Soil Moisture Content
The European Space Agency launched a programme called Climate Change Initiative (CCI). The CCI programme builds long-term data records based on Earth Observation for several Essential Climate Variables. In this context, the ESA CCI Soil Moisture project provides historical EO data adapted for its use by the climate community. Soil Moisture will be an ECV included in the Climate Data Store from the Copernicus Climate Change Service.
The latest CCI Soil Moisture data release has been used in the scope of SWICCA for the evaluation of the soil moisture indicator. The data used is a Soil Moisture CCI combined product, which merges microwave data from higher frequency radiometers and C-band scatterometers, and offers daily surface soil moisture with a spatial resolution of 0.25 degrees, from 1978 to 2014 (for further information, visit ESA CCI Soil moisture portal). For the purpose of this study, 30 years of data were selected from the CCI (from 1978 to 2009, keeping the 30 years period as close as SWICCA’s from 1971 to 2000) and averaged into the same 10-day window when counting the days from the first day of each year, to be compared with SWICCA Soil Water Content indicator.
Example: Soil Moisture for Emilia Romagna, Italy (Climate-proof Irrigation Strategies)
Thirty years of data were selected from the ESA CCI soil moisture active and passive merged data and averaged into the same 10-day window for comparison with SWICCA soil moisture indicator for the Emilia-Romagna use case.
Example: Soil Moisture for Júcar River, Spain (Extensive Drought Operations)
Thirty years of active and passive merged data were selected from the ESA CCI soil moisture dataset and averaged into the same 10-day window for comparison with SWICCA soil moisture indicator for the Júcar River.
Soil Moisture Content - Other Satellites
When working with different EO sources, data quality, and how the measured variable fits the reference indicator is an important aspect. Microwave brightness temperature provided by EO satellites is sensitive to soil moisture because water in soils has a large impact on the soil dielectric constant. The lower the microwave frequency, the higher is the relative sensitivity of brightness temperature to soil moisture. Additionally, lower frequencies show lower sensitivity to vegetation and other perturbing factors such as roughness and atmospheric disturbances. Therefore, from a remote-sensing perspective, L-band microwave radiometry is among the best ways to estimate soil moisture. However, the first space borne L-band radiometer, SMOS (Soil Moisture and Ocean Salinity), was launched in 2009, and SMAP (Soil Moisture Active Passive) in 2015. These lower frequency radiometers are not integrated in the ESA CCI at this point, thus we have shown SMOS seperately here, although new CCI releases are expected to integrate them in the future.
Example: Soil Moisture from SMOS for Júcar River, Spain (Extensive Drought Operations)
In order to show the strong potential of L-band radiometry for climate studies, and the possibility of obtaining consistent soil moisture records at increased resolutions using EO sources, a dataset providing SMOS data at an improved resolution has been generated by isardSAT over one of the SWICCA use cases: the Júcar River basin. The results shown are based on the disaggregation of SMOS data, originally at 40 km-resolution, and estimates the soil moisture variability within a low resolution pixel at the target 1 km resolution using optical data from MODIS. The method used for this downscaling is explained here.
Earth observations data from satellite are becoming a mature technology providing high resolution data at global scale. At this moment, EO water temperature data time span still does not cover 30 years of data but this comparison aims at providing a first approximation to compare upcoming data for Copernicus missions (e.g. Sentinels) with modeled data from SWICCA Copernicus Climate Change Service.
SWICCA portal provides water temperature change compared to SWICCA reference period (1971-2000). The water temperature changes are predicted through the E-HYPE model which provides temperature changes at catchment resolution (horizontal scale). In the case of lakes, it means that the horizontal resolution is equal to the size of the lake.
During lake stratification, the water temperature derived from E-HYPE model corresponds to the mean temperature within the epilimnion. For all other times, water temperature is the mean temperature within the whole lake.
