Just press Continue on each page, and tick agreement when asked, to run the validation.
Feel free to customize the validation by tweaking the options.
You should start seeing validation results in a couple of minutes.
How do you want your validation sample?
Choose a model and its variables
Each model is downloaded from a different ESGF data server. Choose one model, and one to three of its variables: only those files are downloaded.
| Model | Sea surface temperature | Surface nitrate | Sea surface salinity |
|---|
This creates a directory called oceanval_demo, with data in it. Remove it when you have finished with the demo.
This shows how OceanVal works. It is not how to validate a climate model.
Only one year, 2010, is matched up here. A climate model is not meant to reproduce the weather of any particular year, so judging how well one does needs many years of output, not a single one.
The model output
- Model
- Simulation
- Variables
- Covers
- Files
Identifying the files and variables
Please wait!
Do you have your own observational data?
A recipe describes a set of observations once, so that you can use it again in any matchup by name, like recipe={"chlorophyll": "occci"}, which is one of OceanVal's own. Your recipes are kept in .oceanvalrc files: in the directory you run OceanVal from, for that directory only, or in your home directory, for everywhere. Which kind of data is the recipe for?
Check the units
OceanVal has just identified whether the units are the same or whether the observational data needs to be adjusted to be in the model units. Please check the unit conversions below before proceeding.
Each observation becomes observation × Multiply by + Add before it is compared with the model. Multiply by rescales the units: roughly 1.025 turns µmol/kg into mmol/m3. Add shifts the zero point: 273.15 turns °C into K. Leave both boxes empty when no conversion is needed, and fill in any row OceanVal could not work out. Its suggestions are in bold. The units must be confirmed, with the tick box beside Continue, before the matchup starts.
Reading the units from the simulation…
Gridded datasets
OceanVal's gridded recipes, and your own gridded data (netCDF).
| Matchup | Simulation variable | Simulation units | Observation variable | Observation units | Multiply by | Add |
|---|
No gridded datasets are being matched up.
Point datasets
OceanVal's ICES recipes, and your own point data. Your own point data is csv files, which have no units, so it is up to you to make sure they match the simulation's.
| Matchup | Simulation variable | Simulation units | Observation variable | Observation units | Multiply by | Add |
|---|
No point datasets are being matched up.
OceanVal matches each observation in these point datasets to the simulation output on the same day, as their point_time_res includes day. But the simulation output for their variables is coarser than daily: monthly, or once every few days. So most of their observations may find no simulation output on the same day, and be left out of the matchups.
Matching them by year and month instead compares each observation with the simulation output for its month. OceanVal suggests that, in bold, but you can choose for all of them at once, for each one, or keep them as they are.
| Variable | Dataset | Time resolution | Matched by now | Match observations by |
|---|
Reading the simulation's files
OceanVal is finding the simulation's output files, and the variables in them.
The simulation has changed, so OceanVal filled in the variables it recognised again. The rows you gave yourself are kept.
Choose the variables to analyze
| Environmental variable | Variable name in the simulation |
|---|
These simulation variables are time means, and their times look like the end of the period each mean covers: a day's mean dated 00:00 the next day, or a month's dated 00:00 on the 1st of the next month. If so, a January mean would count as February's, and the year's last mean would fall in the next year.
Their files have no time bounds to say for certain, so OceanVal does not shift anything unless you choose it. It suggests shifting every simulation time back one day, in bold, so that each mean falls in the period it covers. The dates of the files are read once you have chosen.
| Variable | Files | Times now | Shifted | Why they look like ends |
|---|
Looking at the files' times
OceanVal is checking whether the times of the files are the start or the end of the period each mean covers.
| Variable | Model variable | Files | Time resolution | Dates |
|---|
Everything you choose is worked out for every simulation, over the same years, which must be years all of them have, and the cell thickness (a variable in each simulation's output, or a file) is the same in all of them.
Select one or more of the analyses
Choose where to save the analysis
Reading the simulations' files
OceanVal is finding the simulations' output files, and the variables in them.
Each value becomes value × Multiply by + Add as it is read, so every analysis (and a water column's integral) is of the converted values. Multiply by rescales the units: 1000 turns mol/m3 into mmol/m3. Add shifts the zero point: -273.15 turns K into °C. Leave both boxes empty when the units are the same, only written another way. OceanVal's suggestions are in bold. The units must be confirmed, with the tick box beside Continue.
These simulation variables are time means, and their times look like the end of the period each mean covers: a day's mean dated 00:00 the next day, or a month's dated 00:00 on the 1st of the next month. If so, a January mean would count as February's, and the year's last mean would fall in the next year.
Their files have no time bounds to say for certain, so OceanVal does not shift anything unless you choose it. It suggests shifting the times of the simulations below back one day, in bold, so that each mean falls in the period it covers. The other simulations are not shifted. The dates of the files are read once you have chosen.
| Simulation and variable | Files | Times now | Shifted | Why they look like ends |
|---|
Looking at the files' times
OceanVal is checking whether the times of the files are the start or the end of the period each mean covers.
The simulation output with the highest temporal resolution has been identified and is listed below. These files will be used for analysis. The simulations are compared over the years they all have.
The simulations can be on different model grids. To compare them cell by cell, and to take means over exactly the same area, they need to be put on one grid.
Cells with no value in any one simulation are made missing in all of them. Once the simulations are on one grid, a cell that is missing in one simulation (land, say, or a place it has no water) is set to missing in every simulation, so that all of them are compared over exactly the same cells. This is judged from the first time step of each simulation's output, for each variable and each layer.
A lon/lat box and resolution
Every simulation is put on a regular lon-lat grid of this box and resolution, with nctoolkit's to_latlon, by the method chosen below.
The grid of a baseline simulation
The baseline's grid is the one the comparisons are made on. Each of the other simulations is regridded to it, with nctoolkit's regrid, by the method chosen below. The baseline itself is not regridded.
These simulation variables are time means, and their times look like the end of the period each mean covers: a day's mean dated 00:00 the next day, or a month's dated 00:00 on the 1st of the next month. If so, matched up as they are, a January mean is compared with February's observations, and the year's last mean falls in the next year and is left out.
Their files have no time bounds to say for certain, so OceanVal does not shift anything unless you choose it. It suggests shifting every simulation time back one day, in bold, so that each mean falls in the period it covers.
| Variable | Files | Times now | Shifted | Why they look like ends |
|---|
The following mapping will be assumed between variables and simulation files
| Variable | Years | Simulation variable | Observation source | File pattern to be used | Time resolution | Dates |
|---|
Not the right files?
Change which file names are used. OceanVal will read the simulation again with them and return to the recipes, to check the simulation variables, keeping your choices where they still apply.
A question for you
Earlier output is not shown.
Your matchup script is saved
Nothing has been matched up and no report has been built. To do both later, open a terminal, activate the environment OceanVal is installed in, and run:
Script file: