively. Since 2005, he has been a Researcher with the Institute of Methodologies for Environmental Analysis (IMAA), National Research Council (CNR), Potenza. He has been a Coinvestigator of several European Union, Italian Space Agency (ASI), and CNR projects. He is the author of several peer-reviewed scientific journal articles and has served as a Referee for many international journals. His main interest is in the field of satellite data analysis for environmental research, particularly regarding the development and assessment of advanced satellite techniques for natural hazard investigation. Dr. Lacava is a member of the American Geophysical Union.
Patrick Matgen was born in Ettelbrück, Luxembourg, on July 13, 1978. He received the M.Sc. degree in environmental engineering from the Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, and the Ph.D. degree from the Water Resources Section, Delft University of Technology, Delft, The Netherlands, with his dissertation being on the retrieval of surface and subsurface water from microwave remote sensing observations and the integration of the data with flood prediction systems. He is currently the Project Leader with the Centre de Recherche Public–Gabriel Lippmann (Grand Duchy of Luxembourg), Belvaux, Luxembourg, where he is responsible for microwave remote sensing and hydrologic-hydraulic modeling projects. His expertise is documented in more than 40 peer-reviewed scientific journal articles.
1621
Luca Brocca was born in Genoa, Italy, in 1978. He received the M.S. degree in environmental engineering and the Ph.D. degree in civil engineering, both with excellence, from the University of Perugia, Perugia, Italy, in 2003 and 2008, respectively. Since 2009, he has been a Researcher with the Research Institute for Geo-Hydrological Protection (IRPI), National Research Council (CNR), Perugia. He is involved in teaching and tutorial activity for the University of Perugia, and he was a Codirector of Master’s thesis. He is the Italian National Correspondent of the Euromediterranean network of Experimental and Representative Basins. He serves as a Reviewer for more than 20 international journals and actively participated to several research projects in the frame of Italian and European programs and in collaboration with Italian and international institutions. He has been the author and the coauthor of 26 journal papers and more than 60 papers in conference proceedings. Dr. Brocca is a member of the Italian Hydrological Society, the Italian Group of Hydraulics (GII), the International Environmental Modelling and Software Society, and the European Geosciences Union. Marco Bittelli received the B.S. degree in agricultural sciences from the University of Bologna, Bologna, Italy, and the M.S. degree and the Ph.D. degree in soil physics from Washington State University, Pullman. He spent one year as a Postdoctoral Scientist in the Physics Department, Heidelberg University, Heidelberg, Germany. Since 2006, he has been a Researcher with the University of Bologna, where he also teaches soil physics, hydrological modeling, and scientific methods courses both at the undergraduate and graduate levels. He is the Coordinator of an International Master in Soil and Water Conservation, established between the University of Bologna, Washington State University, and the University of Florida, Gainesville. He is a Referee for the National Science Foundation (USA). He serves as a Reviewer for over 15 international journals and actively participated in several research projects in the frame of Italian and European programs and in collaboration with Italian and international institutions. He is the author and the coauthor of 22 journal papers, 3 book chapters, and over 40 papers in conference proceedings. Dr. Bittelli is Responsible of the Soil Physics Division, Italian Society of Soil Science. Nicola Pergola (M’11) received the degree in physics from the University of Rome “La Sapienza,” Rome, Italy, in 1993. Since 1998, he has been a Researcher with the Institute of Methodologies for Environmental Analysis (IMAA), National Research Council (CNR), Potenza, Italy, leading the “Geohazards” Unit, IMAA Satellite Remote Sensing Laboratory. He was a member of the Geohazard “Core Team” within the Integrated Global Observing Strategy project, contributing to define the new global observational strategies for geohazard mitigation. He was recently involved in several FP6 and FP7 projects. He was a Scientific Coordinator of different Italian Space Agency (ASI) and CNR projects and a Coinvestigator of several European Union and international (e.g., European Space Agency, North Atlantic Treaty Organization, and International Association for the promotion of cooperation with scientists from the independent states of the former Soviet Union (INTAS)) projects. He is an Associate Editor of the international scientific journal Geomatics, Natural Hazards and Risks (Taylor & Francis). He is the author of about 50 papers on Institute for Scientific Information journal and more than 150 publications and proceedings with peer reviewing and international distribution. He serves as a Referee for several relevant international journals. His main interest is in the field of the development of advanced satellite data analysis, particularly regarding high-temporal-resolution sensors, like the National Oceanic and Atmospheric Administration–Advanced Very High Resolution Radiometer, Earth Observing System–Moderate Resolution Imaging Spectroradiometer, and Meteosat Second Generation–Spinning Enhanced Visible and Infrared Imager, for environmental research and applications. In particular, his research activity is mainly focused on the development of robust satellite techniques for natural hazard investigation from space and on multitemporal satellite data analysis in the space–time domain, particularly using infrared satellite radiances. Dr. Pergola is a member of the European Geosciences Union.
1622
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
Tommaso Moramarco received the M.S. degree in civil hydraulic engineering from the University of Bari, Bari, Italy, in 1989. He is currently a Researcher with the Research Institute for Geo-Hydrological Protection (IRPI), National Research Council (CNR), Perugia, Italy. Since 1989, he has conducted research in the field of hydrological processes addressed to flood forecasting and hydraulic risk mitigation. He has developed original works in these areas, producing a great number of original contributions in the leading hydrologic and hydraulic journals. In particular, his works include hydrometeorological monitoring, synthesis of the effects of spatial variability in hydraulic quantities and scaling, rainfall spatial distribution analysis, entropy theory applied to natural channels, modeling of overland flow over hillslopes, modeling of surface water over watersheds, flood forecasting, hydraulic risk, climate change, drought, safety of dams, artificial neural networks, and genetic algorithm. Since 2001, he has played a lead role in the hydrologic research at IRPI, CNR, and he has been a Coordinator and a Scientist responsible of the projects on hydrometeorological monitoring, flood forecasting, and hydraulic risk in the frame of Italian and European programs. He has advised authorities for establishing measures for flood protection also addressed to the forward planning. He is a Referee of leading hydrologic and hydraulic journals, an Associate Editor of Journal of Hydrologic Engineer [American Society of Civil Engineer (ASCE)], and a Guest Editor of the Hydrology and Earth System Science journal. Mr. Moramarco was the recipient of the Normal Medal from ASCE for his two works on flood routing in natural channels. Afterward, the American Academy of Water Resource Research has conferred on him the Diplomate, Water Resource Engineer.
