Atmosphere/Ionosphere

In this work, we present a statistical assessment of the assimilation of Precipitable Water Vapor (PWV) measurements into a Numerical Weather Model (NWM). The aim is to quantify in how far the assimilation of PWV map time series provided by SAR interferometry and GNSS can improve WRF modeling of atmosphere physics. Results are compared with PWV estimates provided by GNSS. The study area is located within the Upper Rhine Graben (URG) in France and Germany and extends from 48.7°N to 49.6°N in latitude and from 7.4°E to 8.9°E in longitude. This area is characterized by the river Rhine flowing within an about 35 km large valley. The valley is surrounded by forested mountainous regions, for example, the Black Forest in the East. This affects the weather that is mainly cold and dry in winter, but active and highly variable in summer. A network of eight permanent GNSS receivers has been installed and used to estimate the neutrospheric phase delay, mainly related to the atmospheric water vapor. The GNSS-derived PWV has been estimated by processing observations from all visible GNSS satellites with elevations higher than the cutoff elevation angle of 7°. Each of these estimates represents an average value over a conical section of the neutrosphere with a radius depending on the cutoff elevation angle. Furthermore, these values values represent a temporal mean since each PWV estimate is obtained using GNSS observations over a time window of one hour. A time series of 17 SAR images, acquired by ENVISAT ASAR in descending mode from December 15, 2003 to December 8, 2008 with a minimum 35 days temporal resolution, was acquired over an area of 100 km×100 km, centered on 49°11’N, 8°2’E. The times series of SAR images was processed using Persistent Scatterer Interferometry (PSI). As the URG region is characterized by small surface deformations with velocities below 0.5 mm per year, the neutrospheric phase can easily be distinguished from the surface displacement phase.The Weather Research and Forecasting (WRF) model has been run to simulate 3D fields of the atmospheric parameters, i.e., temperature, pressure, relative humidity, and geopotential at the acquisition times of the SAR images. The initial conditions and the boundary conditions have been adjusted by using the ERA-Interim analyses with 6-h temporal and 0.75° 0.75° horizontal resolutions. The above parameters have been used to estimate 3D fields of wet refractivity. Starting from this wet refractivity information, PWV maps have been computed using a ray-tracing technique.The PWV maps estimated from the WRF outputs have been compared with measurements provided by PSI and GNSS. The PWV maps estimated from WRF outputs after the assimilation of absolute PWV maps have been compared with GNSS measurements. It has been observed that after the assimilation of absolute PWV  maps the WRF-based PWV estimates are closer to GNSS measurements and follow the same trend in time for a few hours depending on the location of the GNSS station. Furthermore, it has been observed that the assimilation of InSAR maps of PWV provides better results than the assimilation of GNSS PWV measurements only.

Recently SAR interferometry has been used as a means to derive Precipitable Water Vapor (PWV) maps characterised by a high spatial resolution if compared to the currently available PWV measurements (e.g. GNSS and radiometers). There have been also the first attempts to assimilate those InSAR-derived maps of PWV into a Numerical Weather Model. However, the issue of how effectively use InSAR has not yet been tackled. An important point such as the high spatial density of InSAR measurement of PWV needs to a deeper understanding of the assimilation of InSAR measurements. Can InSAR maps of PWV affect the parameterizations results (e.g. the convective one) of Numerical Weather Models? Is the high spatial resolution of InSAR maps really needed? In this study, we show how the assimilation of precipitable water vapour (PWV) data in Numerical Weather Models (NWMs) can change the system thermodynamics to improve the accuracy of numerical forecast on local heavy rainfalls. Recent 3D-Variational data assimilation (3DVAR) experiments showed that the PWV values estimated from interferometric data provided by the Envisat-ASAR sensor with high spatial resolution can play an important role in improving the correct amount and spatial distribution of moisture in the atmosphere [1, 2]. With the Sentinel-1 A/B C-band sensors its possible generate maps of PWV over large areas (Figure 1) with a length of hundreds of kilometers and a width of about 250 km (country-spanning areas), a spatial resolution of 5×20 m and an absolute revisiting time of 6 days or fewer when combined with other sensors, opening new perspectives to the application of SAR meteorology concept [3]. We used the Weather Research and Forecast Data Assimilation (WRFDA) model, at micro-scale resolutions (1 km), over the Iberian Peninsula (focusing on the southern region of Spain) and during a convective cell associated with a local heavy rainfall event, to study the impact of assimilation PWV maps obtained from SAR interferometric phase calculated using images acquired by the Sentinel-1 satellite. It's worth noting that, in this case, the model without assimilation PWV maps fails to reproduce the amount and the region of heavy rainfall (Figure 2). The assimilation of interferometric PWV maps with high spatial variability by the WRF model, promoted alterations in the buoyancy force over the study area and consequently increased the atmospheric instability, were new convection cells were generated over the correct area. We assessed the results using in-situ meteorological data and with a meteorological radar. The availability of interferometric PWV maps on a routine basis can help to capture the high variability of the water vapour distribution at micro-scales. In this study, we show that the knowledge of the PWV with high spatial resolution can change the system thermodynamics to improve the NWP accuracy.

