High-resolution soil moisture information has significant potential for improving hydrological modelling, Numerical Weather Prediction (NWP) and climate modelling efforts, and further allows for monitoring environmental threats such as droughts and floods. Passive microwave remote sensing of soil moisture only offers very coarse spatial resolution and is challenging at high latitudes because of high sub-pixel variability due to snow, ice, water bodies, rocks and strong topography being present in many coarse satellite pixels, thus complicating the retrieval process. A promising solution to this problem is mapping soil moisture using high-resolution synthetic aperture radar data (SAR) from Sentinel-1A&B.

Several approaches exist for soil moisture retrieval from microwave backscatter, e.g., physical models, semi-empirical models, empirical models and change detection approaches. Calibration of these retrieval methods tends to take place over a specific location and for land cover and soil texture characteristics different from those found at northern latitudes. To address these shortcomings, we are evaluating and testing different retrieval approaches over Norway in a new project funded by the European Space Agency and the Norwegian Space Centre. We do this by comparison of the methods among themselves, and by comparison of the retrieval datasets to independent data from in situ stations and the SURFEX land surface model.

By evaluating different retrieval methods against in situ stations and the land surface model, our main goal is to find the best possible retrieval methods using C-band SAR data from Sentinel-1A&B over northern latitudes. This could allow for the application of different retrieval methods to regions with different characteristics. In this contribution we will present initial results of the project and our experiences with mapping soil moisture over northern latitudes using Sentinel-1A&B backscatter data.

Area C - Board 246

Due to the substantial increase in the availability of remote sensing data, numerous studies have been carried out in recent years, in which the effectiveness and applicability of machine-learning algorithms in retrieving soil moisture (SM) from satellite images has been evaluated and compared. However, conflicting algorithm performances and model selection challenges have been frequently reported. In addition, only in very few studies, machine learners have been combined, to address their individual inefficiencies in order to optimize SM retrieval. In this paper, we propose an ensemble-based approach that couples random forest (RF), support vector regression (SVR) and artificial neural network (ANN) with the integration of environmental covariates. The aim of our study is to retrieve SM estimates from Sentinel-1 satellite images for the semi-arid landscapes of Ghana. Key model input datasets include, among others, VV and VH backscatter from Sentinel-1 images, and Sentinel-2-NDVI. Local incident angle, slope, surface roughness and aspect are considered as environmental covariates. In order to validate the retrieved SM estimates, we used in-situ SM, measured with a PR2 capacitance profile probe (0 — 10 cm depth) at 34 locations on key benchmark soils. These measurements were done simultaneously to the Sentinel-1 acquisitions for a period of 5 months (n = 340). The measurement locations were stratified with a hybrid methodology that coupled the global weighted principal component analysis and a cost-constraint conditioned Latin hypercube algorithm. Through this study, we intend to make two contributions. Firstly, time-series of in-situ SM and environmental covariates, used as proxies to represent SM variability and landscape heterogeneities, are related to Sentinel-1 backscatter values to validate the SM estimates. Secondly, in the ensemble-based approach we use the individual machine learners’ optimized outputs as inputs. Thus, we hypothesize that the ensemble-based method can improve the overall accuracy of SM retrieval as compared to individual machine learning algorithms. The obtained nearly real-time SM information can guide agricultural management systems that depend on large-scale spatio-temporal SM information.

Area C - Board 247

Knowledge about soil moisture over agricultural fields, especially with high spatial and temporal resolution, is important for several applications ranging from hydrology to precision agriculture. Microwave remote sensing technologies have proven the ability to derive soil moisture at high spatial resolution over large areas. To derive soil moisture from microwave data, different approaches have been proposed in the literature, ranging from empirical (Water Cloud Model) over semi-empirical (Oh et al., Dubois et al.) to more physical based (Integral Equation Method) radiative transfer (rt) models. The availability of consistent time series of Sentinel-1 SAR data enables the opportunity to use rt-models within in a data assimilation scheme for continuous monitoring of soil moisture on a high resolution basis. The available temporal resolution of time series from Sentinel-1 A/B over parts of Europe is around one to two days. This high temporal coverage however, results in data sets with different imaging geometries. The geometry can differ in case of the azimuth angle (ascending or descending) as well as in the case of the incidence angle. For the latter one, most of the rt-model consider the incidence angle whereas the azimuth angle is addressed in only a few. Another question that arise is how good the rt-model can account for different incidence angle or if the SAR data should be preprocessed to the same incidence angle before used in a soil moisture retrieval scheme.

