Mapping locations onto CEMS-Flood river networks

page under construction ------------------------------

Are there any differences in the remapping process for EFAS & GloFAS? Which information can be merged into a general description? 

• A lot of terms are duplicated on these pages:  GloFAS mapping locations onto the river network & Mapping locations onto the GloFAS river network .  Karen O'Regan combined and updated page. 

As we cannot trust that the location of a station in the model is completely representative of the location of the station in the 'real world' (based off provided coordinates alone), the following protocol to map the stations on the river network should be followed. The overall concept of mapping locations onto the river networks is the same for both EFAS and GloFAS.

This page describes the protocol of mapping new locations onto the river network, using the example of GloFAS. In this context, ‘new locations’ are usually river discharge observation stations provided by GloFAS users. The locations represent the catchment of the river at their outlet points. During the process of mapping, the ‘best-fit’ of location in the GloFAS model corresponds with the location of the station on the river in the ‘real world’, based on the coordinates of the station from the data provider. 
 Users should rely on this protocol as a guideline in case they have to extract point time series. To extract point time series, the locations must be mapped onto the GloFAS river networ. 

Tools to use

• Google Earth or Google Maps (to verify the station metadata).
• High detail vector river shapefile (e.g. hydroSHEDS at https://hydrosheds.cr.usgs.gov/datadownload.php or http://gaia.geosci.unc.edu/rivers/
• GloFAS river network, represented by the upstream area map (https://cds.climate.copernicus.eu/cdsapp#!/dataset/cems-glofas-historical?tab=form)
• GIS software to display shapefiles and maps (e.g. QGIS at https://www.qgis.org/en/site/forusers/index.html)
• A good enhancement could be to use maps (e.g. Bing or satellite maps) in QGIS. This would provide a more complete workspace in QGIS with the real maps, the high-resolution river layer, and the GloFAS river network in one space overlayed onto each other. Steps to achieve this in QGIS-3:
1. Go to Settings/Options/Network and click on "Use proxy for web access".
2. Go to plugins and choose QuickMapServices and click install.
3. Maps will be available in the Web tab

• TSview software to display time series from .csv, .txt, excel or LISFLOOD files (https://git.ecmwf.int/users/mocm/repos/tsview/browse)

Step-1: Verify station metadata

• Make sure the station metadata is correctly provided, identify if any metadata are missing. These should be provided geographic coordinates (lat/lon in degrees), station and river names, and upstream area (if available).
• Search for the station location on Google Maps or Google Earth
• Download the GloFAS upstream area map and load it into QGIS or some other GIS software. It can also be useful be helpful to add a high detail vector river shapefile which can help to identify the real rivers. Alternatively, you can also load the Google Maps/Earth into the GIS software and analyse all layers together.
• Analyse the location of the stations with the other provided metadata and check if they appear to match with the location in Google Earth/Maps.
• Check how the user-provided metadata compares with the maps and the GloFAS river network (represented by the upstream area map). Analyse the upstream area values of the GloFAS river network in the vicinity of the provided river point location.
• This step is a manual, subjective process.

Step-2: Identify river network match and find the right river channel

• Use the GIS software and the loaded upstream area maps, river shapefile, and Google Maps (Earth) (or if the Google products are not available in the GIS software then open in a separate website).
• First, check how well the GloFAS river network represents the real rivers in the area.
• Check if the rivers, the main channel(s), and tributaries compare perfectly, well, or if there are minor/major discrepancies.
• Maybe the rivers are not represented at all (e.g. rivers are so close in reality which cannot be resolved in GloFAS), or in some cases, the GloFAS network is simply wrong for some reason (e.g. rivers flow into each other incorrectly, etc.), or the provided station drains only a very small area (i.e. below 1000 km² or even maybe only few hundred km²) which GloFAS cannot accurately represent.
• Alternatively in some areas of the world real rivers can be difficult to locate even on Google Maps/Earth (i.e. in very dry areas where ephemeral rivers flow only intermittently, or where rivers can change their channels like in big deltas), so in such cases, the GloFAS river network will more likely be incorrect or incomplete.
• Then locate the right GloFAS river channel.
• As an ideal scenario, the provided coordinates allow identifying the river section without uncertainty on Google Maps/Earth.
• The provided upstream area (if available), and thus the match between the provided and modelled areas will often help to increase the confidence, or in some cases, where the location is not so clear on Google Maps (e.g. the point is right at the confluence, and it is not clear which channel it is gauging), the provided upstream area will most likely help to decide about the correct channel (i.e. whether it is the main channel, or the tributary, etc.).
• In some rare cases the right channel cannot be identified due to problems with the GloFAS river network or uncertainties on the provided location (maybe erroneous coordinates, or uncertain position of real rivers in the area), or simply as a consequence of too small catchment area (i.e. below 1000 km² or even maybe only few hundred km²) of the provided station.
• To help increase the confidence of the river channel match, if possible, also compare the observation time series (if it is available simply plot the time series) with the GloFAS reanalysis time series.
• The GloFAS-ERA5 reanalysis can be downloaded from the Copernicus Climate Data Store (https://cds.climate.copernicus.eu/cdsapp#!/dataset/cems-glofas-historical) and the reanalysis data extracted for the selected GloFAS river location.
• Especially, if there is still some uncertainty about the river match, then moving the selected GloFAS point around, and comparing the reanalysis time series from those points with the observation time series, can highlight problems, wrong provided location, incorrect provided upstream area, etc.
• This step can be especially helpful if the mapping seems problematic and the best model point is ambiguous.
• This step is a manual, subjective process.

Step-3: Compare the river flow time series with nearby stations

• If the station has a time series associated with it compare the data to nearby stations to check if the time series can be trusted or if it requires further investigation.
• In QGIS, individuate nearby upstream/downstream stations.
• Plot the time series using tsview, excel, python, or whichever software can allow you to plot multiple time series
• In case the comparison gives ambiguous results, flag the station and comment as appropriate.

Step-4: Map the station and assign a flag according to mapping issues when necessary

• Based on Step-2 and Step-3, the best river pixel should be chosen on the selected GloFAS river channel and a flag (described in the table below) should be assigned to the station. By doing that a user can quickly identify if a station has an issue and filter out stations from any analysis.
• This is also done manually.
• Confidence level and flag assignment:

In all cases when the provided station is not mapped to the nearest GloFAS model pixel, then a note should be added into the GloFAS station information database about the mapping situation and the nature of the problems or difficulties (i.e. upstream area mismatch, shift of location, GloFAS river network problems, problems with the provided coordinates, difficult area with uncertain rivers, or reservoirs or lakes, etc).