Postprocessing wind forecasts with deep learning in complex terrain

2021 ◽  
Author(s):  
Daniele Nerini ◽  
Jonas Bhend ◽  
Christoph Spirig ◽  
Lionel Moret ◽  
Mark Liniger
Keyword(s):  
Prediction Models ◽  
Weather Prediction ◽  
Arbitrary Point ◽  
Horizontal Resolution ◽  
Supervised Machine Learning ◽  
Regression Problem ◽  
Numerical Predictions ◽  
Operational Forecasts ◽  
First Results ◽  
Elevation Model

<p>To improve and automate the quality of weather forecasts to the public, MeteoSwiss is redesigning its statistical postprocessing suite. The effort aims at producing calibrated probabilistic predictions to any arbitrary point in space and up to a 15-day lead time, by seamlessly integrating multiple numerical weather prediction models into a unique consensus forecast.</p><p>For hourly wind forecasts (mean, gust, and direction), the task is formulated as a regression problem in a supervised machine learning framework, where station measurements are used as labels, and co-located NWP forecasts as features. To improve the estimates at ungauged locations, additional static topographical features are derived from a 50m digital elevation model. The probabilistic component is included by training the neural network not to produce a deterministic prediction, but the parameters of a conditional probability function. To this end, the Continuous Ranked Probability Score (CRPS) is used as a loss function.</p><p>The dataset includes a range of surface parameters at hourly resolution produced by the operational forecasts from three NWP models (the deterministic COSMO-1 model, at 1 km horizontal resolution; the 21-member COSMO-E, 2 km; and the 51-member ECMWF IFS ENS at about 18 km). The data cover the whole of Switzerland over a period spanning more than four years (mid 2016 to end of 2020). Wind measurements from over 500 surface weather stations are included as reference dataset. The study uses a train-validation-test split in both space and time to assess the ability of the postprocessing model to generalize to unseen locations and times.</p><p>The results indicate that, despite the challenging nature of the problem, the postprocessing model can improve over the baseline NWP forecasts in terms of CRPS on the test set. In particular, the model is effectively correcting for biases relating to altitude error and other misrepresentations in the NWP topography. The results show that it is feasible to downscale numerical predictions to a substantially higher spatial resolution. Moreover, the conditional probabilities shows consistent improvements in terms of calibration, although it remains a significant portions of undetected peak events (positive outliers), possibly to be related to unpredictable phenomena (e.g., thunderstorm gusts). Finally, first results seem to suggest that the gain in prediction skill is mainly driven by a better statistical reliability rather than higher statistical resolution.</p>

scholarly journals Validation of the First Years of GPM Operation over Cyprus

2018 ◽  
Vol 10 (10) ◽  
pp. 1520 ◽  
Author(s):  
Adrianos Retalis ◽  
Dimitris Katsanos ◽  
Filippos Tymvios ◽  
Silas Michaelides
Keyword(s):  
High Resolution ◽  
Prediction Models ◽  
Weather Prediction ◽  
Monthly Basis ◽  
Rain Gauges ◽  
Precipitation Measurement ◽  
Precipitation Estimates ◽  
First Results ◽  
Global Precipitation Measurement ◽  
Global Precipitation

Global Precipitation Measurement (GPM) high-resolution product is validated against rain gauges over the island of Cyprus for a three-year period, starting from April 2014. The precipitation estimates are available in both high temporal (half hourly) and spatial (10 km) resolution and combine data from all passive microwave instruments in the GPM constellation. The comparison performed is twofold: first the GPM data are compared with the precipitation measurements on a monthly basis and then the comparison focuses on extreme events, recorded throughout the first 3 years of GPM’s operation. The validation is based on ground data from a dense and reliable network of rain gauges, also available in high temporal (hourly) resolution. The first results show very good correlation regarding monthly values; however, the correspondence of GPM in extreme precipitation varies from “no correlation” to “high correlation”, depending on case. This study aims to verify the GPM rain estimates, since such a high-resolution dataset has numerous applications, including the assimilation in numerical weather prediction models and the study of flash floods with hydrological models.