The presented study aims to assess the effect of not considering the azimuth angle within available rt-models on soil moisture retrieval. Therefore we use dense Sentinel-1 time series acquired over the Munich-North-Isar site at different incidence and azimuth angles. We show the dependency of the model performance on variations in both angles and will discuss the need for more complex models or how to treat above mentioned effects in a data assimilation framework. The effect of the lack of process description in the rt-models on soil moisture retrieval will be discussed as well.

Area C - Board 248

Soil moisture is an important component in investigations of land-surface climate and hydrology. Soil moisture is recognized by the Global Climate Observing System of fundamental importance in the study of climate and environment. The soil moisture content is linked to evaporation and the heat fluxes from land to the atmosphere and are in some regions strongly coupled with precipitation. It is also an important variable for diagnostic of surface conditions in view of hydrologic and agricultural processes as the surface runoff, infiltration or drought development.

In this work we investigate the relation between the phase of interferometric SAR images and the time varying soil moisture. The main advantage of SAR interferometry with respect to the use of radar cross-section is that the information about soil moisture can be recovered using a reduced number of in-situ measurements. In particular, we combine three interferograms obtained from three SAR images, of the same area acquired at different times, to derive maps of bi-coherence and phase triplet. This last quantity allows to disentangle the phase contribution due to soil moisture from those related to microwave propagation in atmosphere and terrain displacements. Hence, a scattering model is used to estimate the time series of soil moisture from the sequence of phase triplet images. Results are compared to those obtained using the interferometric phase and coherence to emphasize the importance to split the effects due to propagation (e.g. atmosphere) from those related to volume scattering. The study area is located in Portugal at a farmland near the Tagus River estuary, south of the Sorraia River mouth. It is composed of large rectangular plots of clay soil with flat topography. Five sensors were deployed, and soil moisture was measured (at a depth of 5 cm) from 1 September 2017 until April 2018 in an hourly basis. For the same location and period, two sets of interferograms were computed from 18 consecutive C-band Sentinel-1 SAR images, using both VV and VH polarimetric channels. The first set was computed combining all two consecutive SAR images with 6 days of revisiting time, the second one was computed combing all two consecutive SAR images with 12 days of revisiting time. The results show that there is a linear correlation between observed and modelled interferometric phase. In this experiment observed phase showed a more sensitivity to soil moisture change when compared with modelled phase. The correlation between observed and modelled coherences was not evident as was with the phase. However, it was found a negative correlation between coherence and soil moisture change. The most promising result comes from the clear linear correlation found between modelled and observed phase triplets although a saturation effect was found that hinder its use in case of very dry soils. A comparison between fully polarimetric L-band images and interferometric C-band results is carried out to investigate advantages and drawbacks in using Sentinel-1 interferograms to estimate soil moisture values. Furthermore, the use of a time series of Sentinel-2 images to distinguish the different fields based on their use, and to identify targets which can be ued as points with no temporal changes in soil moisture is also investigated.
ACKNOWLEDGMENTS
Fundação para a Ciência e Tecnologia (FCT) – project UID/GEO/50019/2013 and SFRH/BPD/96069/2013.

Area C - Board 249

Soil moisture is an important parameter for agricultural crop monitoring. It presents a key parameter of irrigation management and water stress. In the first part of this work, the C-band Sentinel-1 radar data are analyzed and physically interpreted with the support of experimental data for winter wheat crops. The data are collected over two fields of winter wheat with drip irrigation located in the Haouz plain in the center of Morocco. Sentinel-2 images are also used to extract the normalized difference vegetation index (NDVI). Two polarizations (VV and VH) and three angles of incidence (34.5°,35.2° and 45.6°) are investigated in this study. Furthermore, two different informations extracted from Sentinel-1 products are used; the backscattering coefficient and the interferometric coherence. The results of this analysis showed that the interferometric coherence is totally sensitive to the development of wheat, while the backscatter coefficient is widely linked to changes in soil moisture content to the full development of wheat when the overlap of many contributions makes the signal more complicated to interpret. In order to better understand the backscattering mechanism of the wheat, the second part of this work is dedicated to modeling different contributions from the wheat field to the total backscatter. For this purpose, the Water Cloud Model is coupled with Oh model. The coupled model (CWCM) is calibrated and validated for both polarizations VV and VH. The comparison between the modeled backscattering coefficients by the coupled model and those extracted from Sentinel-1, yielded a root mean square error (RMSE) values range between 1.31 dB and 1.88 dB for VV and between 2.23 dB and 2.58 dB for VH. In the third part of this study, soil moisture content is retrieved using only radar data without the need for any ancillary data. First, the wheat water content is retrieved from the interferometric coherence and then used as an input of the coupled CWCM. It is found that the inversion approach is able to retrieve the soil moisture for VV, with a correlation coefficient R, RMSE and bias of 0.65, 0.07 m3/m3 and 0.004 m3/m3 ,respectively. For the VH polarization, the obtained R, RMSE and bias are about 0.53, 0.07 m3/m3 and 0.002 m3/m3. Finally, it can be concluded, that the proposed approach, which is applicable from the swing until the harvest using only the Sentinel-1 data, yielded very promising results in retrieving the soil moisture over the winter wheat crop in semi-arid regions.

