Data rescue of national and international meteorological observations at Deutscher Wetterdienst

Historic observational data records are an important contribution for climate reconstructions and analysis of past weather events. Particularly in remote and data sparse regions, such as the open ocean, newly rescued data can significantly improve the knowledge about weather and climatic conditions in earlier decades and centuries. Deutscher Wetterdienst (DWD) holds several collections of original historical weather records from land stations and ships. They comprise not only observations from Germany, but also of the global oceans and land stations in many parts of the world. All German state-owned meteorological observations beginning with the Prussian Meteorological Institute in 1848 are collected in the main archive of DWD in Offenbach. DWD’s branch office in Hamburg holds the marine archive starting with the collections of the German Naval Observatory, ’Deutsche Seewarte’, which existed from 1874 to 1945. It includes marine data records from ships, as well as land stations in many parts of the world (e.g. from former German colonies) and signal stations situated at the coasts of the North and Baltic Sea. The documentation, digitisation and quality check of the enormous quantity of handwritten journals of all four data archives is still ongoing. The digitised data will be freely accessible to all interested scientists and are also continuously submitted to international data archives, such as ICOADS and ISPD. Through these data sets, the data are also an important input for regional and global reanalyses. The presentation will give an overview of the historical archives of Deutscher Wetterdienst and will show the recent progress of the digitization efforts and ongoing analysis of the data.

Modeling impact of climate warming on cotton growth and phenology in Pakistan from 1961 to 2010 based on provincial data Assessment of temperature extremes based on departures from long-term reanalysis and high-resolution ensemble forecasts over Indian region A kriging method for a gridded quantitative precipitation estimate over Alaska with uncertainty bounds

Abstract
Ocean reanalyses are becoming increasingly important to reconstruct and provide an overview of the ocean state from the past to the present-day. In the scope of the Copernicus Marine Environment Monitoring Service (CMEMS), the Black Sea reanalysis (BS-REA) is produced by using an advanced variational data assimilation method to combine the best available observations with a state-of-the-art ocean general circulation model. The hydrodynamical model is based on Nucleus for European Modeling of the Ocean (NEMO), implemented for the Black Sea (BS) domain with horizontal resolution of 1/27• x 1/36•, and 31 vertical levels. NEMO is forced by ECMWF ERA5 atmospheric reanalysis and climatological precipitation. The model SST is relaxed to daily objective analysis fields from CMEMS SST TAC. The model is online coupled to OceanVar, a 3D-Var ocean data assimilation scheme, to assimilate sea level anomaly (SLA) along-track observations from CMEMS SL TAC and available in situ vertical profiles of temperature and salinity from both SeaDataNet and CMEMS INS TAC products. Temperature fields present a continuous warming in the layer between 25-150 m, within which the BS Cold Intermediate Layer resides. SST shows a basin-wide positive bias. The root mean square difference (RMSD) can reach 0.75 ºC along the Turkish coast in summer. SLA has the largest RMSD close to the shelf due to the high mesoscale activity along the Rim current. The system has produced very accurate estimates which makes it suitable for understanding the BS physical state in the last decades. Nevertheless, in order to improve the quality of the BS-REA, new developments in ocean modelling and data assimilation are still important, and sustaining the BS ocean observing system is crucial.

Abstract
This study evaluates the latest Japanese Reanalysis for Three Quarters of a Century (JRA-3Q) conducted by the Japan Meteorological Agency (JMA), focusing on a semi-period of pre-satellite era of 1960s and 1970s. JRA-3Q, which is based on the JMA's operational system with 6-hourly 4D-Var data assimilation as of December 2018, is the third Japanese global atmospheric reanalysis spanning late 1940s onwards, using an atmospheric model with a reduced horizontal resolution of TL479 and 100 vertical layers up to 0.01 hPa. Because only few global-covered observational datasets during the pre-satellite era are available, the JRA-3Q is mainly evaluated in reanalysis intercomparison and about temporal consistency and spatial homogeneities. Emphasis of this evaluation during the non-satellite era is placed on the representation of tropical circulation, the time consistency of reanalysed fields between the pre-satellite and satellite eras, and the quality of the stratospheric ozone and water vapor. The surface circulation over the tropical Africa is improved by reducing spurious anticyclonic circulation anomalies found in JRA-55. Stratospheric ozone is also improved by incorporating adequate ozone depletion substances, sea-surface temperature as well as the development of the ozone model. The quasi-biennial oscillation is not as good as that in JRA-55 with a shorter period of around one year in the middle stratosphere and diminished amplitude in the lower stratosphere.

Abstract
Reconstructions of climatological features in the tropical areas of South America may be proven challenging due to the scarcity of reliable, long-term data records. To address this problem, retrospective analyses -aka (re)analyses -are regularly used to provide spatially continuous, long-term time series of several atmospheric and land-surface variables in studies of South American climate. Overall, reanalysis comprehends forecast models and data assimilation systems. Data assimilation in a global reanalysis comprises computationally expensive techniques to generate initial conditions used in the embedded forecasting systems. Global reanalysis products also offer a wide range of opportunities to monitoring atmospheric conditions at regional scales, for instance, providing initial and boundary conditions to regional downscaling. The downscaling of a global reanalysis in the tropical-tosubtropical South America by a regional numerical model may as well be enhanced by means of empirical methods, such as spectral (dynamic) nudging and satellite-based precipitation assimilation, both employed in the present study. Examples of the combined application of the two methodologies in regional downscaling will be illustrated through the reconstruction of extreme events that occurred in the continental South America depicted by the interannual variability of South American monsoon precipitation, and the severe weather episodes near the Brazilian coastline.

Abstract
The stratospheric Brewer-Dobson circulation (BDC) is a key element in the climate system as it controls variations in ozone and other trace gases which impact the radiation budget. We investigate the BDC in the new ERA5 meteorological reanalysis and compare with results from its predecessor ERA-Interim, based on residual circulation dignostics and on simulations of stratospheric age of air with the transport model CLaMS. Our results show a substantial uncertainty in the representation of the BDC in reanalyses regarding both the climatology and trends. In particular, the BDC is significantly slower in ERA5 than in ERA-Interim, manifesting in weaker tropical upwelling, diabatic heating rates and larger age of air, mainly related to weaker subtropical gravity wave drag. In the tropical lower stratosphere, heating rates are 30-40% weaker in ERA5 than in ERA-Interim, likely correcting a bias in ERA-Interim. At 20km and in the NH stratosphere, ERA5 mean age values are around the upper margin of the uncertainty range from historical tracer observations, indicating a somewhat slow-biased BDC. The stratospheric age of air trend in ERA5 over 1989-2018 is negative and is related to an increase in tropical upwelling. However, the age decrease is not linear but steplike, potentially caused by multi-annual variability or changes in the observations included in the assimilation. Particularly regarding trends on decadal time scales, the different reanalyses can largely differ.

Abstract
The Atlantic Meridional Overturning Circulation (AMOC) involves a northward movement of warm upper waters accompanied by a southward movement of cold waters at depth, and carries up to 25% of the northward global atmosphere-ocean heat transport in the northern hemisphere. The natural variations in the AMOC can affect climate over decadal timescales, so there is an obvious need for better, more quantitative, forecasts of the future behavior of the AMOC. An accurate AMOC prediction system could potentially provide a valuable early warning of imminent climate change. As part of ongoing efforts to improve forecasting, the NCEP's Environmental Modeling Center (EMC) is developing the prototype version of the Next Generation Global Ocean Data Assimilation System (NG-GODAS). The NG-GODAS uses JEDI-based SOCA (Sea-ice Ocean Coupled Assimilation) as its ocean data assimilation component, and an advanced ocean model (MOM6) is used. The following satellite and in-situ observation data are assimilated in current system: satellite sea surface temperature/sea surface salinity, in-situ temperature and salinity, absolute dynamic topography, and sea ice concentration. The AMOC at 26.5N will be investigated using the porotype NG-GODAS, and a comparison would be presented of ocean reanalyses with different observing networks, to the observations and transport estimates from the RAPID mooring array across 26.5• N in the Atlantic.