The Life-Cycle of Cloud and Precipitation Microphysics in Radar Observation and Numerical Model

2021 ◽  
Author(s):  
Gregor Möller ◽  
Florian Ewald ◽  
Silke Groß ◽  
Martin Hagen ◽  
Christoph Knote ◽  
...  
Keyword(s):  
Prediction Models ◽  
Model Simulation ◽  
Weather Prediction ◽  
Horizontal Resolution ◽  
Polarimetric Radar ◽  
Dual Frequency ◽  
Microphysics Scheme ◽  
Ka Band ◽  
Radar Observations ◽  
Conversion Rates

<p>The representation of microphysical processes in numerical weather prediction models remains a main source of uncertainty. To tackle this issue, we exploit the synergy of two polarimetric radars to provide novel observations of model microphysics parameterizations. In the framework of the IcePolCKa project (Investigation of the initiation of Convection and the Evolution of Precipitationusing simulatiOns and poLarimetric radar observations at C- and Ka-band) we use these observations to study the initiation of convection as well as the evolution of precipitation. At a distance of 23 km between the C-band PoldiRad radar of the German Aerospace Center (DLR) in Oberpfaffenhofen and the Ka-band Mira35 radar of the Ludwig-Maximilians-University of Munich (LMU), the two radar systems allow targeted observations and coordinated scan patterns. A second C-band radar located in Isen and operated by the German Weather Service (DWD) provides area coverage and larger spatial context. By tracking the precipitation movement, the dual-frequency and polarimetric radar observations allow us to characterize important microphysical parameters, such as predominant hydrometeor class or conversion rates between these classes over a significant fraction of the life time of a convective cell. A WRF (Weather Research and Forecasting Model) simulation setup has been established including a Europe-, a nested Germany- and a nested Munich- domain. The Munich domain covers the overlap area of our two radars Mira35 and Poldirad with a horizontal resolution of 400 m. For each of our measurement days we conduct a WRF hindcast simulation with differing microphysics schemes. To allow for a comparison between model world and observation space, we make use of the radar forward-simulator CR-SIM. The measurements so far include 240 coordinated scans of 36 different convective cells over 10 measurement days between end of April and mid July 2019 as well as 40 days of general dual-frequency volume scans between mid April and early October 2020. The performance of each microphysics scheme is analyzed through a comparison to our radar measurements on a statistical basis over all our measurements.</p>

Convection is key to better polar low forecasts

2020 ◽  
Author(s):  
Matilda Hallerstig ◽  
Linus Magnusson ◽  
Erik Kolstad
Keyword(s):  
Prediction Models ◽  
Deterministic Model ◽  
Rapid Development ◽  
Weather Prediction ◽  
Model Performance ◽  
Horizontal Resolution ◽  
Added Value ◽  
Grid Spacing ◽  
Polar Lows ◽  
Polar Low

<p>ECMWF HRES and Arome Arctic are the operational Numerical Weather Prediction models that forecasters in northern Norway use to predict Polar lows in the Nordic and Barents Seas. These type of lows are small, but intense mesoscale cyclones with strong, gusty winds and heavy snow showers. They cause hazards like icing, turbulence, high waves and avalanches that threaten offshore activity and coastal societies in the area. Due to their small size and rapid development, medium range global models with coarser resolutions such as ECMWF have not been able to represent them properly. This was only possible with short range high resolution regional models like Arome. When ECMWF introduced their new HRES deterministic model with 9 km grid spacing, the potential for more precise polar low forecasts increased. Here we use case studies and sensitivity tests to examine the ability of ECMWF HRES to represent polar lows. We also evaluate what added value the Arome Arctic model with 2.5 km grid spacing gives. For verification, we use coastal meteorological stations and scatterometer winds. We found that convection has a greater impact on model performance than horizontal resolution. We also see that Arome Arctic produces higher wind speeds than ECMWF HRES. To improve performance during polar lows for models with a horizontal grid spacing less than 10 km, it is therefore more important to improve the understanding and formulation of convective processes rather than simply increasing horizontal resolution.</p>

scholarly journals Solar Irradiance Forecasts by Mesoscale Numerical Weather Prediction Models with Different Horizontal Resolutions

Energies ◽  
2019 ◽  
Vol 12 (7) ◽  
pp. 1374
Author(s):  
Ohtake ◽  
Uno ◽  
Oozeki ◽  
Hayashi ◽  
Ito ◽  
...  
Keyword(s):  
Numerical Weather Prediction ◽  
Solar Irradiance ◽  
Prediction Models ◽  
Weather Prediction ◽  
Longwave Radiation ◽  
Horizontal Resolution ◽  
Numerical Weather ◽  
Downward Longwave Radiation ◽  
Radiation Processes ◽  
Numerical Weather Prediction Models

This study examines the performance of radiation processes (shortwave and longwave radiations) using numerical weather prediction models (NWPs). NWP were calculated using four different horizontal resolutions (5, 2 and 1 km, and 500 m). Validation results on solar irradiance simulations with a horizontal resolution of 500 m indicated positive biases for direct normal irradiance dominate for the period from 09 JST (Japan Standard Time) to 15 JST. On the other hand, after 15 JST, negative biases were found. For diffused irradiance, weak negative biases were found. Validation results on upward longwave radiation found systematic negative biases of surface temperature (corresponding to approximately −2 K for summer and approximately −1 K for winter). Downward longwave radiation tended to be weak negative biases during both summer and winter. Frequency of solar irradiance suggested that the frequency of rapid variations of solar irradiance (ramp rates) from the NWP were less than those observed. Generally, GHI distributions between the four different horizontal resolutions resembled each other, although horizontal resolutions also became finer.