Area C - Board 251

Soil moisture plays an important role in maintaining life on earth by modulating many physical, chemical and biological processes. Global Climate Observing System recognized soil moisture as an essential climatic variable for preventing extreme events, such as floods, droughts and heat waves, as well as for the management of water resources. Through exchanges with the atmosphere via evapotranspiration, knowledge of soil moisture is also a central component in weather forecast models via general circulation models. Additionally, its continuous monitoring on a large scale and over a long period of time provides insight into possible climate changes. The scope of this study is to carry out an integrated project for estimating soil moisture of an area by combining in situ and remote sensing data in Wallonia, Belgium. In addition, it describes the design, implementation, calibration of a soil moisture network for environmental monitoring. This work investigates the performance of time-domain reflectometry (TDR) probes and capacitance sensors for the regional mapping of soil moisture. The soil moisture measurement will be assessed in different test sites located in agricultural fields and grasslands. Then, in situ data will be combined with remote sensing data for estimating soil moisture at a regional scale. While remote sensing can provide an effective methodology for mapping surface moisture content at a regional scale, it has to be calibrated with in situ quantitative measurements of soil moisture. We will investigates the capabilities Sentinel-2 and Sentinel-1 data to put forward a first interpretation of the local data at the Walloon scale taking into account the soil properties, the geological and the geomorphological characteristics of the territory. This study will draw recommendations regarding (1) the performance of both types of soil moisture sensors, (2) the implementation of a soil moisture network in Wallonia and, (3) the methodology to combine in situ and remote sensing data. Due to climate and rainfall pattern changes it is critical to monitor soil moisture at a regional scale, in this context this network will provide a help to face events like drought and provide support to agriculture.

Area C - Board 252

1. INTRODUCTION
Soil moisture is a key indicator for the management of a resource as precious/scarce as water. Sectors such as agriculture, environmental monitoring, and territory management, need to know the soil hydrological stress to take the appropriate actions. This work describes the second airborne optical and GNSS-R campaign conducted in the framework of the HUMIT project to test the readiness of data fusion techniques [1] to retrieve high resolution soil moisture. The results of the HUMIT support the feasibility of a 6U CubeSat optical and GNSS-R mission for soil moisture monitoring, the ³Cat-3/MOTS. The flight was carried out on October 22nd 2018 near the town of Balaguer (41.79°, 0.80°), over an agricultural area of approximately 4 km².

2. INSTRUMENTATION DESCRIPTION
The plane carried 4 different instruments: a hyperspectral sensor (AisaEAGLE-II, discontinued and not available online; replaced by the AisaFENIX [2]), a thermal sensor (TASI [3]), a reflectometer (CORTO [4]), and an L-band radiometer (ARIEL [5]). In order to retrieve the direct GNSS signal the plane carried several GPS antennas for navigational purposes. One more GPS antenna was added to the plane to retrieve the reflected signal, a Left Hand Circular Polarization (LHCP) down looking antenna at L1/L2 bands. The AISA instrument is a pushbroom hyperspectral sensor (254 bands) in the VNIR (Visible and Near InfraRed), from the 406 nm to the 993 nm. The TASI (Thermal Airborne Spectrographic Imager) instrument is a hyperspectral sensor (32 bands) in the TIR (Thermal InfraRed), between the 8050 nm and the 11450 nm.

3. CAMPAIGN OBJECTIVES
The main scope of the experiment is to acquire optical data with the AISA and the TASI instruments to perform a data fusion with the soil moisture data retrieved from the CORTO instrument. The optical resolution at the flight height (1200 m) is under 1 m, which will increase the spatial resolution of the soil moisture measurement (approximately 10 m). The L-band radiometer will be used, later on, to perform an instrumental performance inter-comparison once all data has been processed. It will also be evaluated the perforance of the data fussion algorithms using Sentinel 2 data instead of high resolution VNIR data (AISA sensor). The closest Setinel 2 pass is from the day after the flight (October 23rd, 2018).