[1] E. Pichelli et al., “InSAR water vapor data assimilation into mesoscale model MM5: Technique and pilot study,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 8, no. 8, pp. 3859–3875, Aug. 2015.

[2] P. Mateus, R. Tomé, G. Nico, and J. Catalão, “Three-Dimensional Variational Assimilation of InSAR PWV Using the WRFDA Model,” IEEE Transactions on Geoscience and Remote Sensing, vol. 54, no. 12, pp. 7323–7330, 2016.

[3] P. Mateus, J. Catalão, and G. Nico, “Sentinel-1 interferometric SAR mapping of Precipitable Water Vapor over a country-spanning area”, IEEE Transactions on Geoscience and Remote Sensing (submitted). 

The influence of signal delay due to the varying atmospheric refractivity can be significant in individual interferograms. This signal is generally considered to be noise in deformation studies, but it can also potentially be used to improve weather models [1]. This application has an enormous potential, because, contrary to deformation studies, every acquired SAR image contains valuable information on the state of the atmosphere.

Until recently, revisit times of SAR satellites were too low to operationalize this application of SAR data. With the launch of the Sentinel-1 satellites, the theoretical revisit time reduced to less than 2 days for mid-latitude regions, which potentially enables operational implementation in weather models. The availability of other current and future SAR satellites further strengthens this opportunity. Even though practical operational applications will be strongly dependent on data latency (downlink, throughput, processing, and dissemination), we aim to develop and demonstrate the business-case.

Here we analyze and demonstrate the assimilation of InSAR-derived atmospheric measurements to numerical weather models. We perform a quantitative analysis of the differential integrated refractivity (DIR) values, as observed by InSAR and a Limited Area Model (LAM). LAMs have become popular for short-range numerical weather prediction on km-scale [2], while medium-range models such as the ECMWF have global coverage and a resolution of about 0.25 degrees. Here we use the HARMONIE LAM, which is developed by the HIRLAM and ALADIN consortia [3] and currently used as the operational weather model for the Netherlands.

As the DIR-fields are highly correlated to the vertically integrated water vapor over the full air column, which is a difficult to measure but important variable for a well-performing weather model, the contribution of InSAR proves to be very significant.

We compare the DIR values and their uncertainties as observed by InSAR with the DIR values, derived from weather model predictions. The results show that there are many similarities on wide scales, but that the precision of the integrated refractivity values from InSAR is much higher. We analyze the scales of both datasets using spectral energy distributions, which show that (i) SAR data holds much more detailed information on scales below 10 kilometers, and (ii) the dynamic range of the DIR observations seems to be underestimated by the weather model. The InSAR data are then assimilated in the numerical weather prediction to find the best compromise (or ‘analysis’) between the model simulation (‘first-guess’) and the observations [4].

[1] Hanssen, R. F., Weckwerth, T. M., Zebker, H. A., & Klees, R. (1999). High-resolution water vapor mapping from interferometric radar measurements.Science, 283(5406), 1297-1299. 

[2] Degrauwe, D., Caluwaerts, S., Voitus, F., Hamdi, R., & Termonia, P. (2012). Application of Boyd's periodization and relaxation method in a spectral atmospheric limited-area model. Part II: Accuracy analysis and detailed study of the operational impact. Monthly Weather Review, 140(10), 3149-3162.

[3] Navascués, B., Calvo, J., Morales, G., Santos, C., Callado, A., Cansado, A., & García-Colombo, O. (2013). Long-term verification of HIRLAM and ECMWF forecasts over southern europe: History and perspectives of numerical weather prediction at AEMET. Atmospheric Research, 125, 20-33.

[4] Marseille, G. J., Barkmeijer, J., de Haan, S., & Verkley, W. (2016). Assessment and tuning of data assimilation systems using passive observations. Quarterly Journal of the Royal Meteorological Society, 142(701), 3001-3014.