Valerio Tramutoli was born on December 28, 1957. He received the M.S. degree in physics from the University of Rome “La Sapienza,” Rome, Italy. Since 1990, he has been with the Department of Engineering and Physics of the Environment (DIFA), University of Basilicata, Potenza, Italy, where he has been a Senior Researcher since 1993 and holding courses of satellite remote sensing in the Faculty of Sciences and Faculty of Engineering since 1997. Since 1991, he has been a Visiting Scientist in the main international centers involved in the Earth’s observation by satellite, taking part in several international projects and initiatives of the main space agencies like the European Space Agency, National Aeronautics and Space Administration, National Space Development Agency, and Italian Space Agency (ASI). He has been the National Coordinator of the SEISMASS Project (funded by ASI) Principal Investigator, or responsible of DIFA participation, to several international projects funded by the North Atlantic Treaty Organization and by European Community in the framework of the Science for Peace, FP6-IST, FP6-INTAS, FP6+FP7 Global Monitoring for Environment and Security (GMES) initiatives. Since 2010, he has been the Coordinator of the European project PRE-EARTHQUAKES funded by EC in the framework of the FP7-GMES-Space Program. His research activity has been focused on the development of new satellite sensors and techniques for natural, environmental, and technological hazard monitoring and mitigation. In this context, in 1998, he proposed the original Robust AVHRR Techniques (now, Robust Satellites Techniques) change detection approach successfully used in a large spectrum of applications. He has been among the few Scientists invited to participate, since 2001, to the Geohazard Core Team, Integrated Global Observing Strategy, instructed (by the most important space agencies, e.g., Committee on Earth Observation Satellites, Food and Agriculture Organization, United Nations Educational, Scientific and Cultural Organization, and International Council of Scientific Unions) to define the new observational strategies for geohazard mitigation for the next decade. He act as a Referee for the most important journals in the field and as a Project Evaluator for several funding agencies at European and extra-European levels. Since 2009, he has been a Member of the Editorial Board of the Geomatics, Natural Hazards and Risk international journal. He is the author or coauthor of more than 150 papers published in international journals, scientific books, and international conference proceedings. Mr. Tramutoli has been the Chair, a Cochair, an Organizer, or an Invited Speaker in the most important international conferences [European Geosciences Union, American Geophysical Union, and the International Union of Geodesy and Geophysics (IUGG)]. Since 2007, he has been a permanent member of IUGG Inter-Association Working Group on Electromagnetic Studies of Earthquakes and Volcanoes.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
A First Assessment of the SMOS Soil Moisture Product With In Situ and Modeled Data in Italy and Luxembourg Teodosio Lacava, Member, IEEE, Patrick Matgen, Luca Brocca, Marco Bittelli, Nicola Pergola, Member, IEEE, Tommaso Moramarco, and Valerio Tramutoli
Abstract—The European Space Agency Soil Moisture and Ocean Salinity (SMOS) mission was launched on November 2, 2009. Providing accurate soil moisture (SM) estimation is one of its main scientific objectives. Since the end of the commissioning phase, preliminary global SMOS SM data [Level 2 (L2) product] are distributed to users. In this paper, we carried out a first assessment of the reliability of this product through a comparison with in situ observed and modeled SM over three different sites: One is located in Luxemburg, and two are located in Italy. The period from August 1, 2010, to July 1, 2011, has been analyzed, giving us the opportunity to evaluate the satellite response to different SM states. The selected period is important for hydrological predictions as it is typically characterized by a sequence of transitions from dry to wet and from wet to dry conditions. In order to compare SMOS and ground SM measurements, a two-step approach has been applied. First, an exponential filter has been applied to approximate root-zone SM, and second, a cumulative distribution function matching has been employed to remove systematic differences between satellite and in situ observations and model simulations of SM. Our results indicate rather good reliability of the filtered and bias-corrected SM estimates derived from the first SMOS L2 products. Bearing in mind that an updated/advanced version of the SMOS SM product has been recently produced, our preliminary results already seem to confirm the potential of SMOS for monitoring of water in soils. Index Terms—Remote sensing, soil moisture (SM), Soil Moisture and Ocean Salinity (SMOS), validation.
I. I NTRODUCTION
T
HERE is ample evidence that hydrological processes are significantly conditioned by a river basin’s wetness state, e.g., [1] and [2]. Many field studies show that the distribution Manuscript received March 31, 2011; revised September 28, 2011, December 17, 2011, and January 18, 2012; accepted January 22, 2012. Date of publication March 12, 2012; date of current version April 18, 2012. T. Lacava and N. Pergola are with the Institute of Methodologies for Environmental Analysis (IMAA), National Research Council (CNR), 85050 Potenza, Italy (e-mail: [email protected]; [email protected]). P. Matgen is with the Centre de Recherche Public–Gabriel Lippmann, 4422 Belvaux, Luxembourg (e-mail: [email protected]). L. Brocca and T. Moramarco are with the Research Institute for GeoHydrological Protection (IRPI), National Research Council (CNR), 06128 Perugia, Italy (e-mail: [email protected]; [email protected]). M. Bittelli is with the Department of Agro-Environmental Science and Technology, University of Bologna, 40126 Bologna, Italy (e-mail: marco. [email protected]). V. Tramutoli is with the Department of Engineering and Physics of the Environment (DIFA), University of Basilicata, 85100 Potenza, Italy (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TGRS.2012.2186819
of a catchment’s water is controlled by soil water storage, with runoff yield abruptly rising when a certain storage threshold is exceeded, e.g., [3]–[5]. Penna et al. [6] argue that the soil water stored in hillslope areas is released only during wetter conditions, when flow paths between the hillslope and riparian zone become connected. Therefore, monitoring the closeness to thresholds is essential to accurately predict stream responses to rainfall events. One approach to do this is through the measurement of surface soil moisture (SM) (SSM), the latter being considered as a proxy of soil water storage that enables the periodical assessment of the readiness of a river basin to generate a fast discharge response. Apart from in situ measurements and hydrological models, SM can be estimated with sensor data obtained from satellite platforms. In the last years, the capability of Earth observation systems to provide reliable SM measurements has been widely investigated. In particular, the data acquired by microwave sensors, both active and passive, have been used, providing detailed information about SM variability in the space–time domain [7]–[9]. The launch of the Soil Moisture and Ocean Salinity (SMOS) mission, the European Space Agency (ESA) dedicated SM mission, in November 2009 [10]–[13], clearly indicates the need and the will of the international scientific community of having a better SM estimation from space. The main appeal of the Microwave Imaging Radiometer with Aperture Synthesis (MIRAS) on board SMOS over existing sensors is that the emitted microwave radiation is measured at L-band (1.4 GHz) at a reasonable spatial resolution (the average ground resolution is 43 km) whereas previous spaceborne radiometers operated at frequencies of 6.6 GHz (C-band) and higher. At L-band and depending on soil characteristics and wetness conditions, 5 cm of soil is probed on average [14]. As the thickness of the sampled layer by the radiometers increases with increasing wavelength and due to the fact that the vegetation optical depth and other perturbations less impact lower frequencies, it is expected that L-band radiometers such as MIRAS show the highest potential for SM retrieval studies [7], [15]. Moreover, the vegetation and soil contributions to the signal can be separated more easily as MIRAS images any point of the surface at several angles [11], [12]. One of the several possible retrieval models for inverting SM from dualpolarization multiangular brightness temperature observations of SMOS is the L-band Microwave Emission of the Biosphere (L-MEB) model [16]–[18]. The validity of the model has been verified in a number of tower-based case studies over mostly
0196-2892/$31.00 © 2012 IEEE
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
1613
TABLE I M AIN C HARACTERISTICS OF THE T EST S ITES U SED FOR T HIS S TUDY (LAT: L ATITUDE , LON: L ONGITUDE , E LEV: E LEVATION , MAR: M EAN A NNUAL R AINFALL , AND MAT: M EAN A NNUAL T EMPERATURE )
