Case description

In this exercise we will use Metview to explore the various ways ensemble forecast forecasts can be processed and visualised. The case we will investigate is related to a low pressure system crossing the UK on 10 August 2014: it caused high winds and heavy rain especially in the South-Western part of the country.

...

XXX Download data

Verify that the data are as expected.

Remember to give your icons useful names!

Setting the map View

With a new Geographical View icon, set up a cylindrical projection with its area defined as

South/West/North/East: 43/-20/60/10

Set up a new Coastlines icon with the following:

• the land coloured in cream
• the coastlines thick black
• use grey grid lines at every 5 degrees.

...

Evaluating the forecasts

In this part of the exercise we will create the plot shown below:

...

In this plot each map contains the various forecasts for the maximum 10 m wind gust in a 6-hour period from various model runs:

• top left: the closest latest forecast to before the event
• top right: the forecast run 4 days before the event
• bottom left: the ensemble mean for the ENS  forecast run 4 days before the event
• bottom right: the ensemble spread for the ENS forecasts run 4 days before the event

Setting the View

With a new Geographical View icon, set up a cylindrical projection with its area defined as

South/West/North/East: 43/-20/60/10

Set up a new Coastlines icon with the following:

• the land coloured in cream
• the coastlines thick black
• use grey grid lines at every 5 degrees.

Defining the layout

With a new Display Window icon define a 2x2 layout so that each plot should contain your view. 

Visualising the latest forecast

The GRIB file fc_latest_oper.grib contains the latest forecast run preceding the event. Drag it into the top left map and customise it with the wgust_shade Contouring icon and the title_oper Text Plotting icon. Animate through the fields to see the location of the areas heavily hit by the storm.

Visualising the operational forecast

The GRIB file fc_oper.grib contains the operational forecast run at 7 August 00 TC UTC (four 4 days before the event). Drag it into the top right map and customise it in the same way as the previous plot. You will see that the wind storm was not present in the forecast.

Visualising the ensemble mean

The GRIB file fc_ens.grib contains the control forecast member and the 50 perturbed members of the ENS run at 7 August 00 TC UTC (four 4 days before the event). We will compute and visualise the mean of the ensemble members using a Macro.

The computations will be done in Macro, so create Create a new Macro and edit it. First, read the GRIB file in:

...

Our GRIB contains three time steps (78, 84 , and 90 and 96 hours, respectively) and we would like to compute the ensemble mean for each one separately. To achieve this goal we will write a loop going through the time steps. First, define the fieldset that will contain the results

...

Next, add this piece of code to define the loop (we store the time steps in a list):

tsLst=[78,84,90,96]

loop step in tsLst

    ...your code will go here ...

end loop

...

e_mean = e_mean & f

By doing so the loop's body is completed. We finish the macro by returning the resulting fieldset:

return e_mean

Drag it your Macro into the bottom left map and customise it with the wgust_shade Contouring icon. You would also need a custom Text Plotting icon for the title. Take a copy of the one used for the previous plots (called title_oper) and tailor it to your needs. When you analyse the plot you will notice that the ensemble mean hints for higher wind gusts in the our area of interest.

Visualising the ensemble spread

The ensemble spread is the standard deviation of the ENS members. You We can compute it in a very similar way to the ensemble mean. The only difference is that this time you we need to use the stdev() function instead of mean(). Now it is your task to write a Macro for it. Once you finished your Macro drag it into the bottom right map and customise it with the wgust_spread_shade Contouring icon and with a custom Text Plotting icon. You will see that the ensemble spread is fairly high in the investigated area.

...

Checking the probabilities

In this part we will estimate the risk of the wind gust being higher than a certain threshold. We will compute the probability of the wind gust exceeding 22 m/s (≈ about 80 kkm/mh) and generate the plot shown below:

You can We will compute the probabilities with a Macro in a very similar Macro to the ones you wrote way as we did for the ensemble mean (and standard deviation). The difference is that this time you we need to compute a probability for each time step. Now it is your task to write a Macro and compute the probabilities. The computation should go like this

Now duplicate the ensemble mean Macro and edit it. Find the code line computing the mean and replace it with this code block:

f=f > threshold22	
f=100*mean(f)

Here we supposed that all the ENS members are stored in fieldset f and the threshold value in variable threshold. The first line in the code above, performs a logical operation on the fieldset and results in a new fieldset having 1s . In this new fieldset we have only 1s and 0s:

• the value is 1 in each gridpoint where the condition meets (i.e. the value is larger than the threshold)

...