scholarly journals Assessing parametrization uncertainty associated with horizontal resolution in numerical weather prediction models

2014 ◽  
Vol 372 (2018) ◽  
pp. 20130284 ◽  
Author(s):  
Glenn Shutts ◽  
Alfons Callado Pallarès
Keyword(s):  
Prediction Models ◽  
Weather Prediction ◽  
Model Error ◽  
Horizontal Resolution ◽  
Phase Changes ◽  
Ensemble Prediction ◽  
Condensation Temperature ◽  
Ensemble Prediction System ◽  
The Mean ◽  
Temperature Tendency

The need to represent uncertainty resulting from model error in ensemble weather prediction systems has spawned a variety of ad hoc stochastic algorithms based on plausible assumptions about sub-grid-scale variability. Currently, few studies have been carried out to prove the veracity of such schemes and it seems likely that some implementations of stochastic parametrization are misrepresentations of the true source of model uncertainty. This paper describes an attempt to quantify the uncertainty in physical parametrization tendencies in the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System with respect to horizontal resolution deficiency. High-resolution truth forecasts are compared with matching target forecasts at much lower resolution after coarse-graining to a common spatial and temporal resolution. In this way, model error is defined and its probability distribution function is examined as a function of tendency magnitude. It is found that the temperature tendency error associated with convection parametrization and explicit water phase changes behaves like a Poisson process for which the variance grows in proportion to the mean, which suggests that the assumptions underpinning the Craig and Cohen statistical model of convection might also apply to parametrized convection. By contrast, radiation temperature tendency errors have a very different relationship to their mean value. These findings suggest that the ECMWF stochastic perturbed parametrization tendency scheme could be improved since it assumes that the standard deviation of the tendency error is proportional to the mean. Using our finding that the variance error is proportional to the mean, a prototype stochastic parametrization scheme is devised for convective and large-scale condensation temperature tendencies and tested within the ECMWF Ensemble Prediction System. Significant impact on forecast skill is shown, implying its potential for further development.

The impact of vertical resolution on tropical cyclone simulation

2020 ◽  
Author(s):  
Wei Huang ◽  
Mengjuan Liu ◽  
Xu Zhang ◽  
Jian-wen Bao
Keyword(s):  
Tropical Cyclone ◽  
Prediction Models ◽  
Weather Prediction ◽  
Vertical Mixing ◽  
Forecast Model ◽  
Horizontal Resolution ◽  
Model Solution ◽  
Vertical Resolution ◽  
Solution Convergence ◽  
The Impact

<p>It is well known that horizontal resolution has a great deal of impact on tropical cyclone simulations using numerical weather prediction models.  It is relatively less discussed in the literature how vertical resolution affects the solution convergence of tropical cyclone simulations.  In this study, the resolved kinetic energy spectrum, the Richardson number probability density function and resolved flow features are used as metrics to examine the behavior of solution convergence in tropical cyclone simulations using the Weather and Forecast Model (WRF).  It is found that for convective-scale simulations of a real tropical cyclone case with 3-km horizontal resolution, the model solution does not converge until a vertically stretched vertical resolution approaches 200 layers or more.  The results from this study confirm the results from a few previous studies that the subgrid turbulent mixing, particularly, the vertical mixing, plays a significant role in the behavior of model solution convergence with respect to vertical resolution.  They also provide a basis for the vertical grid configuration selection for the operational tropical cyclone model of Shanghai Meteorological Service.</p>

The impact of SST on the weather forecast quality in the Bulgarian Antarctic Base area on Livingstone Island

2020 ◽  
Author(s):  
Boriana Chtirkova ◽  
Elisaveta Peneva
Keyword(s):  
Prediction Models ◽  
Weather Prediction ◽  
Wrf Model ◽  
Weather Forecast ◽  
Horizontal Resolution ◽  
Initial Time ◽  
Test Cases ◽  
Validation Data ◽  
The Impact ◽  
Forecast Quality