4. VALIDATION AND FIELD CAMPAIGN
Simultaneously with the flight it was perform a fiel campaig to collect data for validation purposes. Soil samples, Theta Probe soil moisture measurements and soil roughness profiles were collected. Initial ground truth analysis shows soil moisture ranging between 0.15 and 0.33 m³/m³.

5. PRELIMINARY RESULTS
The processing of the data retrieved is under process and will be presented in the LPS19 to its full extent. The multispectral and thermal images undergo a 3-step correction process: radiometric, geometric and atmospheric. In the same way, the reflectivity obtained from the CORTO instrument undergoes antenna pattern compensation and geolocation. In this process it is used topography information from the ICGC database (2 m grid DEM). And finally, soil roughness data will be used in the transformation from reflectivity to soil moisture.

[1] 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 transactions on Geoscience and Remote Sensing, Vol. 49, no. 9, September 2011.
[2] SPECIM AisaFENIX sensor. Available (online): http://www.specim.fi/products/aisafenix-hyperspectral-sensor/, [last visited November 11th, 2018].
[3] iTRES TASI sesor. Available (online): http://www.itres.com/tasi-600/ [last visited November 11th, 2018].
[4] T. Martini, “UAV-based GNSS-R systems for soil moisture monitoring”, Final Degree Project, tutor: A. Alonso, advisors: A. Camps, and L. Schenato, Padova, October 6th, 2015.
[5] Balamis ARIEL sensor. Available (online): http://www.balamis.com/ariel/, [last visited November 11th, 2018].

Area C - Board 253

Human society increasingly depends on information derived from Earth Observation (EO) data. In particular, there is a growing demand for information facilitating agriculture, forestry, soil moisture and hydrology. To this purpose the number of space-borne satellites is projected to increase dramatically over the next years, in the form of international scientific missions from various agencies but also small cube-sat constellations. The general trend currently is to create individual land surface or soil moisture products for each satellite mission. In that regard, a variety of individual products is created for each single-sensor approach and for similar land surface parameters such as soil moisture. However, such single-sensor approaches do not take into account potential synergies between different sensors which might enhance the retrieval.
Within the H2020 project MULTIPLY (www.multiply-h2020.eu), a data assimilation platform has been developed for the consistent retrieval of bio and geophysical parameters from various space-borne sensors over the optical and microwave spectral domain. We combine, for example, data from SAR observations with optical remote sensing data, compatible radiative transfer models and auxiliary data from various information sources (observational, prior, multi-temporal) to optimally retrieve the observed satellite information. Finally we use the optimized state vector for the synergistic retrieval of land surface parameters such as soil moisture at high resolution.
In this study, we assess the retrieval of soil moisture by using the MULTIPLY platform over agricultural landscapes and discuss the benefits of such a system. It will be shown that the retrieval of soil moisture from remote sensing data greatly benefits from a joint retrieval from optical and microwave data. In particular the retrieval of soil moisture is greatly improved by incorporating LAI retrievals from optical sensors accounting for vegetation characteristics.
Results are validated against field measurements over the Munich-North-Isar site, Germany showing a good performance of the developed retrieval scheme.

Area C - Board 254

A large number of agricultural, crop monitoring and water management applications require sub-kilometric frequent surface Soil Moisture (SM) observations. L-band acquisitions are especially suited for SM retrieval since they are less susceptible to attenuation by vegetation and are less sensible to surface roughness of the soils than other parts of the microwave spectrum such as C or X – bands.
However, with the currently available missions (SMOS/SMAP) the native spatial resolution is of ~40 km. Recently, a number of different approaches based on the Sentinel-1 C-band synthetic aperture radar (SAR) have been developed, also the pair of Sentinel-1 satellites enable for SM retrieval on a scale of 1 km using change detection-like algorithms. Nonetheless, the C-band radar backscatter is sensitive to a number of factors apart from soil moisture like the vegetation opacity and the surface roughness.
In this study, we used an Artificial Neural Network (ANN) algorithm for retrieving the SM at 1 km resolution. For the algorithm development and training of the ANN we used dense in-situ SM networks, sites with multiple measurements over the area covering a significant portion of the L-band radiometer footprint. Notably we used the data from networks: HOBE, OZNet, Little Washita, Reynolds Creek, Little River, Walnut Gulch, Rehmedus, and Beerambadi in South-India.
The core concept of the algorithm consists of downscaling the coarse scale SMOS L3 brightness temperatures (TB) rather than the Soil Moisture itself. The downscaled is done using the Sentinel-1 backscatter which is calibrated, filtered, orthorectified and then averaged over a 1 km grid. The data from three years: 2015, 2016, 2017 from the networks was divided randomly in 4-1 proportion, the larger one served to test the architecture of the ANN and to train it. The remaining 20% and the data from International Soil Moisture Network served for point-by-point validation of the global retrievals.