Abstract
Indian Monsoon Data Assimilation and Analysis (IMDAA) is the currently available highest resolution (12 km), long-term (40 years, 1978-2018, extended to 2020), satellite era regional reanalysis over south Asian monsoon region. National Centre for Medium Range Weather Forecasting (NCMRWF), Ministry of Earth Sciences, Government of India distributes the IMDAA reanalysis through reanalysis web portal, https://rds.ncmrwf.gov.in/. This study illustrates location specific verification of surface temperature estimates from IM-DAA and its comparison with ERA5 global reanalysis at the selected locations over Indian landmass for 19 years during the 21st century, from 2000 to 2018. Fourteen stations selected, such that (i) they are representatives of different homogenous temperature regions of India, and (ii) continuous in-situ surface observations are available during the period of study with minimum gap. Verification of surface temperature estimates (maximum, minimum and mean) shows that, quality of IMDAA is better over the Tropics and coastal regions (south of 20•N), whereas ERA5 outperforms over extra-tropics (north of 20•N). This could be due to better representation of global circulation in the global reanalysis ERA5, and local effects more effectively in the regional reanalysis IMDAA. High resolution feature of IMDAA (12 km) compared to ERA5 (31 km) better captures surface temperature over the orographic region. Surface maximum and minimum temperature estimates show that IMDAA has a comparatively hot summer and cool winter over north India, whereas the reverse in the ERA5. In general, ERA5 shows better correlation with in-situ observations than IMDAA; however, the mean of the three datasets differs significantly (p < 0.05 ) Keywords: IMDAA, ERA5, Surface temperature, Regional reanalysis, Global reanalysis, NCM-RWF, Verification Evaluation and model performance, Tropics, Extra tropics, Orography, India Abstract Given the positive impact by the assimilation of Aeolus horizontal line of sight wind profiles particularly in the tropics at several operational centers around the world, we want to specifically investigate how beneficial this new observation type is for the initialization of tropical cyclones (TCs) in NOAA's Finite Volume Cubed Sphere Global Forecasting System (FV3GFS).To maximize the benefits of Aeolus, several key aspects that make data assimilation in the vicinity of TCs complicated, need to be addressed. By having inaccurate a-priori estimates of the state of the atmosphere under complex flow structures combined with the use of suboptimal quality control (QC) procedures can have detrimental impacts to the analysis. Static QC and background checks, which are based on blacklisting and fist-guess rejections can make it arbitrary to take decisions, such as keeping observations with initially large departures from the model background or how much weight should be given to an observation during the analysis update stage. To address suboptimal quality controls, we implemented the assimilation of Aeolus Mie-cloudy and Rayleigh-clear observation regimes with additional Variational Quality Control (VarQC) on NOAA's FV3GFS. This VarQC algorithm can assign adaptive weights and address non-Gaussianity aspects of Aeolus observations. VarQC can also be beneficial to TC analysis and forecast as it considers information about the local TC flow, the a-priori estimates of relevant sources of error, and the analysis state in a synergistic manner. In this presentation, we describe the benefits of applying VarQC to the Aeolus observations for improving the quality of the analysis and forecast in NOAA's FV3GFS during TC activity.
Extending a forward operator for visible satellite channels by near-infrared and aerosol capabilities Abstract Satellite images in the solar spectrum provide high resolution cloud and aerosol information and thus present a promising observation type for data assimilation and model evaluation. While visible channels contain information on the cloud distribution, cloud optical thickness and cloud structure, near-infrared channels are in addition more sensitive to cloud microphysical properties and can be used to distinguish between water and ice clouds. Moreover, solar channels are sensitive to aerosols, so their assimilation can be expected to improve forecasts of cloud and aerosol distribution, and thus also solar radiation. However, mainly due to a lack of sufficiently fast forward operators for visible and nearinfrared radiances, operational data assimilation systems so far use only the thermal channels. With recent development of MFASIS, a 1D radiative transfer (RT) method that is similarly accurate but orders of magnitude faster than conventional RT solvers for the solar spectrum, it has become possible to utilize visible channels.Here we discuss MFASIS's limitations preventing it from simulating near-infrared channels accurately and present a solution increasing the accuracy significantly for near-infrared channels. Furthermore, it will be demonstrated that replacing MFASIS's look-up table by a neural network reduces computational costs significantly, thus allowing for additional input parameters. Those parameters enable us to describe the vertical aerosol distribution for multiple species. This extends the application of MFASIS towards the assimilation of aerosol-affected radiances. The new approaches presented are tested using IFS and ICON model output. Abstract Oil spills at sea pose a serious threat to the coastal environment. To control and limit unreported spills, it is essential to identify pollution sources, and satellite imagery can be an effective tool for this purpose. We present in this work a Bayesian inference approach to identify the source parameters of a spill from contours of oil slicks detected by satellite images. The approach adopts an observation error model based on a non-local measure of the dissimilarity between the predicted and observed contours. A Markov chain Monte Carlo (MCMC) technique is then employed to sample the posterior distribution of five parameters of interest: the x and y coordinates of the source of release, the time and duration of the spill, and the quantity of oil released. To make the estimation of the posterior distribution computationally feasible, a Polynomial Chaos-based surrogate of the oil spill model is used within MCMC. To that end, a feature-based object localization method based on image moments is proposed to approximate contours, or binary images, in the form of integral quantities, for which surrogate models can be built. Two synthetic experiments of a spill released from a fixed point source are investigated, where a contour is completely observed in the first case, while two contours are partially observed at different times in the second case. In both experiments, the proposed framework is able to provide good estimates of the source parameters along with a level of confidence reflected by the uncertainties within. In the case of partial observations, the estimated parameters can be used to reconstruct the missing parts of an observed slick from which an oil spill model can be initiated to better forecast the spread of oil.

Abstract
In addition to measuring properties of hydrometeors and their motion, radars are capable of measuring the average refractive index between the radar and ground targets, or rather its temporal change. This measurement, physically similar to the GNSS refractivity, is focused on the first tens of meters of the atmosphere and on a few tens of kilometers around the radar, and are available under all weather conditions. Radar-measured refractivity data are mindboggingly precise and are very sensitive to humidity changes, as well as being representative of conditions at meso-beta scales instead of at a point. Because refractivity measurements are difficult to interpret by people, they have not seen much use in the forecaster-centric world of operational radars. But they have the potential of being particularly useful to constrain surface properties and humidity in the lower boundary layer. In this poster, examples of measurements and applications of radar-measured refractivity will be presented.
Operational direct assimilation of radar reflectivity volumes with KENDA at Arpae-SIMC

Abstract
This study aims to improve the precipitation forecasts from numerical weather prediction models through effective assimilation of satellite-observed precipitation data. We have been developing a global atmospheric data assimilation system NICAM-LETKF, which comprises the Nonhydrostatic ICosahedral Atmospheric Model (NICAM) and Local Ensemble Transform Kalman Filter (LETKF). This study pioneers to assimilate radar reflectivity measured by the Dual-frequency Precipitation Radar (DPR) onboard the Global Precipitation Measurement (GPM) core satellite into the NICAM. We conduct the NICAM-LETKF experiments at 28-km horizontal resolution with explicit cloud microphysics of a single-moment 6-class bulk microphysics scheme. To simulate GPM DPR reflectivity from NICAM model outputs, the Joint-Simulator (Hashino et al. 2013; JGR) is used. Our initial tests showed a better match with the observed reflectivity by assimilating GPM DPR reflectivity into NICAM forecasts. However, the results from a 1-month data assimilation cycle experiment showed general degradation by assimilating GPM DPR reflectivity. For better use of GPM DPR reflectivity data, we estimated a model cloud physics parameter corresponding to snowfall terminal velocity by data assimilation. Parameter estimation reduced the snowfall terminal velocity, and successfully mitigated the gap between simulated and observed Contoured Frequency by Altitude Diagram (CFAD). The estimated parameter also improved temperature and humidity fields in the mid-to lower troposphere, and precipitation forecasts.

Abstract
The current surface-based observing network has an insufficient density to fully characterize atmospheric processes at scales of Meso-β (and below). It is therefore unable to capture many relevant meteorological phenomena, especially in the boundary layer. Spacebased measurements are unlikely to resolve this deficiency. A new generation of ground-based remote sensing instruments, often called "profilers", has meanwhile become commercially available. These instruments are able to provide continuous measurements of kinematic, thermodynamic and cloud/aerosol particle related variables, mostly in the form of vertical profiles. Benefits of assimilating such data were recently seen in field campaigns. It is therefore timely to ask whether such profilers can also be used successfully in an operational setting. The presentation will give an overview of the project "Pilotstation" at DWD, which is investigating various options for a qualitative extension of the surface-based observing network. A testbed approach is employed to assess data availability, quality, observation impact as well as operational sustainability for the following profilers: Doppler lidar, microwave radiometer, water vapor broadband-DIAL, cloud radar and Raman lidar. Abstract Space-borne radar observations are currently emerging has an observation kind important to consider within Numerical Weather Prediction applications. Like for the forward simulation of passive microwave observations, radar data simulations require to make multiple assumptions including on the scattering properties of hydrometeors. With the objective of simulating both active and passive microwave instruments within a single framework using the same radiative transfer assumptions into a widely-used tool in the NWP community, a first version of active sensor module has recently been released within Version 13 of the RT-TOV software by the EUMETSAT NWP SAF. This initial version supports the simulation of both the GPM/Dual frequency Precipitation Radar and the Cloudsat/Cloud Precipitation Radar. Simulations of the GPM/DPR, performed with RTTOV V13 and the ARPEGE global model running operationally at Météo-France will be shown. Comparisons will be performed with observations, both on a case study as well as on a large number of samples. In particular, a sensitivity of the simulations to the hydrometeor fraction profile specifications will be discussed. Ward, Kobe, Hyogo 650-0047, Japon, Japan

