Soil moisture controls several important Earth system processes, and has hence many scientific and practical applications in a diverse number of fields such as climatology, hydrology, agronomy or ecology. An increasing number of spaceborne active and passive microwave sensors provide measurements of the moisture content in the soil. All these microwave sensors are mounted on polar-orbiting satellite platforms, which means that they map soil moisture at best twice a day at their ascending and descending passes respectively. The best temporal coverage is achieved by the three Advanced Scatterometer (ASCAT) instruments on board of the three concurrently operated METOP satellites, which together cover practically the whole Earth every day in both their ascending and descending passes. However, the spatial resolution of the ASCAT measurements is rather coarse, from about 15 km to 50 km depending on the level or pre-processing. From all available Synthetic Aperture Radar (SAR) satellite missions, the best temporal sampling of the Earth’s land surface areas is achieved by the Sentinel-1 satellite constellation, varying from less than three days over Europe to twelve days over the least-prioritized regions. These spatio-temporal characteristics of existing microwave sensors mean that they are incapable of capturing many important Earth system processes such as the formation of convective precipitation and flash floods. This limitation can be overcome by the G-CLASS (Geosynchronous – Continential Land-Atmosphere Sensing System) mission, which is one of the three Earth Explorer 10 mission candidates.
G-CLASS proposes a polarimetric C-band radar operating from geosynchronous orbit to provide continuous monitoring over Europe and Africa. It will be able to acquire backscatter measurements at 20-50 m resolution every few hours, and lower resolution imagery at the kilometre scale within minutes (but note that there is an important operational trade-off between geographical coverage and spatial and temporal resolutions). The observations will fill the current measurement gaps, and have a real impact on our ability to use satellite radar to improve our understanding of fine-scale, within-catchment variability of surface moisture. For G-CLASS, the goal will be to retrieve soil moisture either at 1 km scale every 15 min or at 300 m every 3 hours with an accuracy comparable or exceeding state-of-the-art microwave soil moisture data sets. In this contribution, we will discuss the suitability of different soil moisture retrieval approaches for reaching these goals based upon the experiences of applying change detection, radiative transfer modelling and machine learning approaches to ASCAT and Sentinel-1 backscatter data. Overall, we expect that existing algorithms will be able to deliver good quality of the retrievals over low to moderately vegetation regions, while scientific advances will be needed to deal with known issues affecting active C-band soil moisture retrievals over Europe and Africa such as sub-surface scattering, structural changes in vegetation and soil, and diurnal effects. The much improved temporal sampling available from G-CLASS is expected to improve our capability to use the retrieved soil moisture data for estimating higher-level data products such as profile soil moisture content or rainfall, and to ease their assimilation in various types of Earth system process models (land surface models, hydrologic models, etc.). Together these advances will improve services for agriculture and water resource management and help society respond to threats such as floods and landslides.

Drought arguably the climate and weather phenomenon that most strongly impacts societies worldwide, causing severe socioeconomic and ecologic damage. While models unanimously project an overall increase in aridity and drought occurrence in the future, observational evidence has so far been inconclusive. Discrepancies between the various drought definitions and existing drought indices have been one of the factors contributing to the current low confidence in global-scale observed trends. In this study we investigate global trends in meteorological dry spells using a simple and intuitive diagnostic: the maximum annual number of consecutive dry days (CDDs). In contrast to popular drought indices, the number of CDDs is direct measure of rainfall scarcity, is easy to quantify based on rain data, free of parametrizations, and independent from other proxies.

Previous analyses of trends in CDDs are based on in situ rain gauges. While they may cover relatively long periods, they still suffer from spatial and temporal discontinuity, changes in accuracy in time, and various other quality-control issues. Because the time-span of available precipitation data records is constantly increasing, current satellite-based products are becoming an alternative to obtain spatially and temporally consistent global trends in CDDs. In particular, the Tropical Rainfall Measuring Mission (TRMM) has now been operational for over twenty years, thus offering a unique opportunity to analyse temporally-coherent single-platform precipitation data. Here, we use TRMM3B42 3-hourly data for 1998–2018 – gap-filled with CMORPH-v1.0 – to calculate and analyse the maximum annual number of CDDs worldwide.

Preliminary results show that 30-40% of the continental land monitored by TRMM experiences significant positive trends in the maximum annual number of CDDs. In depth analysis reveals that about 60% of this signal lies in regions that are affected by monsoons, and originates at most during the spring dry season. Further analyses relate the trends in CDDs to the intensity of monsoonal circulation and rainfall, and the role of land–atmospheric feedbacks. Our findings emphasize the vulnerability of monsoon regions, where agriculture is mostly rainfed and population relies on rainfall availability and variability. An increasing length of dry spells as we progress into the future might lead to devastating socioeconomic and ecologic consequences in these regions.