homogeneous land surface conditions [15], [19]. In a similar experiment, Panciera et al. [20] applied the L-MEB model with the proposed set of model parameters over grassland and cropland cover types in Australia. They showed that the targeted accuracy level of 0.04-m3 /m3 volumetric SM could only be reached over grassland-covered sites while significant underestimations of SSM were found over crop-covered areas. Initial results with actual SMOS data seem to confirm the good sensitivity of MIRAS to SM variations [21]–[24]. A “worrying problem” [25] for passive microwave remote sensing, particularly radiometry with wavelengths longer than C-band, is the radio frequency interference (RFI). As a result, brightness temperature may be degraded over densely populated areas at the risk of rendering SM retrievals impossible [26]. Such a problem becomes even more relevant when it happens in what should be a protected band reserved for the Earth Exploration Satellite Service, space research, and radio astronomy [27]. The first results show that L-band SMOS data, in some places, are also contaminated by RFI [28]–[32]. Kerr [12] qualifies the pixel heterogeneity and the significant differences in behavior from different targets that go with it as the most stringent limitation currently impacting the accuracy of SMOS-derived SM estimates. Other possible limitations that are currently not well mastered are related to topography, frozen soils, snow cover, urban areas, permanent water bodies, bare dry soil, litter, and forest emission. It is well known that an important issue when validating coarse-resolution satellite SM products with ground measurements relates to the disparity of spatial scales between the two data sets [33], [34]. However, several studies have shown that point measurements of SM can be representative of larger areas (e.g., [35]–[37]), thereby reducing the commensurability problem. However, systematic differences between remote-sensingderived and in situ observations are usually detected even though the temporal dynamics are very similar, e.g., [38]. To address this problem, several techniques have been developed for matching the variability of satellite data with in situ ones, e.g., [8], [9], [33], [39], and [40]. In this context, the objective of this paper is to carry out a field evaluation of the SMOS Level 2 (L2) SM product with ground measurements carried out in three experimental research basins: One is located in Luxembourg, and two are located in Italy. Such an approach gives the opportunity to assess SMOS SM data sensitivity under different observation conditions (i.e., latitude, topography, snow presence, land cover uses, vegetation presence, etc.). In particular, the study covers the period August 1, 2010–July 1, 2011, thereby allowing us to evaluate the capability of the sensor to capture the transitions
from dry to wet as well as from wet to dry conditions and to verify possible limitations related to the presence of snow. The SMOS SM product is compared with both in situ and modeled SM data to provide a robust assessment of its reliability. II. M ETHODS Due to the difference in sensing depth between satellite (e.g., ∼5 cm for MIRAS) and in situ sensors (5–10 cm for this study; see Table I), the semiempirical approach proposed by Wagner et al. [41], also named “exponential filter,” is used to obtain the root-zone SM (RZSM) product from the SSM directly sensed by satellite sensors. In addition, in order to match satellite time series (both SSM and RZSM) to in situ ones, the cumulative distribution function (CDF) matching approach (e.g., [39]) is employed. This two-step approach is briefly described in the following along with the soil water balance model that has been used for the computation of the modeled SM data. A. Exponential Filter The profile SM values are often reasonably well correlated with SSM values (e.g., [42]) because both are an expression of meteorological conditions during the preceding days and weeks. The exponential filter, introduced by Wagner et al. [41], is a simple and effective method to retrieve profile SM values from surface ones. It relies on the analytical solution of a differential equation, assuming that the variation in time of the average profile SM is linearly related to the difference between the surface and the profile values. The method is dependent on only one parameter, the characteristic time length T , that characterizes the temporal variation of SM within the root-zone profile. Despite its simplicity, the algorithm was found reliable in predicting profile SM values based on SSM information using both in situ observed [43]–[45] and modeled [46] data. Because of the higher sensing depths of L-band radiometers, it can be expected that there is a reduced need for implementing a transformation routine. However, it is also worth mentioning that a low-pass filter also tends to reduce random noise affecting the investigated raw time series. For a detailed description of the method, the reader may refer to [41] and [47]. In this study, the recursive formulation of the method is used [48]. B. CDF Matching Approach The systematic differences between remote-sensing-derived and in situ observations of SM prevent the absolute agreement between the two time series. Consequently, the comparison
1614
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
of these two time series is often aided by normalizing the remotely sensed data to better match the distribution of ground measurements. In this study, the CDF matching approach (e.g., [39]) that can be considered as an enhanced nonlinear technique for removing systematic differences between two data sets is implemented. Through this method, the satellite data are rescaled in such a way that its CDF matches the CDF of in situ measurements. The CDF matching is performed on the SSM (and RZSM) time series, yielding SSM-CDF (and RZSM-CDF).
C. Soil Water Balance Model In addition to in situ observations, modeled SM data are used for the validation of the SMOS SM products. The structure of the soil water balance model used in this study was derived by using SM observations carried out in an experimental catchment located in central Italy [2]. The model has been applied in different test sites located across Europe with good performances obtained from all sites [9], [40]. The model, which requires routinely measured meteorological variables (rainfall and air temperature) as input data, incorporates only five parameters, and the default time step is 1 h. Moreover, the model was found reliable even when it was calibrated only with a limited number of observations [2], thus allowing to confidently apply the model over large areas and for periods different from those employed for parameter calibration. It is also worth mentioning that the water balance model is forced with rainfall and temperature data averaged over an area of 15 km by 15 km surrounding each test site. As the spatial resolution of the hydrological model is similar to the footprint of the SMOS satellite, the modeled SM data are representative of an area larger than the one covered by in situ measurements, which arguably reduces the problem of the spatial mismatch between ground and satellite data.
III. S TUDY A REAS AND SM M EASUREMENTS The three test sites used for the validation of the SMOS SM products are located in Italy (two sites) and Luxembourg (one site). These sites have been used in previous studies [11], [49] for the validation of coarse-resolution satellite SM products retrieved by active and passive microwave sensors. The calibration of the sensors against gravimetric SM measurements was carried out for each site, obtaining a good accuracy (less than 0.02 m3 /m3 ). Therefore, the in situ SM data can be assumed to be sufficiently accurate and representative of the SM variability in the investigated areas. Table I synthesizes the main characteristics of each site in terms of location, soil texture, vegetation cover, and measurement depth, while Fig. 1 shows the location of the study sites with respect to the SMOS satellite footprint. For all the investigated sites, in situ SM observations are available for the period from August 2010 to July 2011 with a temporal resolution of 1 h. The SMOS SM evaluation is carried out with both in situ observations and modeled data as reference data. In the following, the main characteristics of the three test sites are described. Note that, in all test sites, rainfall and temperature data are collected as forcing data for driving the soil water balance model. A. Italy—CER The Cerbara (CER) site is located in central Italy, and it was installed by the Hydrology Service of Umbria region for civil protection activities. The area is characterized by a Mediterranean semihumid climate with an average precipitation of 900 mm/year and a mean annual temperature of 13 ◦ C. The SM probe, based on time-domain reflectometry (TDR) technique (ThetaProbe ML2X, Delta-T Device), was set up in November 2009 and continuously measures volumetric SM in the soil column at 10-, 20-, and 40-cm depths; the data at 10-cm depth are considered for this study.
D. Performance Index
B. Italy—SPC
For each comparison, the two following statistical indexes are used to evaluate the SMOS SM product accuracy: (1) RM SD = (SMSAT − SMINSITU )2 (SMSAT − SMINSITU )2 (2) R= 1− (SMINSITU − SMINSITU )2
The San Pietro Capofiume (SPC) site is located in northern Italy (Emilia–Romagna region), and it was installed by the Service of Hydrology, Meteorology and Climate of the Regional Agency for Environmental Protection in Emilia–Romagna (ARPA-SIMC), mainly for agricultural purposes. The area is characterized by a Mediterranean climate with an average precipitation of 750 mm/year and a mean annual temperature of 16 ◦ C. For this site, TDR volumetric SM measurements (TDR100, Campbell Scientific Inc.) have been collected at seven depths at an hourly time step; the data at 10-cm depth are used for this study.
BIAS = (SMSAT − SMINSITU )
(3)
where RM SD is the root mean squared difference between the in situ (observed and modeled) SMINSITU and the different satellite SM products SMSAT , the overbar is the mean operator, R is the correlation coefficient, and BIAS is the mean difference between in situ and satellite products. It has to be noted that, instead of RM SE (where E stands for “error”), we used RM SD to underline that also ground measurements contain errors (instrumental and representativeness) and, hence, they cannot be considered as the “true” SM.
C. Luxembourg In the southern part of Luxembourg, the experimental Bibeschbach (BIB) catchment (10.8 km2 ) is considered as the third test site. It is characterized by a gently rolling landscape and a typical humid temperate climate with an average precipitation of about 740 mm/year and a mean annual temperature
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
Fig. 1.