<p>The weather forecast of good quality is essential for the humans living and operating in the Bulgarian Antarctic Base. The numerical weather prediction models in southern high latitude regions still need improvement as the user community is limited, little test cases are documented and validation data are scarce. Not lastly, the challenge of distributing the output results under poor internet conditions has to be addressed.</p><p>The Bulgarian Antarctic Base (BAB) is located on the Livingstone Island coast at 62⁰S and 60⁰W. The influence of the Southern ocean is significant, thus important to be correctly taken into account in the numerical forecast. The modeling system is based on the WRF model, configured in three nested domains down to 1 km horizontal resolution, centered in BAB. The main objective of the study is to quantify the Sea Surface Temperature (SST) impact and to recommend the frequency and way to perform measurements of the SST near the base. The focus is on prediction of right initial time and period of “bad” weather events like storms, frontal zones, and severe winds. Several test cases are considered with available measurements of temperature, pressure and wind speed in BAB during the summer season in 2017. The numerical 3 days forecast is performed and the model skill to capture the basic meteorological events in this period is discussed. Sensitivity experiments to SST values in the nearby marine area are concluded and the SST influence on the model forecast quality is analyzed.</p>

scholarly journals An Evaluation of the Impact of Assimilating AERI Retrievals, Kinematic Profilers, Rawinsondes, and Surface Observations on a Forecast of a Nocturnal Convection Initiation Event during the PECAN Field Campaign

2019 ◽  
Vol 147 (8) ◽  
pp. 2739-2764 ◽  
Author(s):  
Samuel K. Degelia ◽  
Xuguang Wang ◽  
David J. Stensrud
Keyword(s):  
Prediction Models ◽  
Weather Prediction ◽  
Mesoscale Convective System ◽  
Convective System ◽  
Operational Forecasts ◽  
Outflow Boundary ◽  
Convection Initiation ◽  
The Impact ◽  
Surface Observations ◽  
Wind Profilers

Abstract Numerical weather prediction models often fail to correctly forecast convection initiation (CI) at night. To improve our understanding of such events, researchers collected a unique dataset of thermodynamic and kinematic remote sensing profilers as part of the Plains Elevated Convection at Night (PECAN) experiment. This study evaluates the impacts made to a nocturnal CI forecast on 26 June 2015 by assimilating a network of atmospheric emitted radiance interferometers (AERIs), Doppler lidars, radio wind profilers, high-frequency rawinsondes, and mobile surface observations using an advanced, ensemble-based data assimilation system. Relative to operational forecasts, assimilating the PECAN dataset improves the timing, location, and orientation of the CI event. Specifically, radio wind profilers and rawinsondes are shown to be the most impactful instrument by enhancing the moisture advection into the region of CI in the forecast. Assimilating thermodynamic profiles collected by the AERIs increases midlevel moisture and improves the ensemble probability of CI in the forecast. The impacts of assimilating the radio wind profilers, AERI retrievals, and rawinsondes remain large throughout forecasting the growth of the CI event into a mesoscale convective system. Assimilating Doppler lidar and surface data only slightly improves the CI forecast by enhancing the convergence along an outflow boundary that partially forces the nocturnal CI event. Our findings suggest that a mesoscale network of profiling and surface instruments has the potential to greatly improve short-term forecasts of nocturnal convection.

scholarly journals Met Office Unified Model Tropical Cyclone Performance Following Major Changes to the Initialization Scheme and a Model Upgrade

2016 ◽  
Vol 31 (5) ◽  
pp. 1433-1449 ◽  
Author(s):  
Julian T. Heming
Keyword(s):  
Prediction Models ◽  
Positive Impact ◽  
Weather Prediction ◽  
Horizontal Resolution ◽  
Unified Model ◽  
Track Forecast ◽  
Forecast Errors ◽  
Forecast Bias ◽  
Initialization Scheme ◽  
The Impact

Abstract The Met Office has used various schemes to initialize tropical cyclones (TCs) in its numerical weather prediction models since the 1980s. The scheme introduced in 1994 was particularly successful in reducing track forecast errors in the model. Following modifications in 2007 the scheme was still beneficial, although to a lesser degree than before. In 2012 a new trial was conducted that showed that the scheme now had a detrimental impact on TC track forecasts. As a consequence of this, the scheme was switched off. The Met Office Unified Model (MetUM) underwent a major upgrade in 2014 including a new dynamical core, changes to the model physics, an increase in horizontal resolution, and changes to satellite data usage. An evaluation of the impact of this change on TC forecasts found a positive impact both on track and particularly intensity forecasts. Following implementation of the new model formulation in 2014, a new scheme for initialization of TCs in the MetUM was developed that involved the assimilation of central pressure estimates from TC warning centers. A trial showed that this had a positive impact on both track and intensity predictions from the model. Operational results from the MetUM in 2014 and 2015 showed that the combined impact of the model upgrade and new TC initialization scheme was a dramatic cut in both TC track forecast errors and intensity forecast bias.

Sign in / Sign up

Export Citation Format

Share Document