Accurate knowledge of soil moisture is recognized as essential for hydrological studies and flood risk assessment. The benefit of integrating remotely sensed Soil Moisture (SM) in rainfall-runoff modelling is potentially very high but previous studies have shown controversial results in terms of the ability of surface observations of soil moisture to improve runoff predictions. Among others, main problems that have been highlighted are related to 1) the existence of multiplicative bias components between satellite retrievals and models which may produce incorrect variance error estimation and thus provide non-optimal assimilation results (Kornelsen et al. 2015); 2) the optimality criterion inherent in all least squares-based approaches (and therefore in all Kalman filtering-based techniques) which penalizes the prediction of flood extremes (Saifuddin et al. 2017).
In this study, the effects of remotely sensed SM assimilation on hydrological model performances were extensively explored as a first goal, involving a large number of catchments (+700) located across Europe for almost a decade. Satellite observations derived from the active and the passive European Space Agency Climate Change Initiative (ESA CCI) products were ingested into MISDc-2L model (Brocca et al. 2011, Massari et al., 2018) by using the Ensemble Kalman Filter (EnFK) technique. Rainfall and temperature data were collected from the European Climate Assessment & Dataset E-OBS (https://www.ecad.eu/), while discharge data were obtained from the Global Runoff Data Centre (GRDC, https://www.bafg.de/GRDC).
The role of the rescaling approach adopted during the pre-processing of SM observations before the assimilation was analysed, considering its effect in terms of multiplicative bias between rescaled satellite-based and reference model SM estimates (i.e. with differences between datasets showing a systematic relationship with the estimate values). In this respect, two different rescaling methods, i.e. CDF-matching (CDF) and Triple Collocation Analysis (TCA), were considered, that imply the matching of total variance and only its signal component, respectively, between reference and rescaled datasets. The evidences of conditional bias in runoff predictions due to remotely sensed SM assimilation were checked, with simulated discharges that result in larger and smaller values being under- and overestimated, respectively.
Preliminary results showed that CDF led to a larger conditional bias than TCA, which means that CDF is less optimal for predicting flood extremes as it results in a much larger underestimation of flood peaks.

Agricultural drought is one of the most critical hazards with regard to intensity, severity, frequency, spatial extension and impact on livelihoods compared to other natural hazards. In the last decades, the occurrence of drought episodes is increasing, with direct and indirect impacts on economy, environment and livelihoods. Agricultural drought involves a deficit in plant-available water that could compromise the crop yields. A vast suite of approaches has been developed to monitor and characterize agricultural drought, as based on either climatic ground-based data, soil moisture data or a variety of remote-sensing drought proxies. We recently proposed the Soil Moisture Agricultural Drought Index (SMADI), a simple and intuitive index that allows determining agricultural drought events based on the physical conditions of the soil-vegetation ensemble. It is scalable over space and time and integrates remote sensing datasets on land surface temperature (LST), vegetation indices (e.g., the NDVI) and surface soil moisture (SSM). Previous experiments with SMADI at local, continental and global scales showed a reasonable match with other agricultural drought indices and with registered events of drought.
In this work, we further investigate the physical link of optical and microwave data using statistical learning methods for anomalous change detection. The objective is to automatically capture possible non-linear relationships that may be missed by the index and also include microwave vegetation optical depth (VOD) as an additional feature in the input space. In the proposed model, SMOS-based SSM and VOD are used to characterize the water content in soils and vegetation, and MODIS-derived LST and NDVI are used as descriptors for the surface temperature and the vegetation chlorophyll content, respectively. Based on the experiments and learned relations, we will provide a model solely based on Earth Observation (EO) data to characterize the physical mechanisms driving drought-prone conditions in agro-ecosystems. A selection of drought episodes registered in the Emergency Events Database (EM-DAT) for years 2010 to 2015 is used as a benchmark for the model validation. The proposed model will be compared to SMADI and the suitability of both approaches for the early detection of drought in current disaster management systems will be assessed. In this regard, the Copernicus Imaging Microwave Radiometer (CIMR) high priority candidate mission can provide continuity of L-band observations beyond SMOS, and optical LST and NDVI data could be obtained from ESA Sentinels.

High-frequency, high-resolution quantitative rainfall estimates are key inputs to the accurate assessments of heavy rain associated with extreme weather and storms, and potential for flash-flooding. However, for a comprehensive, nation-wide rainfall analysis, there is no single source of rainfall data that possesses the desirable characteristics of coverage, resolution, accuracy and data latency for real-time applications. And while data accuracy and timeliness (i.e. low latency) are characteristics of utmost importance for real-time applications, the ability to reliably capture the spatial and temporal patterns of rainfall across a range of scales may be just as important for accurate catchment streamflow modelling and flood prediction.