1615
Location and land cover of the three test sites in Luxembourg and Italy with the boundaries of the SMOS pixels for those areas.
of 9 ◦ C. The site is mainly covered by forest and agricultural land on loamy soils. Since 2005, the basin has been equipped with a set of 40 sensors (ECH2O, DecagonTM) measuring the volumetric SM of the topsoil layer at a depth of 4–7 cm using the capacitance principle. The measurement sites are located on the edge of grasslands or agricultural plots and are distributed throughout the catchment. The SM sensors are connected to data loggers that record an hourly dielectric constant of the medium. A calibration with regular TDR point measurements is performed to convert the signal into volumetric SM content. The calibration equation is established for each measurement site individually. All measurements exhibit a similar temporal pattern of SM fluctuations. For this study, the area-averaged volumetric SM of the near-surface SM probes (∼5 cm depth) is considered. IV. SMOS S ATELLITE P RODUCTS The SMOS Soil Moisture User Data Product 2 (SMUDP2), provided by ESA in the framework of a Cat-1 project, has been used in this study. In particular, such data, delivered by ESA [50], have been generated by the SMOS L2 SM processor, taking Level 1c (L1c) and other auxiliary data (i.e., brightness temperature, polarization, radiometric uncertainty, incidence
angle, azimuth angle, elliptical footprint semi-axes) as input data. Using recalibrated L1c data, an updated version of the L2 data has been produced recently by the Centre d’Etudes Spatiales de la Biosphère [51]. SMUDP2 data have a spatial resolution of 43 km on average and a geolocation error of 400 m; the data are mapped on an icosahedral Snyder equal area hexagonal grid of aperture 4 and resolution 9 (ISEA 4H9) with a spatial resolution of 15 km (see also Fig. 1). In the observation period (i.e., August 1, 2010–July 1, 2011), SMUDP2 has been generated regularly, except in the period December 27, 2010–January 1, 2011, probably due to some SMOS anomalies [52]. V. R ESULTS First, the modeled SM data are computed for each test site by optimizing the soil water balance model parameters against in situ observations. Fig. 2 shows the comparison between modeled and observed data for the three sites, highlighting the good model performance. In particular, the correlation coefficient R values between the observed and modeled data equal 0.97, 0.93, and 0.84 for the CER, SPC, and BIB sites, respectively (for BIB, no modeled data were achievable for August 2010); the root mean squared difference RM SD is always lower than 0.025 m3 /m3 . The agreement between modeled and observed
1616
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
Fig. 2. Time series of in situ observed and modeled SM for (a) CER, (b) SPC, and (c) BIB sites. The hourly rainfall pattern is also shown.
data indicates that the ground observations are not much affected by measurement errors or any other type of error, thus representing an appropriate benchmark for SMOS data except for their spatial representativeness. It is worth noting that, because of the calibration of model parameters with ground measurements, the time series of in situ and modeled SM cannot be considered as two independent validation data sets. However, the integration of in situ measurements and model results enables a more comprehensive and efficient assessment of the reliability of the SMOS SM product as it reduces the impact of ground data uncertainties related to measurement errors in the case of in situ data as well as forcing data and model parameters in the case of modeled data. For the comparison between ground and satellite data, the following procedure is adopted. For each test site, the satellite pixel whose centroid is the nearest to the in situ location is identified and considered for the subsequent analysis. Then, the in situ and modeled SM dates corresponding to the acquisition time of the SMOS passes are retrieved, and the corresponding data values are used for the SMOS evaluation. Next, the satellite RZSM product is computed starting from the SSM time series by optimizing the value of the characteristic time length parameter T obtained by maximizing the correlation between the RZSM and the in situ time series. Finally, the CDF matching approach is applied to satellite data (SSM and RZSM) to match
their CDF to the CDF of the in situ data, thereby yielding biascorrected SSM-CDF and RZSM-CDF time series. The results of all the comparisons between SMOS SM products (SSM and RZSM) and in situ data in terms of RM SD, BIAS, and R values are reported in Table II both with respect to the original values and the rescaled data using the CDF matching approach. The sample size and the optimized value of the characteristic time length parameter T , used for RZSM computation, are also indicated. Therefore, the field evaluation is carried out separately with three soil wetness products (i.e., with and without the application of the exponential filter and the bias removal). As all intermediate results are provided in the manuscript, the utility and the effect of each processing step can be investigated separately. Fig. 3 shows the time series of the three SMOS SM products (SSM, SSM-CDF, and RZSMCDF) against in situ observations for the three sites while Fig. 4 shows the corresponding scatter plots. In Figs. 3 and 4, it can be seen that the SSM-CDF and RZSM-CDF SM products are able to reproduce the temporal pattern of ground observations with sufficient accuracy, even at a fine temporal scale. On the other hand, when the original SSM product is considered, a tendency to underestimate the ground observations of SM is apparent for all the sites (see also the BIAS values reported in Table II). This observation is well in line with other studies dealing with SMOS SM validation [53], [54].
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
1617
TABLE II S UMMARY OF THE R ESULTS OF THE C OMPARISON B ETWEEN IN SITU AND SMOS SM P RODUCTS (SSM: S URFACE SM, RZSM: ROOT-Z ONE SM, R: C ORRELATION C OEFFICIENT, RMSD: ROOT M EAN S QUARE D IFFERENCE , CDF: C UMULATIVE D ISTRIBUTION F UNCTION M ATCHING , T: C HARACTERISTIC T IME L ENGTH , AND N: S AMPLE S IZE ). T HE RMSD AND THE BIAS A RE E XPRESSED IN m3 /m−3
Fig. 3. Time series of in situ SM, OBS, and SMOS SM products for (a) CER, (b) SPC, and (c) BIB sites (SSM: Surface SM, RZSM: Root-zone SM, and CDF: Cumulative distribution function matching approach).
For the investigated 11-month period, while 209 SMOSderived SM estimates are available for the CER site, the sample size is close to 300 for the other two study areas (equal to 309 for the SPC site and 292 for the BIB site). For the BIB site, no data are available for the month of December due to
a continuous presence of snow, which did not allow any SM retrieval from SMOS data. The difference in the number of acquisitions can be explained also by the occasional occurrence of RFI that has hampered SM retrieval, mainly for the CER site.
1618
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
Fig. 4. Comparison of in situ SM, OBS, and SMOS SM products for (a) CER, (b) SPC, and (c) BIB sites (SSM: Surface SM, RZSM: Root-zone SM, CDF: Cumulative distribution function matching approach).
Despite the existing scale mismatch between the satellite footprint and the area covered by ground measurements, the results in terms of correlation are rather satisfactory with R-values ranging between 0.314 and 0.637 (on average, equal to ∼0.518) for the original SSM and the SSM-CDF products. On the other hand, the absolute SM values are characterized by high RM SD values that are equal, on average, to 0.108 m3 /m3 , mainly due to the large underestimation that is visible in Fig. 3. After removing systematic differences through the CDF matching approach, RM SD values are significantly decreased with an average value that is now equal to 0.054 m3 /m3 . On the other hand, it can be seen that the agreement is higher when modeled data are used rather than in situ measurements (both in terms of R and RM SD), arguably because the spatial scale covered by the model is similar to the SMOS pixel size. If the exponential filter is applied, thus obtaining the RZSM product, R values increase significantly (0.617–0.810), but the RM SD is still characterized by values (0.059–0.123 m3 /m3 ) that are fairly similar to those that were found for the unprocessed SSM product. On the other hand, the filtered and rescaled SMOS SM product, i.e., RZSM-CDF, provides the best results with R values ranging between 0.591 and 0.877 and RM SD values between 0.024 and 0.046 m3 /m3 . With respect to the obtained T values, for the comparison with in situ observations, T is equal to 15 days for all the sites while lower values were obtained with modeled data (on average, equal to nine days). The obtained T values are higher than the ones obtained by Kerr et al. [11] and Matgen et al. [49] who compared different SM products, obtained by the Advanced Scatterometer and Advanced Microwave Scanning Radiometer sensors, with in situ observations over the same test sites. This unexpected result supports the assumption that the main effect of applying an exponential filter is to reduce random errors affecting the time series. This result is arguably due to the high temporal variability of the first delivery of the SMOS SM product (see Fig. 3). As a consequence, the exponential filter tends to smoothen the temporal fluctuations, and by reducing the noise level in the raw data provides a better agreement with ground observations. This would explain why higher values for T were obtained than in previous studies dealing with sensors characterized by lower sensing depths. The result seems to indicate that the exponential filter leads to a removal of noise in the SMOS data set rather
Fig. 5. (a) Correlation coefficient R and (b) root mean squared difference RM SD between in situ observed and modeled and SMOS SM data for all the investigated sites and by separating (asc) ascending and (desc) descending passes (SSM: Surface SM, RZSM: Root-zone SM, and CDF: Cumulative distribution function matching approach).