Area C - Board 255

In recent years, the possibility of estimating Soil Moisture (SM) at different resolution scales improved greatly with the launch of the latest satellite microwave sensors and in particular of the Soil Moisture Active and Passive (SMAP) radar + radiometer and Sentinel-1 (S-1) Synthetic Aperture Radar (SAR). However, the tradeoff between temporal and spatial resolution offered by each of these two sensors is still unable to meet the requirements of many users. SMAP is indeed able to observe the most part of the earth surface almost daily, but it is limited by the coarse spatial resolution, which hampers the detection of hydrological patterns in small catchments. On the other hand, Sentinel-1 is capable to acquire data at high resolution but with less frequent revisiting, thus being unable to follow the quick variations of SM that can help, for instance, in predicting extreme events.
The possibility of merging SMAP and S-1 data through Artificial Neural Networks (ANN) for obtaining an improved SM product has been exploited in this study. An ANN application integrating SMAP and S-1 to obtain a SM product at 10 Km resolution and enhanced accuracy with respect to the SMAP radiometric one has already been proposed in [1]. In this study, we focused on the use of ANN to improve the spatial and temporal resolution of the SM obtained by merging data from both sensors.
The proposed algorithm has been implemented and tested on an area in central Italy, roughly corresponding to the Umbria region. The area is characterized by a hilly landscape, including agricultural areas and forests, with heterogeneities that set important constraints to the potential of SMAP observations for SM monitoring. All the SMAP and S-1 acquisitions available between 2014 and 2017 on the area have been taken into account for implementing, training, and validating the algorithm.
Due to the unavailability of distributed SM in-situ measurements, the reference SM values for implementing and training the algorithm have been derived from the Soil Water Balance Model (SWBM, [2]), a well-assessed hydrological model that was able to simulate accurately the observed SM in the study area.
After training, the algorithm has been validated using in situ SM data from five stations in Umbria obtained from the International Soil Moisture Network (ISMN, [3]).
The obtained results highlight the potential of the SM generated by the algorithm for operational hydrological applications at regional scale, particularly in heterogeneous environments as the Italian territory.
REFERENCES
[1] E. Santi, S. Paloscia, S. Pettinato, L. Brocca, L. Ciabatta, D. Entekhabi, 2018. On the synergy of SMAP, AMSR2 AND SENTINEL-1 for retrieving soil moisture. International Journal of Applied Earth Observation and Geoinformation, 65, pp. 114-123, ISSN 0303-2434, https://doi.org/10.1016/j.jag.2017.10.010.
[2] L. Brocca, S. Camici, F. Melone, T. Moramarco, J. Martinez-Fernandez, J.-F. Didon-Lescot, R. Morbidelli (2014), “Improving the representation of soil moisture by using a semi-analytical infiltration model”, Hydrological Processes, 28(4), pp. 2103-2115, doi:10.1002/hyp.9766.
[3] Dorigo, W.A.; Wagner, W.; Hohensinn, R.; Hahn, S.; Paulik, C.; Xaver, A.; Gruber, A.; Drusch, M.; Mecklenburg, S.; van Oevelen, P.; et al. The International Soil Moisture Network: A data hosting facility for global in situ soil moisture measurements. Hydrol. Earth Syst. Sci. 2011, 15, 1675–1698.

Area C - Board 256

Soil texture heterogeneity is known to be one of the main sources of soil moisture spatial variability. With the recent development of high resolution maps of basic soil properties such as soil texture and bulk density, relevant information to estimate soil moisture variability within a satellite product grid cell is available. We use this information for the prediction of the sub-grid soil moisture variability for each SMOS, SMAP, and ASCAT grid cell. The approach is based on a method that predicts the soil moisture standard deviation as a function of the mean soil moisture based on soil texture information. It is a closed-form expression using stochastic analysis of 1D unsaturated gravitational flow in an infinitely long vertical profile based on the Mualem-van Genuchten model and first-order Taylor expansions. We provide a look-up table that indicates the soil moisture standard deviation for any given soil moisture mean, available at https://doi.org/10.1594/PANGAEA.878889. The resulting data set helps identify adequate regions to validate coarse scale soil moisture products by providing a measure of representativeness of small-scale measurements for the coarse grid cell.