The goal of producing (near) real-time high-resolution accurate estimates of hourly rainfall on a national scale is only likely to be achieved through the statistical blending of two or more rainfall data sources which may include rain gauges, radar, satellite, and numerical weather prediction model output. Moreover, a real-time blended analysis is likely to gain wider use if it is accompanied by a spatially explicit representation of estimation accuracy, most usefully as a rainfall ensemble. In addition to providing a quantitative assessment of analysis uncertainty, an ensemble provides a means of computing probabilistic rainfall information, including of exceedence probability.

This paper describes a method to generating gridded rainfall ensembles to quantify uncertainty in a national-scale real-time blended radar-satellite-NWP rainfall analysis for Australia. Currently the blended rainfall analyses represent 60-minute accumulations of rainfall at 2-km resolution nationally and generated at 30-minute intervals throughout the day. Data used include the satellite rainfall retrievals from the GPM IMERG near real-time products, rain rates derived from Himawari-8, forecast rainfall accumulations from the Bureau of Meteorology’s ACCESS-R system, and a national mosaic of gauge-corrected radar rainfall. Several methods for blending these multiple rainfall sources of (near) real-time rainfall data were assessed, including the approaches underpinning MSWEP (Beck et al., 2016), STEPS (Seed et al., 2013) and a scale-dependent blending method (Renzullo et al., 2017).

Gridded rainfall ensembles were generated through the perturbation of the blended rainfall analyses with spatially structured noise. Structure is derived from the rainfall analyses themselves, thus ensuring that each ensemble member possesses the scaling properties of the blended analysis. Furthermore, we paid close attention to the error variances of the perturbations to ensure that the ensemble spread satisfies reliability criteria. Finally, we describe our approach of extending the ensemble generation from local to national scales by concatenating series of spatially contiguous local analyses, and provide a detailed evaluation of the ensembles by assessing the statistical reliability of the rainfall information products using independent rain gauge measurements.

The Mediterranean Sea is a unique meteorological environment and a weather forecasting challenge. Over this area severe weather events of different nature often originate and develop to hit coastal regions, causing major damages and casualties.
Conventional ground-based instruments (e.g., raingauges and weather radar networks) are often inadequate to monitor events during their offshore development. On the other hand, in spite of their irregular time sampling, Low Earth Orbit (LEO) satellite systems carrying passive and active microwave (MW) sensors can be fully exploited for the analysis, characterization, and monitoring of heavy precipitation systems over this area.
In this respect, the NASA/JAXA Global Precipitation Measurement (GPM) Core Observatory (CO), equipped with the most advanced microwave radiometer (GPM Microwave Imager, GMI) and the first spaceborne Dual-frequency Precipitation Radar (DPR), allows for the 3-D analysis of precipitation structures. Moreover the GPM constellation of MW radiometers ensures 1-3 hourly global coverage of precipitation.
In this presentation, we show the GPM mission role in the characterization and monitoring of precipitation associated to extreme events, that hit the Italian peninsula in the last years.
We first test the capabilities of advanced satellite precipitation products (GPM and H SAF) for quantification and monitoring of precipitation associated with a very intense and localized single cell that caused a flash flood in Livorno, on the coast of Tuscany, Italy, in the night between 9 and 10 Sept. 2017. In this case, the satellite precipitation products based on the GPM constellation of MW radiometers offer a wide spectrum of results, showing that better performances come from algorithms tailored for the specific region, while algorithms designed for global application are less sensitive to such small-scale thunderstorms.
Then, GPM-CO measurements, together with those from other satellite-based and ground-based sensors, are used for a detailed observational analysis of the impressive hailstorm that hit the Gulf and the city of Naples (Italy) on September 05, 2015. This analysis provides observational evidences of extremely rare features of such severe hailstorm and demonstrates the GPM-CO potentials in providing unique spaceborne measurements of 3-D structure of precipitation.
Finally, an analogous study is carried out for Numa, a Mediterranean tropical-like cyclone occurred on 16-18 November 2017 over southern Italy. GPM-CO measurements are used, in conjunction with LINET (LIghtning NETwork) data, to show how the rain-band structure evolves as the storm develops into its mature phase. High-resolution simulations, carried out with the Regional Atmospheric Modeling System (RAMS) to analyze the key mechanisms leading to the formation and evolution of Numa, show the impact of the data assimilation of radar DPR reflectivity on the precipitation forecast.
This study shows that in order to fully exploit satellite data and products for studying and monitoring severe weather in the Mediterranean area, and, ultimately to improve forecasting (and nowcasting) capabilities in this complex region, efforts should be undertaken to provide products tailored for these areas (working on algorithm calibration), and provide the error structure of the products (working on validation). This would enable users to properly apply the most suitable product for the specific need, including hydrological applications, and data assimilation of precipitation-related fields over the sea where these systems often initiate. Long data records of satellites precipitation-related measurements and reliable products are becoming available and will contribute in the assessment of climate change signatures in the Mediterranean area, where such severe events are becoming more and more frequent.