than to a conversion of SSM into estimates of SM in deeper layers. To further investigate the SMOS reliability, the ascending and the descending passes are identified, and the same comparisons are carried out separately for the two orbits. Fig. 5 summarizes and compares the results in terms of R and RM SD values for all the sites. The selection of data corresponding to the ascending passes between 03:00 and 05:00 Greenwich Mean Time (GMT) slightly improves the results with, for SSM, an average increase of correlation that is equal to 0.054 and a reduction of 0.003 m3 /m3 of the RM SD values. Only considering the descending passes, between 17:00 and 19:00 GMT, clearly reduces the performance. This behavior can be ascribed to several factors: 1) ascending data, particularly for the period investigated, are more stable than descending ones in terms of temperature variations [55], and 2) residual RFI effects, if
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
present, are less important for the ascending passes for which human activity is comparatively low. However, a more in-depth investigation is clearly needed to corroborate these findings. One motivation for the SMOS mission is to improve the effectiveness of flood forecasting systems [12]. Our study provides some promising first results in this respect. The transition from dry to wet and saturated conditions, which is well visible in the in situ and modeled SM time series (Fig. 2) between October 1, 2010, and December 1, 2010, is quite well captured by the RZSM-CDF signal (see Fig. 3). Similarly, in early spring, between March 1, 2011, and May 1, 2011, the RZSMCDF signal over the BIB and SPC sites closely follows the gradual transition from nearly saturated to very dry conditions, while there is a slightly delayed response in the SMOS data for the CER site. This means that, in threshold-mediated catchments where fast storm flow is initiated once a certain SM threshold is approached, there is clearly a potential for filtered and bias-corrected SMOS-derived SM data to serve as a proxy of soil water storage that may help to verify and eventually improve hydrological model predictions. However, it is also obvious from our results that the detection of SM threshold changes is difficult to achieve with the unprocessed SMOS SSM data. Moreover, it is important to note that, in all test sites, the dynamic range of SM fluctuations is rather high, with SM varying by more than 0.25 m3 /m3 between the beginning and end of the investigation period. While our results clearly illustrate the capability of SMOS data for monitoring season-dependent SM variations, one could argue that, even with targeted 0.04 m3 /m3 accuracy, it still remains difficult to effectively detect event-specific SM fluctuations (see, for example, the sudden SM fluctuation observed in June 2011 in both CER and SPC sites that is not captured by the SMOS signal). However, the distinction between periods potentially exhibiting high runoff yields and periods with limited runoff production appears to be feasible with the accuracy levels expected from the SMOS L2 SM data, which is confirmed by the results demonstrated in this paper.
1619
of the SMOS L2 data is significantly larger than the ranges associated with in situ measured and modeled data. Second, the application of the exponential filter and the CDF matching operator yields RZSM estimates that are highly correlated with in situ measured and modeled SM (R between 0.591 and 0.877), indicating a satisfactory capability of SMOS for monitoring SSM variability. The RM SD values between the postprocessed SMOS L2 SM data and the in situ measurements are 0.042 m3 /m3 for the CER site, 0.037 m3 /m3 for the SPC site, and 0.029 m3 /m3 for the BIB site. Moreover, the postprocessed SMOS L2 SM captures the transition from dry to wet as well as from wet to dry conditions that is characteristic of the passage from autumn to winter and from winter to spring, respectively. This result highlights an interesting potential of SMOS SM data for improved hydrological predictions. In fact, by reliably distinguishing between the two main hydrological states, SMOS may help to monitor a hydrological system’s closeness to critical thresholds. Due to the limited sample size and the fact that this is the first delivery of the SMOS SM product, the results of this study need to be considered as preliminary and, clearly, need to be both confirmed and deepened in future investigations. ACKNOWLEDGMENT
VI. C ONCLUSION
The authors would like to thank the Umbria Region (Italy), the Regional Agency for Environmental Protection in Emilia–Romagna (Italy), and the Centre de Recherche Public–Gabriel Lippmann (Luxembourg) for providing the in situ soil moisture (SM) and hydrometeorological data. This research was carried out in the framework of the Category 1 Project 4699: “Assessment of EO-based soil wetness variation maps for geohazards monitoring and mitigation” funded by the European Space Agency (ESA) in the framework of First SMOS Data Announcement of Opportunity. Additional data over Luxembourg were acquired in the framework of the ESA Category 1 Project 8696: “On the value of SMOS MIRASderived volumetric SM for hydrological monitoring and prediction applications: A field evaluation over Luxembourg.”
This paper has investigated the potential of SMOS for capturing the SM variability in three study areas characterized by different hydrometeorological regimes. We are aware that, because of the differences in scale and coverage between the point measurements and the SMOS observations, a justified evaluation with in situ SM measurements remains difficult. However, and in spite of this study’s focus on specific areas (and limited time periods), we feel that there are elements of this investigation that are of more general significance. Also, it is important to take into account that the data set that we used refers to the very first SMOS SM product release. First, the combination of SMOS MIRAS brightness temperature observations with the SMOS L2 SM processor provides SSM estimates that, on all sites, are correlated reasonably well (R between 0.345 and 0.637) with in situ and modeled SM. More specifically, on all sites, the SMOS L2 SM data have a mean value that is significantly lower than the mean of the in situ measurements. Moreover, the dynamical range
