CEMS-Flood data is associated with a certain terminology that can be difficult to grasp. This page summarises some of the most important terms that can be useful when reading the wiki pages and dealing with the data.
The terms are listed in alphabetical order.
|An analysis is our best simulated estimate of reality, used as a proxy for observations which do not exist everywhere. It is the archived lead time 0 of a forecast. For example, the archived EFAS simulation forced with observation (sfo in MARS) is an analysis.
|Archived forecasts are forecast runs generated in real time 'fashion' for dates in the past. Unlike the reforecasts, they are not re-run when the model changes, so the model version used is not consistent throughout time.
|A dataset against which another dataset is compared. It can be a reference dataset or the forecast outputs from a previous version of the system.
|A class defines different groups of data in MARS. Class ce stands for Copernicus Emergency Management Services. This is important for accessing data through the Climate Data Store and MARS.
|A climatology is a collection of information describing the climate. For example, a flood climatology will include information about flood events happening over a long period. These include the statistics of the climate, such as mean or standard deviation of certain flood events or the flood percentiles (floods happening with certain frequency) or return period magnitudes (extreme floods expressed in recurrence interval).
|A hydrological simulation forced with observations where the input have gone through a quality check procedure. It is generally available with a delay of several days to months. It is recommended to be used as reference dataset or initial conditions of reforecasts.
|Control forecast (cf)
|The control forecast is the special, unperturbed member of the ensemble forecast. It is run from the best initial conditions (that are used in the HRES) on the lower resolution of the ensemble system. It should be used in combination with the perturbed forecasts. The ensemble control forecast data is archived in MARS under type 'cf'.
|A single forecast realisation from a Numerical Weather Prediction (NWP) model. Typically, deterministic forecasts are associated with high spatial resolution that require more computational resource to run. Deterministic forecasts do not contain any information on the forecast uncertainty.
|A number of simulations (members) using the same Numerical Weather Prediction (NWP) model, but run from varying initial conditions. Examples are the ECMWF-ENS and the COSMO-LEPS ensemble forecasts. The ensemble members are run from perturbed conditions (pushed away from the analysed/observed state), with regards to the best estimate of the initial state of the Earth system. The spread (standard deviation) of the ensemble should ideally reflect the model error. The ensemble forecast is a probabilistic forecast where the distribution of future scenarios is given by the equally likely ensemble members. The mean of an ensemble forecast has been shown, on average, to have higher predictive skill than the corresponding high-resolution HRES forecast beyond the first few days of the forecast horizon. The ensemble forecasts generally show a smoother run to run evolution than a deterministic forecast as the members (all or most) very rarely change in the same way. Ensemble forecasts generally contain a control member, and a set of perturbed members.
|Ensemble Streamflow Prediction (ESP)
|A hydrological forecasting method developed in the mid-1970s for reservoir operations in the US. It consists of starting a hydrological model with the latest initial hydrological conditions (e.g., soil moisture, river level, snow cover, etc) and forcing it with historical meteorological observations (e.g., precipitation, temperature, evaporation, etc), or reanalysis. Each ensemble member thus corresponds to a year of historical meteorological observations. For example, to generate hydrological forecasts on the 1st of February 2016 for the next 7 months of lead time, the hydrological model are started with the hydrological conditions observed on the 1 February 2016. Meteorological forcing are then taken from each available year of meteorological observations for the 1st of February - 31st August (+7 months).
ERA5 is a comprehensive reanalysis from 1950 to near real time, which assimilates as many observations as possible in the upper air and near surface. The ERA5 atmospheric model is coupled with a land surface model and a wave model. It is generated by ECMWF for the Copernicus Climate Change. For further details, visit ERA5 documentation.
|Experiment version (expver)
|An experiment version defines the type of simulation. Expver 1 stands for the operational data and is the official release of the data. Other numbers define test versions and experimental data, and should only be used with caution.
|Extended-range (or sub-seasonal) forecast
|Forecast produced with lead times beyond 15 days but less than the seasonal range. The ECMWF extended range forecast is an ensemble forecast with a lead time of 46 days issued bi-weekly (Mondays and Thursdays) at 00UTC.