Earth Observation data is derived from different instruments on-board of satellites. The first satellites measuring land surface temperature (including water temperature) were launched on 1994 and first measurements were available from 1995 (Figure 1).
Figure 1. List of satellites/instruments measuring Land Surface Temperature (1991-2015). Source: ESA GlobTemperature project.
The comparison from SWICCA and EO data could be split into two different time periods: a first period based on the SWICCA reference period (1971-2000) and a second period based on the SWICCA prediction for the first 30 years of this decade (2020 period). For both periods, EO data available never covers the minimum time span required by climate studies (30 years) so the results of the comparison would only show an initial approximation for interpreting whether the earth observations fall into the range of the SWICCA results.
Example: Water temperature for Lake Glan in Sweden
EO data available overlapping SWICCA reference period covers the period corresponding from January 1997 to December 2000. GlobTemperature data per each year has been plotted (see Figure 2) over SWICCA water temperature data for the reference period (January 1971 – December 2000).
Figure 2. Water Temperature for Glan Lake in Sweden (58.63, 15.93). In black there is the monthly average for the reference period of SWICCA’s indicator from 1971 to 2000 for E-HYPE and S-HYPE models. The grey intensity indicates the density of results with similar values in the SWICCA ensemble (the darker the grey, the more similar values). The rest of the colours represent the monthly average from EO data for years 1997 (purple), 1998 (red), 1999 (orange) and 2000 (blue).
The comparison shows that there is a discrepancy between EO data and SWICCA modelled data, with EO from the discrete years being lower than the ensemble range of the SWICCA 30 year period. SWICCA data represents estimates from dynamic modelling, while EO data represents estimates using satellites. The different approaches thus give different results. One reason could be that SWICCA water temperature modeled data represents for most of the cases (see exceptions below) the mean temperature within the whole lake of SWICCA. Contrarily, data derived from satellites represent the water surface temperature because satellite signals can scarcely penetrate water depths below water surface. In addition, SWICCA and EO data works at different horizontal resolution: SWICCA at catchment resolution and EO data available at 0.05 degrees. The spatial resolution of the data derived from satellites is a constraint because it combines land and surface temperature for the same pixel. In Figure 3, we can see this behavior for the example of Lake Glan. Lake Glan is observed for 11 pixels and each pixel measurement combines both land and water surface temperature which cannot be disaggregated to provide only water temperature.
Figure 3. EO data coverage for Glan Lake in Sweden. Each box represents a pixel of the EO data combining both land and water surface temperature.
Figure 4. Land/water surface temperature derived from EO data for each pixel (from Figure 3) and each year, covering Lake Glan.
Nowadays, satellite technology has evolved since precursor satellites launched in 1990s and there are several satellites measuring land and water surface temperature at higher resolution (below 1 km) able to disaggregate water and land pixels. Copernicus program from European Commission launched on February 2016 a new satellite called Sentinel-3 which carries an instrument called SLSTR aiming at measuring land and surface water temperature at higher resolution (500m). This satellite is already providing some water temperature data but these data does not cover a full year yet. An alternative solution is using water temperature retrieved from Landsat satellite series from NASA (Figure 5) which provides water temperature information at 30m resolution. This allows avoiding land contamination and providing water surface temperature exclusively.
Figure 5. Water Temperature for Lake Glan in Sweden. In pink, it is represented the monthly average from EO data for year 2016. The other colours represent the different RCPs in the SWICCA ensemble for the 2020's. Credit: EOMAP.
The method to retrieve water temperature has been developed by EOMAP Gmbh Company. In the case of Lake Glan, time series are built for coordinate Lat: 58.6359 and Lon: 15.9359 (same coordinate as SWICCA model representing this lake) within Lake Glan, averaging direct pixel and their surrounding pixels (3x3 pixel window).
NOTE: Winter measurements were not available due to the lake was frozen or there was too much haze or clouds. EOMAP workflows measure temperature over water bodies (even if there is ice) but ice results have given implausible results so they have been discarded from the analysis.