[1] A. W. Western and R. B. Grayson, “The Tarrawarra data set: Soil moisture patterns, soil characteristics and hydrological flux measurements,” Water Resour. Res., vol. 34, no. 10, pp. 2765–2768, 1998. [2] L. Brocca, F. Melone, and T. Moramarco, “On the estimation of antecedent wetness condition in rainfall-runoff modelling,” Hydrol. Process., vol. 22, no. 5, pp. 629–642, Feb. 2008. [3] H. J. Tromp-van Meerveld and J. J. McDonnell, “Threshold relations in subsurface stormflow 1. A 147 storm analysis of the Panola hillslope,” Water Resour. Res., vol. 42, p. W02 410, Feb. 2006. doi:10.1029/ 2004WR003778. [4] J. Latron and F. Gallart, “Runoff generation processes in a small Mediterranean research catchment (Vallcebre, Eastern Pyrenees),” J. Hydrol., vol. 358, no. 3/4, pp. 206–220, Sep. 2008. [5] C. Spence, “A paradigm shift in hydrology: Storage thresholds across scales influence catchment runoff generation,” Geography Compass, vol. 4, no. 7, pp. 819–833, Jul. 2010. [6] D. Penna, M. Borga, D. Norbiato, and G. Dalla Fontana, “Hillslope scale soil moisture variability in a steep alpine terrain,” J. Hydrol., vol. 364, no. 3/4, pp. 311–327, Jan. 2009. [7] J. C. Calvet, J. P. Wigneron, J. Walker, F. Karbou, A. Chanzy, and C. Albergel, “Sensitivity of passive microwave observations to soil
R EFERENCES
1620
[8]
[9]
[10]
[11]
[12] [13] [14]
[15]
[16]
[17]
[18]
[19] [20]
[21]
[22]
[23]
[24]
[25]
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
moisture and vegetation water content: L-band to W-band,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 4, pp. 1190–1199, Apr. 2011. T. Lacava, L. Brocca, G. Calice, F. Melone, T. Moramarco, N. Pergola, and V. Tramutoli, “Soil moisture variations monitoring by AMSU-based soil wetness indices: A long-term inter-comparison with ground measurements,” Remote Sens. Environ., vol. 114, no. 10, pp. 2317–2325, Oct. 2010. L. Brocca, S. Hasenauer, T. Lacava, F. Melone, T. Moramarco, W. Wagner, W. Dorigo, P. Matgen, J. Martínez-Fernández, P. Llorens, J. Latron, C. Martin, and M. Bittelli, “Soil moisture estimation through ASCAT and AMSR-E sensors: An intercomparison and validation study across Europe,” Remote Sens. Environ., vol. 115, no. 12, pp. 3390–3408, Dec. 2011. doi:10.1016/j.rse.2011.08.003. Y. Kerr, P. Waldteufel, J. P. Wigneron, J. M. Martinuzzi, J. Font, and M. Berger, “Soil moisture retrieval from space: The Soil Moisture and Ocean Salinity (SMOS) mission,” IEEE Trans. Geosci. Remote Sens., vol. 39, no. 8, pp. 1729–1735, Aug. 2001. Y. Kerr, P. Waldteufel, J. P. Wigneron, S. Delwart, F. Cabot, J. Boutin, M. J. Escorihuela, J. Font, N. Reul, C. Gruhier, S. E. Juglea, M. R. Drinkwater, A. Hahne, M. Martín-Neira, and S. Mecklenburg, “The SMOS mission: New tool for monitoring key,” Proc. IEEE, vol. 98, no. 5, pp. 666–687, May 2010. Y. Kerr, “Soil moisture from space: Where are we?,” Hydrogeol. J., vol. 15, no. 1, pp. 117–120, Feb. 2007. H. Barré, B. Duesmann, and Y. Kerr, “SMOS: The mission and the system,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 587–593, Mar. 2008. S. Delwart, C. Bouzinac, P. Wursteisen, M. Berger, M. Drinkwater, M. Martín-Neira, and Y. H. Kerr, “SMOS validation and the COSMOS campaigns,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 695– 704, Mar. 2008. K. Saleh, J. P. Wigneron, P. Waldteufel, P. de Rosnay, M. Schwank, J. C. Calvet, and Y. Kerr, “Estimates of surface soil moisture under grass covers using L-band radiometry,” Remote Sens. Environ., vol. 109, no. 1, pp. 42–53, Jul. 2007. J. P. Wigneron, Y. Kerr, P. Waldteufel, K. Saleh, M. J. Escorihuela, P. Richaume, P. Ferrazzoli, M.-J. Escorihuela, J. P. Grant, B. Hornbuckle, P. deRosnay, J.-C. Calvet, T. Pellarin, R. Gurney, and C. Mätzler, “L-band Microwave Emission of the Biosphere (L-MEB) model: Description and calibration against experimental data sets over crop fields,” Remote Sens. Environ., vol. 107, no. 4, pp. 639–655, 2007. J. Wigneron, A. Chanzy, Y. H. Kerr, H. Lawrence, S. Jiancheng, M. J. Escorihuela, V. Mironov, A. Mialon, F. Demontoux, P. de Rosnay, and K. Saleh-Contell, “Evaluating an improved parameterization of the soil emission in L-MEB,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 4, pp. 1177–1189, Apr. 2011. Array Systems Computing Inc., Algorithm Theoretical Basis Document (ATBD) for the SMOS Level 2 Soil Moisture Processor Development Continuation Project, 2011. [Online]. Available: http://www.cesbio.upstlse.fr/data_all/SMOS-doc/atbd/ATBD_v3_4_20110124.pdf M. J. Escorihuela, A. Chanzy, J. P. Wigneron, and Y. Kerr, “Effective soil moisture sampling depth of L-band radiometry: A case study,” Remote Sens. Environ., vol. 114, no. 5, pp. 995–1001, May 2010. R. Panciera, J. P. Walker, J. D. Kalma, E. J. Kim, K. Saleh, and J. P. Wigneron, “Evaluation of the SMOS L-MEB passive microwave soil moisture retrieval algorithm,” Remote Sens. Environ., vol. 113, no. 2, pp. 435–444, Feb. 2009. I. Corbella, F. Torres, N. Duffo, V. Gonzalez-Gambau, I. Duran, M. Pablos, and M. Martin-Neira, “Some results on SMOS-MIRAS calibration and imaging,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 3768–3771. I. Corbella, F. Torres, N. Duffo, V. Gonzalez-Gambau, M. Pablos, I. Duran, and M. Martin-Neira, “MIRAS calibration and performance: Results from the SMOS in-orbit commissioning phase,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 9, pp. 3147–3155, Sep. 2011. J. T. Dall’Amico, F. Schlenz, A. Loew, and W. Mauser, “SMOS soil moisture validation: Status at the upper Danube Cal/Val site eight months after launch,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 3801–3804. C. Gruhier, Y. Kerr, F. Cabot, A. Mialon, T. Pellarin, P. De Rosnay, and M. Grippa, “SMOS L-band brightness temperatures compared to other passive microwaves sensors over West Africa area,” in Proc. Living Planet Symp., Bergen, Norway, Jun. 28–Jul. 2, 2010. W. Wagner, V. Naeimi, K. Scipal, R. A. M. De Jeu, and J. MartinezFernandez, “Soil moisture from operational meteorological satellites,” Hydrogeol. J., vol. 15, no. 1, pp. 121–131, Feb. 2007.