Abstract
Sudden heavy rain may lead to disasters like flooding and loss of life and property. To reduce the risk, predicting sudden downpours is of key importance. However, predictability of such events is limited to only for a very short range within an hour or shorter because of their abruptness. In this case nowcasting is an effective approach. Detecting sudden heavy rain even 10 minutes before it occurs can reduce the damage drastically. Precipitation nowcasting is the process of short-range prediction based on observation data. In the case of sudden rainfalls, this process is difficult due to the fast evolution of the rain and its chaotic nature. Therefore, we need innovative techniques. The novel Phased-Array Weather Radar (PAWR) offers dense 3D images of reflectivity every 30 seconds. We took advantage of this big data to perform nowcasting using neural networks. We use Residual Neural Networks (RESNet) to compress the images and extract information relevant for the prediction. Next, we use a Convolutional Neural Network (CNN) to make the prediction. Afterwards, we use the same RESNet to map the forecast to the original domain. The RESNet and the CNN are trained jointly for the compression to maximize the prediction accuracy. Our first results show that in most cases we can predict precipitations up to 30 minutes, with an error rate (false positives + false negatives) of 8%. The use of the RESNet allowed to alleviate the memory load and the computational complexity of the prediction. Moreover, training the RESNet and the CNN jointly reduced immensely the prediction noise in non-precipitation regions and improved the accuracy in precipitation regions.
Keywords: Weather Forecasting, Neural Networks, Precipitation Nowcasting * Speaker Impact of ground-based water vapour and temperature lidar profiles on short-range forecast skill by means of hybrid 3DVAR-ETKF data assimilation Zurich-Airport, Switzerland

Abstract
Assimilation of ground-based thermodynamic lidar observations has augmented numerical weather prediction capabilities from nowcasting to the very short-range, short-range, and medium-range. In this study, temperature and water vapour profiles obtained from the temperature Raman lidar and the water vapour differential absorption lidar, respectively, of the University of Hohenheim are assimilated into the Weather Research and Forecasting (WRF) model through a new forward operator. The operator directly incorporates the water vapour mixing ratio, avoiding undesirable cross sensitivities to temperature, enabling complete observation concerning the water vapour contents to be propagated into the model. The assimilation was performed with the three dimensional variational DA system and with the hybrid 3DVAR Ensemble Transform Kalman Filter approach at a convection-permitting resolution. The 3DVAR-ETKF experiment resulted in a 50% smaller temperature and water vapour RMSE than the 3DVAR experiment. The planetary boundary layer height (PBLH) of the analyses also showed improvement compared to available ceilometer data. A single lidar vertical profile impact spreads over a 100 km radius, promising future assimilation of water vapour and temperature data from operational lidar networks. Forecast improvement with respect to PBLH was observed for about 7 hours, while an improvement of integrated water vapour lasts for 4 hours. We also present some significant collaborative effort with the Raman lidar for meteorological observation (RALMO) from the MeteoSwiss. Also, some initial results from the assimilation Atmospheric Raman Temperature and Humidity Sounder (ARTHUS) data will be shown. This study focuses on future instruments onboard MTG-I (Meteosat Third Generation) and MSG-B (MetOp Second Generation): the Flexible Combined Instrument (FCI), the MicroWave Imager (MWI) and the Ice Cloud Imager (ICI). Due to their different spectral ranges, they are sensitive to various and complementary quantities within clouds and precipitation. The objective is to identify the key components of the assimilation system for reaching a synergistic use of these observations in an all-sky context. The ability of the Bayesian inversion to provide complementary information is quantified with simulations from radiative transfer model RTTOV v.13 and lagged forecasts for the observations and the background.
Statistical results based on a wide sample of profiles from the ARPEGE forecast model will enable to build a global evaluation for IR and MW observations, and thus measure the degree of consistency and the differences in the retrievals from IR and MW observations. Two steps are followed: (i) considering a perfect forward model to understand the source of differences from the retrieval method and the spectral range ; (ii) introduction of errors in the radiative transfer model to understand the differences introduced by the hypotheses used. Preliminary results will be shown.