|A period used to bridge the gap between the latest (proxy-)observations (e.g. observed meteorological fields or reanalysis forcing such as ERA5) and the start of the forecast (at 00UTC or 12UTC, several days after the latest observations are available) when running a hydrological forecast chain. It is generated using the most recent high-resolution Numerical Weather Prediction forecast as forcing, hence expected not to be as accurate as the operational simulations forced by (proxy-)observations. The fill-up data is not recommended for use except for very specialised modelling exercise for warm-start of past forecast reruns. The fill-up data is archived in MARS (similarly to the forecasts) under type 'fu'.
|A forecast horizon (also called lead time) is the length of time into the future the forecast is valid for. It is typically expressed in hours or days.
|High-resolution forecast (fc)
A single (or deterministic), high-resolution forecast realisation from a Numerical Weather Prediction (NWP) model. Examples are the ECMWF-HRES and the DWD-Det high-resolution deterministic forecasts. The high-resolution forecast does not contain uncertainty information and is considered less reliable than the mean of the (probabilistic) ensemble forecast beyond the first few days of the forecast horizon. It is usually better (through the higher horizontal resolution) than ensemble forecasts in resolving local features, such as orographic precipitation. However, it is also likely to show higher variability (sometimes even jumpiness) than the ensemble forecast, just as any change in the weather situation, such as geographical displacement (and thus precipitation), will directly manifest in the outcomes. The HRES forecast data is archived in MARS under type 'fc'.
A hindcast is a commonly used term for reforecast. They tend to be used less than reforecast nowadays.
|Hydrological conditions (e.g. soil moisture, snow cover, river discharge, etc.) at the start of the forecast integration.
|A hydrological simulation forced with observations created in near-real time, where the input data could be affected by quality issues such as missing or erroneous values. It is available in near-real time. It is typically used as initial conditions for real-time hydrological forecasts, but should be avoided as reference dataset if a consolidated simulation is available.
|A lead time (also called forecast horizon) is the length of time into the future the forecast is valid for. It is typically expressed in hours or days.
|A commonly used term to describe a multi-decadal historical simulation forced with observation (or reanalysis) that can be used as reference dataset.
|Meteorological Archival and Retrieval System (MARS) is the ECMWF archiving system for storing its forecast data, accessible to eligible users. (see MARS user documentation for details)
|Maximum lead time
|The maximum lead time (or maximum forecast horizon) is the longest lead time a forecast is generated for.
|Forecasts produced for a lead time of several days (typically 3-15 days).
|Numerical weather prediction model (NWP)
|Mathematical models of the atmosphere, oceans, and land processes that predict the future weather based on current weather conditions.
|Perturbed forecast (pf)
|The perturbed members of the ensemble forecast without the unperturbed control member/forecast. Generally used in combination with the control forecast. The order of the numbering of the perturbed forecasts is random. The perturbed ensemble member forecast data is archived in MARS under type 'pf'.
A probabilistic forecast predicts a whole distribution of future scenarios, instead of a single (deterministic) outcome. The distribution of future outcomes gives information on the probability and thus the uncertainty for all these possible events to happen.
|A reanalysis is the same as analysis (estimate of the atmosphere, ocean or land processes) but produced using a fixed model version giving the benefit that the model does not change over the time period of the reanalysis (note that the observations assimilated does still change over time depending on what was available throughout the long period).
|A dataset used to derive climatologies which provide a relevant context to forecasts (i.e how severe the actual forecast in the context of the climatology). It is often a simulation, including historical simulations forced with observations, reanalysis or reforecasts.
|Forecast simulations performed for a set of past dates, based on a configuration as close as possible to the operational setting. They are used for example to evaluate the forecast skills against a benchmark or reference forecast. They sometimes are called hindcasts, especially in the context of seasonal forecasting.
|Forecasts produced with a lead time of several months (typically 30 days to 7 months).
|Forecasts produced for a lead time of few days only (usually maximum 2 or 3 days).
|Simulation forced with observation (sfo)
Hydrological simulation, in the form of long time series forced with observations, reanalysis or other proxy for observations. It is considered to be the best estimate of the reality, can can be considered as a hydrological reanalysis. It can be used to monitor and analyse past events, as initial conditions for the reforecasts, or as reference dataset if covering a long, multi-decadal historical period. The simulation forced with observation data is archived in MARS under type 'sfo'.
|Time increments for simulation dataset such as forecast, reforecast or reanalysis. For a forecast, a time step of 6 hours means that the forecast as a value every 6 hours until the maximum forecast horizon.
The total area that contributes with water to a specific point on the river network. Also known as a catchment area.