[26] C. S. Ruf, S. M. Gross, and S. Misra, “RFI detection and mitigation for microwave radiometry with an agile digital detector,” IEEE Trans. Geosci. Remote Sens, vol. 44, no. 3, pp. 694–706, Mar. 2006. [27] European Space Agency, SMOS Water Mission Winning Battle With Interference, 2010. [Online]. Available: http://www.esa.int/SPECIALS/ smos/SEMB5USOREG_0.html [28] A. Camps, J. Gourrion, J. M. Tarongi, A. Gutierrez, J. Barbosa, and R. Castro, “RFI analysis in SMOS imagery,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 2007–2010. [29] M. K. Hallikainen, J. Seppäanen, J. Hakkarainen, A. Rautiainen, and K. Aalto, “Studies of radio frequency interference at L-band using an airborne 2-D interferometric radiometer,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 2490–2491. [30] N. Skou, J. E. Balling, S. S. Sobjarg, and S. S. Kristensen, “Surveys and analysis of RFI in the SMOS context,” in Proc. IGARSS, Honolulu, HI, Jul. 25–30, 2010, pp. 2011–2014. [31] N. Skou, S. Misra, J. E. Balling, S. S. Kristensen, and S. S. Sobjaerg, “L-band RFI as experienced during airborne campaigns in preparation for SMOS,” IEEE Trans. Geosci. Remote Sens., vol. 48, no. 3, pp. 1398–1407, Mar. 2010. [32] E. Anterrieu, “On the detection and quantification of RFI in L1a signals provided by SMOS,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 10, pp. 3986–3992, Oct. 2011. [33] T. J. Jackson, M. H. Cosh, R. Bindlish, P. J. Starks, D. D. Bosch, M. Seyfried, D. C. Goodrich, M. S. Moran, and J. Du, “Validation of advanced microwave scanning radiometer soil moisture products,” IEEE Trans. Geosci. Remote Sens., vol. 48, no. 12, pp. 4256–4272, Dec. 2010. [34] C. Rüdiger, J. P. Walker, and Y. H. Kerr, “On the airborne spatial coverage requirement for microwave satellite validation,” IEEE Trans. Geosci. Remote Sens. Lett., vol. 8, no. 4, pp. 824–828, Jul. 2011. [35] P. de Rosnay, C. Gruhier, F. Timouk, F. Baup, E. Mougin, P. Hiernaux, L. Kergoat, and V. LeDantec, “Multi-scale soil moisture measurements at the Gourma meso-scale site in Mali,” J. Hydrol., vol. 375, no. 1/2, pp. 241–252, 2009. [36] L. Brocca, F. Melone, T. Moramarco, and R. Morbidelli, “Spatial– temporal variability of soil moisture and its estimation across scales,” Water Resour. Res., vol. 46, p. W02 516, Feb. 2010. doi:10.1029/ 2009WR008016. [37] D. G. Miralles, W. T. Crow, and M. H. Cosh, “Estimating spatial sampling errors in coarse-scale soil moisture estimates derived from pointscale observations,” J. Hydrometeorol., vol. 11, no. 6, pp. 1423–1429, Dec. 2010. [38] A. Loew and F. Schlenz, “A dynamic approach for evaluating coarse scale satellite soil moisture products,” Hydrol. Earth Syst. Sci., vol. 15, no. 1, pp. 75–90, Jan. 2011. doi:0.5194/hess-15-75-2011. [39] M. Drusch, E. F. Wood, and H. Gao, “Observation operators for the direct assimilation of TRMM microwave imager retrieves soil moisture,” Geophys. Res. Lett., vol. 32, p. L15 403, Aug. 2005. doi:10.1029/2005GL023623. [40] L. Brocca, F. Melone, T. Moramarco, W. Wagner, and S. Hasenauer, “ASCAT soil wetness index validation through in-situ and modeled soil moisture data in central Italy,” Remote Sens. Environ., vol. 114, no. 11, pp. 2745–2755, Nov. 2010. [41] W. Wagner, G. Lemoine, and H. Rott, “A method for estimating soil moisture from ERS scatterometer and soil data,” Remote Sens. Environ., vol. 70, no. 2, pp. 191–207, Nov. 1999. [42] D. Penna, H. J. Tromp-van Meerveld, A. Gobbi, M. Borga, and G. Dalla Fontana, “The influence of soil moisture on threshold runoff generation processes in an alpine headwater catchment,” Hydrol. Earth Syst. Sci., vol. 15, no. 3, pp. 689–702, Mar. 2011. [43] C. Albergel, C. Rüdiger, T. Pellarin, J. C. Calvet, N. Fritz, F. Froissard, D. Suquia, A. Petitpa, B. Piguet, and E. Martin, “From near-surface to root-zone soil moisture using an exponential filter: An assessment of the method based on in-situ observations and model simulations,” Hydrol. Earth Syst. Sci., vol. 12, pp. 1323–1337, 2008. [44] C. Albergel, J. C. Calvet, J. F. Mahfouf, C. Rüdiger, A. L. Barbu, S. Lafont, J. L. Roujean, J. P. Walker, M. Crapeau, and J. P. Wigneron, “Monitoring of water and carbon fluxes using a land data assimilation system: A case study for southwestern France,” Hydrol. Earth Syst. Sci., vol. 4, no. 6, pp. 1109–1124, Jun. 2010. [45] A. L. Barbu, J. C. Calvet, J. F. Mahfouf, C. Albergel, and S. Lafont, “Assimilation of soil wetness index and leaf area index into the ISBAA-gs land surface model: Grassland case study,” Biogeosciences, vol. 8, no. 7, pp. 1971–1986, Jul. 2011. [46] B. Bisselink, E. Van Meijgaard, A. J. Dolman, and R. A. M. De Jeu, “Initializing a regional climate model with satellite-derived soil moisture,” J. Geophys. Res., vol. 116, p. D02 121, Jan. 2011.
LACAVA et al.: FIRST ASSESSMENT OF THE SMOS SOIL MOISTURE PRODUCT
[47] A. Ceballos, K. Scipal, W. Wagner, and J. Martınez-Fernandez, “Validation of ERS scatterometer derived soil moisture data in the central part of the Duero Basin, Spain,” Hydrol. Process., vol. 19, no. 8, pp. 1549–1566, May 2005. [48] C. Albergel, C. Rüdiger, D. Carrer, J. C. Calvet, N. Fritz, V. Naeimi, Z. Bartalis, and S. Hasenauer, “An evaluation of ASCAT surface soil moisture products with in-situ observations in southwestern France,” Hydrol. Earth Syst. Sci., vol. 13, pp. 115–124, 2009. [49] P. Matgen, S. Heitz, S. Hasenauer, C. Hissler, L. Brocca, L. Hoffmann, W. Wagner, and H. H. G. Savenije, “On the potential of METOP ASCATderived soil wetness indices as a new aperture for hydrological monitoring and prediction: A field evaluation over Luxembourg,” Hydrol. Process., 2011, to be published. [50] European Space Agency, SMOS Level 2 Data Release to General Science Community, 2010. [Online]. Available: http://earth.esa.int/ object/index.cfm?fobjectid=7246 [51] CESBIO (Centre d’Etudes Spatiales de la BIOsphère), SMOS L2 Data is Reprocessed!, 2011. [Online]. Available: http://www.cesbio.ups-tlse.fr/ SMOS_blog/?p=1898 [52] European Space Agency, SMOS Anomalies—MIRAS Unavailability, 2011. [Online]. Available: http://earth.esa.int/object/index.cfm? fobjectid=7408 [53] A. Al Bitar, Y. Kerr, O. Merlin, P. Richaume, A. K. Sahoo, and E. F. Wood, “Validation of SMOS soil moisture over SCAN/SNOTEL network,” Geophys. Res. Abstracts, vol. 13, p. 1, 2011. doi:EGU2011-8700-1. [54] C. Rüdiger, J. P. Walker, Y. H. Kerr, O. Merlin, A. Mialon, and E. J. Kim, “SMOS L1c and L2 product validation with airborne and in-situ observations over south-eastern Australia,” Geophys. Res. Abstracts, vol. 13, p. 1, 2011. doi:EGU2011-5183. [55] M. Piles, A. Camps, M. Vall-llossera, I. Corbella, R. Panciera, C. Rüdiger, Y. H. Kerr, and J. Walker, “Downscaling SMOS-derived soil moisture using MODIS visible/infrared data,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 9, pp. 3156–3166, Sep. 2011.
Teodosio Lacava (M’12) was born in Potenza, Italy, in 1972. He received the M.S. degree in geological sciences and the Ph.D. degree in methods and technologies for environmental monitoring from the University of Basilicata, Potenza, in 1999 and 2004, respectively. Since 2005, he has been a Researcher with the Institute of Methodologies for Environmental Analysis (IMAA), National Research Council (CNR), Potenza. He has been a Coinvestigator of several European Union, Italian Space Agency (ASI), and CNR projects. He is the author of several peer-reviewed scientific journal articles and has served as a Referee for many international journals. His main interest is in the field of satellite data analysis for environmental research, particularly regarding the development and assessment of advanced satellite techniques for natural hazard investigation. Dr. Lacava is a member of the American Geophysical Union.
Patrick Matgen was born in Ettelbrück, Luxembourg, on July 13, 1978. He received the M.Sc. degree in environmental engineering from the Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, and the Ph.D. degree from the Water Resources Section, Delft University of Technology, Delft, The Netherlands, with his dissertation being on the retrieval of surface and subsurface water from microwave remote sensing observations and the integration of the data with flood prediction systems. He is currently the Project Leader with the Centre de Recherche Public–Gabriel Lippmann (Grand Duchy of Luxembourg), Belvaux, Luxembourg, where he is responsible for microwave remote sensing and hydrologic-hydraulic modeling projects. His expertise is documented in more than 40 peer-reviewed scientific journal articles.