Area C - Board 257

Active and passive sensors are used to retrieve geophysical parameters over agricultural areas. The active SAR sensors have a high spatial resolution and in some cases like Sentinel-1 high temporal repetitivity. The backscatter from S1 and Alos Palsar are used here to investigate the relation with vegetation characteristics. The passive sensors are characterised by a low resolution and high temporal revisit. They are also less impacted by vegetation structure and surface roughness. In this paper we use a rich dataset of +50 monitored plots in the Berambadi and Maddure watersheds in Karnataka south-india. The plots are characterized with a small surface and multiple yearly crop cycles (up to three). Surface roughness, LAI and surface soil moisture were measured at each plots for each overpass of S1 and Alos. The remote sensing data from S1 and Alos is processed for calibration, ortho-rectification, Lee speckle filtering and extracted over each plot. The Passive data is extracted over the watershed area and filtered for Radio Frequency Interference. The results show the differentiated impact of vegetation between C-Band and L-Band. The vegetation cycle is clearly identified with the S1 dataset and to a lower extent from the L-Band SAR ALOS. The low spatial resolution data from SMOS and AMSR-2 show seasonal and yearly dynamics of soil moisture and vegetation dynamics respectively that can be a reference information for a temporal disaggregation approach. Base on the MAPSM algorithm (Tomer et al. 2016) a synergistic approach for the use of SMOS and S1 data is applied for the retrieval of surface soil moisture. The product is compared to the monitored soil moisture and existing datasets (SMAP+S1). The presented results are important for the understanding of of upcoming mission like the futur L-Band SAR mission from Copernicus and NISAR (NASA, ISRO). It is also important for the continuity of C- and L-Band passive missions.

Area C - Board 258

Soil moisture is an essential climate variable which plays a crucial role linking the Earth’s water, energy and carbon cycles between the land and the atmosphere. Although it represents a small percentage of all the water on Earth, it provides key information about evaporation, transpiration, infiltration and runoff. As a result, accurate knowledge of soil moisture is crucial for both global and local applications.
Currently there are two operating missions, SMOS (Soil Moisture and Ocean Salinity) and SMAP (Soil Moisture Active and Passive), launched by ESA and NASA, which are specifically devoted to measure the (surface) soil moisture. Their passive microwave instruments operate at L-band, which is favorable for measuring soil moisture due to the high sensitivity of the microwave signal to the soil dielectric constant at this frequency. The Copernicus Imaging Microwave Radiometer (CIMR) high priority candidate mission is planned to have an L-band radiometer on-board and could provide continuation of the observational data stream.
Accurate high-resolution soil moisture maps are needed to fulfill a growing number of application tasks e.g., prevention of wildfires [1][2], early-detection of forest decline [3], monitoring the evolution of insect pests [4]. To downscale soil moisture from the coarse-scale radiometer resolution to a finer one, downscaling algorithms typically use additional observations at fine scales. A wide variety of downscaling algorithms exists [5]. They may use observations from different sensors, use different ancillary data or rely on different physical assumptions. The performance of these algorithms depends on the physics they are based on, and on the information at fine scale that they use, which may also depend on season, climate and land cover. This makes a direct comparison difficult, since their performance is intrinsically time and region dependent, and the algorithm set-up will therefore reflect into the results.
The baseline downscaling approach for the SMAP mission uses active and passive microwaves. SMAP was designed to have a radar and a radiometer on-board, but the radar stopped operating on July 7, 2015. An alternative active-passive soil moisture product is now available merging SMAP and Sentinel-1 data [6].
Recently a spatially-consistent downscaling algorithm has been developed to obtain high-resolution soil moisture maps at 1 km from the ~30-40 km native resolution of current passive microwave instruments using optical data. This algorithm is based on the semi-empirical downscaling approach implemented at the Barcelona Expert Center (BEC) to blend SMOS and MODIS data into 1 km soil moisture estimates [7]. The novelty is that it introduces the concept of a shape adaptive window, improving its spatial consistency and thereby allowing for its global implementation [8].
The objective of this study is to provide insight into the physical mechanisms that can influence the production (downscaling) of high-resolution soil moisture maps with a special focus on soil moisture anomalies. We will investigate soil moisture maps produced by the BEC (microwave-optical) and the SMAP-Sentinel (active-passive microwave) algorithms presented above and understand the characteristics of the downscaled soil moisture maps with a focus on spatial and temporal anomalies.

