Soil Moisture for Hydrological Modelling and Water Cycle

The objective of the NASA Soil Moisture Active & Passive (SMAP) mission is to provide global measurements of soil moisture and freeze/thaw states. SMAP integrates L-band radar and radiometer instruments as a single observation system combining the respective strengths of active and passive remote sensing for enhanced soil moisture mapping.

Airborne instruments are a key part of the SMAP validation program. Here, we present an airborne campaign in the Rur catchment, Germany, in which the passive L-band system Polarimetric L-band Multi-beam Radiometer (PLMR2) and the active L-band system F-SAR of DLR were flown simultaneously on the same platform on six dates in 2013. The flights covered the full heterogeneity of the area under investigation, i.e. main land cover types and experimental monitoring sites with in situ sensors. Here, we used the obtained data sets as a test-bed for the analysis of three active-passive fusion techniques: A) The SMAP baseline algorithm: Disaggregation of passive microwave brightness temperature by active microwave backscatter and subsequent inversion to soil moisture, B), the SMAP alternative algorithm: Estimation of soil moisture by passive sensor data and subsequent disaggregation by active sensor backscatter and C) Copula-based combination of active and passive microwave data. For method C empirical Copulas were generated and theoretical Copulas fitted both on the level of the raw products brightness temperature and backscatter as well as two soil moisture products.

Results indicate that the regression parameters  for method A and B are dependent on the radar vegetation index (RVI). Similarly, for method C the best performance was gained by generating separate Copulas for individual land use classes.

Although precipitation products derived from satellite observations have reached a good level of maturity over the last decade, various studies underline that ongoing research and development are needed to address the accuracy and the resolution of these products. One potential strategy to improve precipitation products is the use of ancillary land measurements related to precipitation such as surface soil moisture (Crow et al., 2011; Pellarin et al., 2013, Brocca et al., 2014). Soil moisture measurements can be seen as the signature of the precipitation and intrinsically contain spatial and quantitative characteristics of the precipitation.

In this study, we carried out a detailed scientific analysis based on three different methodologies applied over 9 contrasted sites in Europe, Australia, Africa and USA containing high-quality hydrometeorological and soil moisture observations. Satellite soil moisture data from Soil Moisture and Ocean Salinity (SMOS) mission are employed for testing their potential in deriving a cumulated rainfall product at different temporal resolution from 3-hour to 1- and 5-day.

Three different methodologies, with an increasing order of complexity, are examined in the study. First, the SM2RAIN algorithm developed by Brocca et al. (2014) is based on the inversion of the terrestrial soil water balance equation and allows estimating rainfall from SMOS observations alone. Second, the Soil Moisture Analysis Rainfall Tool (SMART) developed by Crow et al. (2011) makes use of a a single-layer land surface model with a Kalman Filter to assimilate SMOS observations and utilizes analysis increments to derive errors in the forcing rainfall. Finally, the Land surface Model Assimilation Algorithm (LMAA), use a multi-layer land surface model with a Kalman Filter to derive errors in the forcing rainfall.

The applicability and accuracy of the three algorithms will be investigated as a function of climatic and soil/land use conditions. Indeed, if previous studies lead to promising results, a particular attention will be paid to assess expected limitations such as soil saturation, freezing/snow conditions, SMOS RFI, irrigated areas, contribution of surface runoff and evapotranspiration, vegetation coverage, temporal sampling, and the assimilation/modelling approach. The 9 selected sites gather such potential problems and will be shown and discussed at the conference.

Soil water is central to both physical and biogeochemical processes within the Earth System. Drying of soils leads to evapotranspiration (ET) becoming limited or “water-stressed” and is accompanied by rises in land surface temperature (LST), land surface-air temperature difference (delta T), and sensible heat flux. Climate models predict substantial changes to the global water cycle however there is a large spread between models in the time scale of ET decay during dry spells.

The e-stress project is developing novel satellite-derived datasets to assess the ability of Earth System Models (ESMs) to capture behaviour that is due to soil moisture controls on evapotranspiration and focuses on how drying soils affect fluxes of sensible and latent heat into the atmosphere.

Satellite records of LST now extend 15 years or more.  MODIS Terra LST is available from 2000 to the present and the Along-Track Scanning Radiometer (ATSR) LST record runs from 1995 to 2012. As part of the e-stress project these datasets have been used to calculate time series of delta T. This paper presents results from an investigation into the variability and trends in delta T during the MODIS Terra mission.

We use MODIS Terra LST with 2m air temperatures from reanalyses to calculate trends in delta T and “water-stressed” area. We make use of other satellite datasets in order to investigate the variability of delta T in relation to soil moisture (ESA CCI Passive Daily Soil Moisture), vegetation (MODIS Monthly Normalized Difference Vegetation Index) and precipitation (TRMM Multi-satellite Monthly Precipitation).  The paper also compares the temporal and spatial variability of delta T with GLEAM evaporation data.

Delta T is expected to show seasonally and spatially varying lagged correlations with antecedent rainfall. Earth System Models must be able to reproduce this ‘soil memory’. These correlations, derived using satellite data, will be discussed in the light of ESM assessment.

In conclusion there have been distinct signals in delta T during recent decades and these provide an independent assessment of hydrologically-forced changes in the land surface energy balance which can be used as a metric for the assessment of Earth System Model and global surface flux products.

Global warming is expected to change the global water cycle, leading to a change in storm tracks and an increase in the frequency and severity of extreme events like floods and droughts. The anticipated changes in water availability are expected to alter ecosystem dynamics and composition through multiple interacting pathways. This would in turn affect vegetation production and the efficiency of ecosystems to sequestrate atmospheric carbon dioxide, thus potentially influencing the pace of global warming. However, the impacts of changes in the global water cycle on ecosystems as predicted by models are uncertain as the link between soil moisture variability and vegetation is only poorly understood at the scale of the models.

This presentation gives an overview of the potential of the multi-satellite ESA CCI soil moisture product to assess a) the variability and trends in soil moisture worldwide, and b) the impact of observed soil moisture variability and change (e.g., trends, droughts) on ecosystem dynamics as characterised by various products based on microwave (e.g., vegetation optical depth) and optical remote sensing (e.g., fAPAR, fluorescence, CCI land cover). At the global level, trends in CCI soil moisture over the period 1988-2014 show a slight tendency towards drier conditions, but with large regional differences in sign and strength. Besides, the year-to-year variability in this period appears to be largely driven by the major climate oscillations, including El Niño Southern Oscillation. The connection between climate oscillations and soil moisture is also reflected by the variability in vegetation productivity data, in particular for semi-arid regions.    

Finally, we will explore how CCI soil moisture data can be directly integrated into a new model-data-integration approach to revise the parameterisation of soil moisture–vegetation interactions in the state-of-the art ecosystem model LPJml

Flood events are extremes situations that need accurate simulations and forecasting systems. The reliable estimation in time and space of the antecedent soil moisture condition is the critical boundary initial condition for flood simulation when using a distributed continuous hydrological model. In these models the soil moisture condition is updated through time from the solution of water and energy balance equations.  Recent developments have made these models more complex by inclusion of more processes and controlling variables, increasing parameter number and uncertainty of their estimates. In this context, remote sensing for  soil moisture retrievals could help to better initialize the hydrological model before flood events.

In this study we investigate: 1) the role of soil moisture initial conditions in the hydrological modeling of Alpine basin floods, 2) the effect and reliability of using satellite soil moisture data to initialize the hydrological model.

A distributed hydrological model, the FEST-EWB, that solves the system of energy and mass balance equations as a function of the representative equilibrium temperature (RET) was used. RET is comparable to the land surface temperature as retrieved from operational remote sensing data, so that a pixel to pixel calibration procedure of soil hydraulic and vegetation parameters for each pixel of the domain is proposed according to the comparison between observed (MODIS) and simulated land surface temperature.

Different remote sensing soil moisture data are used such as SMOS Level-3 products, ERA-interim reanalysis, a combined optical-passive microwave (OPM) approach. The combined OPM, based on SMOS,  was calibrated by using Leaf Area Index (LAI), the Enhanced Vegetation Index (EVI) and Normalized Difference Vegetation Index (NDVI) as a vegetation indicators.

These analyses are performed for the Upper Po river basin (Italy) closed at the river cross section of Ponte della Becca with a total catchment area of about 38000 km2, for different flood events between 2010 and 2014.