1621
Luca Brocca was born in Genoa, Italy, in 1978. He received the M.S. degree in environmental engineering and the Ph.D. degree in civil engineering, both with excellence, from the University of Perugia, Perugia, Italy, in 2003 and 2008, respectively. Since 2009, he has been a Researcher with the Research Institute for Geo-Hydrological Protection (IRPI), National Research Council (CNR), Perugia. He is involved in teaching and tutorial activity for the University of Perugia, and he was a Codirector of Master’s thesis. He is the Italian National Correspondent of the Euromediterranean network of Experimental and Representative Basins. He serves as a Reviewer for more than 20 international journals and actively participated to several research projects in the frame of Italian and European programs and in collaboration with Italian and international institutions. He has been the author and the coauthor of 26 journal papers and more than 60 papers in conference proceedings. Dr. Brocca is a member of the Italian Hydrological Society, the Italian Group of Hydraulics (GII), the International Environmental Modelling and Software Society, and the European Geosciences Union. Marco Bittelli received the B.S. degree in agricultural sciences from the University of Bologna, Bologna, Italy, and the M.S. degree and the Ph.D. degree in soil physics from Washington State University, Pullman. He spent one year as a Postdoctoral Scientist in the Physics Department, Heidelberg University, Heidelberg, Germany. Since 2006, he has been a Researcher with the University of Bologna, where he also teaches soil physics, hydrological modeling, and scientific methods courses both at the undergraduate and graduate levels. He is the Coordinator of an International Master in Soil and Water Conservation, established between the University of Bologna, Washington State University, and the University of Florida, Gainesville. He is a Referee for the National Science Foundation (USA). He serves as a Reviewer for over 15 international journals and actively participated in several research projects in the frame of Italian and European programs and in collaboration with Italian and international institutions. He is the author and the coauthor of 22 journal papers, 3 book chapters, and over 40 papers in conference proceedings. Dr. Bittelli is Responsible of the Soil Physics Division, Italian Society of Soil Science. Nicola Pergola (M’11) received the degree in physics from the University of Rome “La Sapienza,” Rome, Italy, in 1993. Since 1998, he has been a Researcher with the Institute of Methodologies for Environmental Analysis (IMAA), National Research Council (CNR), Potenza, Italy, leading the “Geohazards” Unit, IMAA Satellite Remote Sensing Laboratory. He was a member of the Geohazard “Core Team” within the Integrated Global Observing Strategy project, contributing to define the new global observational strategies for geohazard mitigation. He was recently involved in several FP6 and FP7 projects. He was a Scientific Coordinator of different Italian Space Agency (ASI) and CNR projects and a Coinvestigator of several European Union and international (e.g., European Space Agency, North Atlantic Treaty Organization, and International Association for the promotion of cooperation with scientists from the independent states of the former Soviet Union (INTAS)) projects. He is an Associate Editor of the international scientific journal Geomatics, Natural Hazards and Risks (Taylor & Francis). He is the author of about 50 papers on Institute for Scientific Information journal and more than 150 publications and proceedings with peer reviewing and international distribution. He serves as a Referee for several relevant international journals. His main interest is in the field of the development of advanced satellite data analysis, particularly regarding high-temporal-resolution sensors, like the National Oceanic and Atmospheric Administration–Advanced Very High Resolution Radiometer, Earth Observing System–Moderate Resolution Imaging Spectroradiometer, and Meteosat Second Generation–Spinning Enhanced Visible and Infrared Imager, for environmental research and applications. In particular, his research activity is mainly focused on the development of robust satellite techniques for natural hazard investigation from space and on multitemporal satellite data analysis in the space–time domain, particularly using infrared satellite radiances. Dr. Pergola is a member of the European Geosciences Union.
1622
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 50, NO. 5, MAY 2012
Tommaso Moramarco received the M.S. degree in civil hydraulic engineering from the University of Bari, Bari, Italy, in 1989. He is currently a Researcher with the Research Institute for Geo-Hydrological Protection (IRPI), National Research Council (CNR), Perugia, Italy. Since 1989, he has conducted research in the field of hydrological processes addressed to flood forecasting and hydraulic risk mitigation. He has developed original works in these areas, producing a great number of original contributions in the leading hydrologic and hydraulic journals. In particular, his works include hydrometeorological monitoring, synthesis of the effects of spatial variability in hydraulic quantities and scaling, rainfall spatial distribution analysis, entropy theory applied to natural channels, modeling of overland flow over hillslopes, modeling of surface water over watersheds, flood forecasting, hydraulic risk, climate change, drought, safety of dams, artificial neural networks, and genetic algorithm. Since 2001, he has played a lead role in the hydrologic research at IRPI, CNR, and he has been a Coordinator and a Scientist responsible of the projects on hydrometeorological monitoring, flood forecasting, and hydraulic risk in the frame of Italian and European programs. He has advised authorities for establishing measures for flood protection also addressed to the forward planning. He is a Referee of leading hydrologic and hydraulic journals, an Associate Editor of Journal of Hydrologic Engineer [American Society of Civil Engineer (ASCE)], and a Guest Editor of the Hydrology and Earth System Science journal. Mr. Moramarco was the recipient of the Normal Medal from ASCE for his two works on flood routing in natural channels. Afterward, the American Academy of Water Resource Research has conferred on him the Diplomate, Water Resource Engineer.
Valerio Tramutoli was born on December 28, 1957. He received the M.S. degree in physics from the University of Rome “La Sapienza,” Rome, Italy. Since 1990, he has been with the Department of Engineering and Physics of the Environment (DIFA), University of Basilicata, Potenza, Italy, where he has been a Senior Researcher since 1993 and holding courses of satellite remote sensing in the Faculty of Sciences and Faculty of Engineering since 1997. Since 1991, he has been a Visiting Scientist in the main international centers involved in the Earth’s observation by satellite, taking part in several international projects and initiatives of the main space agencies like the European Space Agency, National Aeronautics and Space Administration, National Space Development Agency, and Italian Space Agency (ASI). He has been the National Coordinator of the SEISMASS Project (funded by ASI) Principal Investigator, or responsible of DIFA participation, to several international projects funded by the North Atlantic Treaty Organization and by European Community in the framework of the Science for Peace, FP6-IST, FP6-INTAS, FP6+FP7 Global Monitoring for Environment and Security (GMES) initiatives. Since 2010, he has been the Coordinator of the European project PRE-EARTHQUAKES funded by EC in the framework of the FP7-GMES-Space Program. His research activity has been focused on the development of new satellite sensors and techniques for natural, environmental, and technological hazard monitoring and mitigation. In this context, in 1998, he proposed the original Robust AVHRR Techniques (now, Robust Satellites Techniques) change detection approach successfully used in a large spectrum of applications. He has been among the few Scientists invited to participate, since 2001, to the Geohazard Core Team, Integrated Global Observing Strategy, instructed (by the most important space agencies, e.g., Committee on Earth Observation Satellites, Food and Agriculture Organization, United Nations Educational, Scientific and Cultural Organization, and International Council of Scientific Unions) to define the new observational strategies for geohazard mitigation for the next decade. He act as a Referee for the most important journals in the field and as a Project Evaluator for several funding agencies at European and extra-European levels. Since 2009, he has been a Member of the Editorial Board of the Geomatics, Natural Hazards and Risk international journal. He is the author or coauthor of more than 150 papers published in international journals, scientific books, and international conference proceedings. Mr. Tramutoli has been the Chair, a Cochair, an Organizer, or an Invited Speaker in the most important international conferences [European Geosciences Union, American Geophysical Union, and the International Union of Geodesy and Geophysics (IUGG)]. Since 2007, he has been a permanent member of IUGG Inter-Association Working Group on Electromagnetic Studies of Earthquakes and Volcanoes.