[1] D. Chaparro, M. Piles, M. Vall-Llossera, A. Camps, “Surface moisture and temperature trends anticipate drought conditions linked to wildfire activity in the Iberian Peninsula,” European Journal of Remote Sensing, vol. 49, issue 1, pp. 955-971, 2016.
[2] D. Chaparro, M. Vall-llossera, M. Piles, A. Camps, C. Rüdiger and R. Riera-Tatché, "Predicting the Extent of Wildfires Using Remotely Sensed Soil Moisture and Temperature Trends," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 9, no. 6, pp. 2818-2829, June 2016.
[3] D. Chaparro, J. Vayreda, M. Vall-llossera, M. Banqué, M. Piles, A. Camps, J. Martínez-Vilalta, “The role of climatic anomalies and soil moisture in the decline of drought-prone forests,”. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 10, issue 2, pp. 503-514, 2017.
[4] M. J. Escorihuela, O. Merlin, V. Stefan, G. Moyano, O. A. Eweys, M. Zribi, ... & S. Ghaout. “SMOS based high resolution soil moisture estimates for Desert locust preventive.
[5] J. Peng, A. Loew, O. Merlin, N. Verhoest, “A review of spatial downscaling of satellite remotely sensed soil moisture,” Reviews of Geophysics, 10.1002/2016RG000543, March 2017.
[6] N. Das, R.S. Dunbar, “Level 2 SMAP/Sentinel Active/Passive Soil Moisture Product Specification Document,” Jet Propulsion Laboratory, D-56548, August 2017.
[7] M. Piles, N. Sánchez, M. Vall-llossera, A. Camps, J. Martínez-Fernández, J. Martínez, V. González-Gambau, “A Downscaling Approach for SMOS Land Observations: Evaluation of High-Resolution Soil Moisture Maps Over the Iberian Peninsula,” IEEE Journal of Sel. Topics in Applied Earth Obs. and Remote Sens., vol. 7, no. 9, pp. 3845-3857, 2014.
[8] G. Portal et al., "A Spatially Consistent Downscaling Approach for SMOS Using an Adaptive Moving Window," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 11, no. 6, pp. 1883-1894, June 2018.

Area C - Board 259

Soil moisture is a key environmental variable governing the partition of energy and mass fluxes between land and atmosphere, and is characterized by high variability in space and time. In the last decades, microwave remote sensing has been successfully used to estimate soil moisture globally. Microwave radiometers and scatterometers typically observe the same location with a daily frequency, thus providing adequate temporal coverage. However, most of the soil moisture products currently available are not suitable for regional hydrological and agricultural applications because of the coarse spatial resolution (~ 10s-100s of km2). In fact, there is significant spatial variation in soil moisture at much finer scales (~ 1-100 m), mainly controlled by topography, soil texture, and vegetation.

In this presentation, we propose a method to downscale coarse scale soil moisture from ASCAT by using static ancillary data, namely soil texture and topographical indices, and a temporally dynamic vegetation variable, fAPAR. The analysis is carried out within a small agricultural catchment located in Lower Austria, characterized by complex topography and different vegetation types (crops, grassland, and forest). The coarse scale soil moisture input is assumed to account for factors governing large-scale patterns (i.e. climate). We tested various downscaling models which differ in terms of complexity (e.g. multiple linear regression and Random Forest) and input variables. The resulting soil moisture maps, produced at 50 m resolution,were evaluated against a high-density in-situ network.

This preliminary study shows the potential of downscaling coarse resolution soil moisture products by including ancillary data (both static and dynamic), all of them available at high spatial resolution.Random Forest provided the most accurate results when using soil, topography, and vegetation in synergy. In particular, adding a proxy for vegetation dynamics improved the correlation of approx. 10% compared to static-only variables, and fAPAR was the single most important feature for the Random Forest model. In conclusion, we show that it is possible to improve the level of detail of the spatial distribution of water in the soil, which is necessary for applications like agricultural monitoring, crop yield forecast, and the prediction of landslides and erosion.

Area C - Board 260

Observations from radar sensors on-board satellites, mostly SAR and scatterometers, can be used to monitor changes in land cover, vegetation density, surface soil moisture (SSM) or flood extent. Recently, backscattering coefficients (σ0) from radar altimetry demonstrated a strong potential to monitor land surface properties (e.g., floods, SSM, …).
In a previous study, σ0 acquired over West Africa (between 0° and 25°N and 5°W - 25°E) at S, C, Ku and Ka bands using nadir-looking altimeters were analyzed over the major bioclimatic zones, soil and vegetation types encountered in this region. SSM was first retrieved using ENVISAT RA-2 backscattering coefficients over semi-arid Sahelian savannahs in the Gourma region of Mali. Contrary to what was observed using other radar data, satellite altimetry backscattering coefficients seem to be less affected by the presence of vegetation. High correlations between SSM and σ0 acquired at Ka-band by SARAL were also observed in West Africa showing the potential of radar altimetry for estimating SSM. Similar observations were also made over several semi-arid sites in Australia. The impact of soil moisture on radar altimetry waveforms at Ku and Ka bands was also estimated through an electro-magnetic modeling study, showing an increase of the power received by the altimeter sensor as SSM increases.
In this study, an inversion approach of SSM from radar altimetry backscattering is proposed calibrated with SSM from ORCHIDEE land surface model (LSM). It is based on the linear relationship observed between SSM and σ0 over sandy areas. A linear fit is estimated between ORCHIDEE SSM and σ0 averaged at grid-scale (0.5°). It is applied to Low Resolution Mode (LRM) high frequency data (several hundred of m) along the tracks of ENVISAT (2002-2010), Jason-2 (2008-2016) and SARAL (2013-2016) on their nominal orbit over Sahel.
Evaluation of the SSM retrievals was performed through comparisons against in situ SSM measurements from several sites in West Africa (mostly in Mali, Niger and Senegal AMMA-CATCH sites), and remote sensing SSM products such as ESA CCI-SM, SMOS-IC, ASCAT and model based estimation like those provided by GLEAM. Preliminary results showed a very good agreement between our SSM estimates and in-situ measurements from Gourma AMMA-CATCH site in Mali.

Area C - Board 261

Quality assurance of satellite-derived soil moisture datasets is key for ensuring users are provided with datasets of sufficient quality to their applications. Currently, such activities are undertaken on an ad-hoc basis or using a protocol designed for a specific product. These results are useful, but differ in terms of output validation data, plots and reports and therefore are difficult to inter-compare.

The aim of the Quality Assurance for Soil Moisture (QA4SM) (www.qa4sm.eodc.eu) project (supported by the Austrian Space Application Programme) is to bring together methodologies and protocols used for the validation and quality control of soil moisture data products and develop this into a coherent methodology allowing the provision of standardised, and above all traceable validation results to the user community. QA4SM is driven by validation best practice guidelines resulting from a collaborative effort of the entire satellite soil moisture validation community, including the Committee on Earth Observation Satellites, the Global Climate Observing System, and the validation teams of the individual satellite missions.

In addition to this standardisation, an easy-to-use functional, web-based test application has been made available to the user community. The application currently allows users to compare medium resolution satellite datasets to in-situ validation data (made available through the International Soil Moisture Network (ISMN) (http://ismn.geo.tuwien.ac.at)) and global land surface models. The application is being enhanced to allow comparison of high resolution soil moisture products to ISMN data. Later updates of the system will allow the upload and comparison of user’s own datasets.

The application is accessible through a web-based graphical front-end which allows users to easily select specific criteria for the comparison process including control of the masking applied to each dataset, the metrics calculated and so on. The validation results are provided as a netCDF file which includes traceability data in the global attributes as well as plots of each metric. Future iterations will also automatically generate a validation report skeleton for the user.

The goal of this presentation is to present the QA4SM system, show initial validation results and collect feedback from the wider satellite soil moisture validation/calibration community.

Area C - Board 262

Monitoring soil moisture conditions at high temporal and spatial resolutions is of major interest for several applications, from agriculture to flood prevention. Microwave remote sensing has been successfully used to estimate soil moisture globally at daily frequency, however, the spatial resolution (~ tens of km) is often too coarse. New satellite missions (planned or already launched) will provide soil moisture at very high spatial resolution in the near future. A framework for the validation of such products is therefore necessary, in order to evaluate the accuracy of these estimates.

Various independent locally and regionally operating in situ networks exist but they do not follow standardized techniques or protocols. Data is collected in different units, at different depths and at various sampling rates. In addition, quality control protocols are rarely applied.

Founded in 2009, the International Soil Moisture Network (ISMN) is a unique service, acting as a centralized data hosting facility for in situ soil moisture observations. Soil moisture measurements are collected from all around the globe, harmonized, in terms of units and sampling rate, advanced quality controls applied and made freely available to users via a web portal (http://ismn.geo.tuwien.ac.at/).

Currently, the ISMN consists of almost 2500 stations spread all over the globe, thus providing a unique tool for evaluation of satellite-derived soil moisture. In fact, the ISMN provides benchmark data for several services like ESA CCI soil moisture, Copernicus C3S, Copernicus Global Land Service and for online validation services such as QA4SM.

Moreover, the ISMN represents a rich data source for validating emerging fine spatial resolution products such as the Sentinel-1 Surface Soil Moisture (at one kilometer resolution) or the SCATSAR-Soil Water Index (one kilometer daily soil water index), which both are based on near-real-time Sentinel-1 SAR observations and currently in preparation towards operational product dissemination in the Copernicus Global Land Service.

In this session we give an overview of the ISMN, its unique features and the benefits of using its data for validating high resolution products (e.g. from Sentinel-1).

Area C - Board 263