• the value is 0 in all other gridpoints.

The second line simply derives Then the probability is simply as the mean of these fields. We multiply the result by 100 to scale it into the 0-100 range for an easier interpretation.

Once you finished your Macro drag it into the bottom right map and customise , visualise your Geographical View icon and drag the Macro into the  plot. Customise it with the prob_shade Contouring icon and with . Also use a custom Text Plotting icon . You will to define the title. As for the probabilities, you should see that there is indeed a reasonable probability hinting for a wind storm.

...

some probability of high wind speeds.

Creating a stamp plot

In this part we will investigate the individual ENS members and create a plot showing them all for a given time step on the same page like this:

...

This plot, for an obvious reason, is called a stamp plot. This is a complex plot so we will write a Macro to generate it.

Create a new Macro and edit it. Drop your Geographical View and the Coastlines icons into the Macro editor. Once you've tidied up the code is generated and tidied up , define a 6x9 layout so that each plot should contain your view:

dw=plot_super_pagesuperpage(pages: mxn_layout(my_view,9,6))

...

The stamp plot will be generated by plotting each perturbed forecast member into a separate map, so we need to write a loop like this:

for numberi=1 to 50 do 

    ...your code will go here ...

end loopfor

Within the loop, simply read the current perturbed forecast member for the given time step:

f=read(data: g,
		number: numberi,
		type: "pf",
		step: step
	  )	

...

title = mtext(text_line_1 : "PF: " & numberi)

Finish the loop by plotting the Last, plot the field into the right map in our layout:

plot(dw[i],title,f,wgust_shade)

Having done so run we have finished the code inside the loop. Now visualise your Macro (this will take a minute or so) and try to identify the ENS members predicting high wind speeds in our area.

...

Creating a spaghetti plot

We finish the case study by looking into the predictability of the large scale flow pattern by generating spaghetti plots for 500 hPa geopotential from the same ENS run as we investigated before.   In a spaghetti plot each ENS member is rendered into the same map using a single isoline value. The plot we want to generate is shown below (it contains the spaghetti plot for 500 hPa geopotential using the 560 gpm isoline value):

...

Create a new Macro and edit it. Drop your Geographical View and the Coastlines icons into the Macro editor . Once the code is generated and tidied up change the map area to

[40,-40,70,20]

...

Next, define the contouring used for the "spaghetti" by dropping the cont_spag Contouring icon into the Macro. A code like this should be generated for you:

cont_spag = mcont(
	contour_label: "off",
	contour_level_selection_type	:	"level_list",
	contour_level_list	:	560,
	contour_line_colour: "blue",
	contour_highlight: "off"
 )

In this mcont() we turned contour labels off so that to keep the plot uncluttered and defined only a single contour value (for 560 gpm). 

Continue with reading in the GRIB file of the ENS forecasts used for the "spaghetti":

 g = read("spag_ens.grib")

The "spaghetti" will be generated by plotting each perturbed forecasts member as a separate layer into the same map. To achieve this goal we need to write a loop like this:

for numberi=1 to 50 do 

    ...your code will go here ...

end loopfor

Within the loop, read all the perturbed forecast members for the all the time steps:

f=read(data: g,
	type: "pf", 
	number: numberi
 )	

Next, plot it into your view with your contour settings:

plot(your_view,f,cont) 

If you run this macro you will see that the "spaghetti" was properly generated but you have too many titles (one for each layer i.e. 50). To get rid of the extra titles you need to define a Text Plotting inside the loop and add it to your plot() commandBy default, if no title definition is specified, Metview adds a title line for each field in the plot. Since we are about to plot 50 fields into the same map this would result in 50 titles in the plot! To avoid having too many titles we use a custom Text Plotting icon:

title=mtext(text_line_1: "your custom title ....")

plot(your_view,f,cont,title)Value: 560 gpm T+<grib_info key='step' where='number=50' /> h" )

Here we used the As for the custom title's text, the idea is that you should use the where statement inside the grib_info tag (as described here)  to make the title appear for one member (the 50th member) only.

Last, plot the field with our contour settings and title:

plot(your_view,f,cont_spag,title) 

Having done so we have finished the code inside the loop. Now visualise your Macro (it will take half a minute or so) Once you sorted the the title out visualise your plot again and animate through the steps to see how the spaghetti is spreading out over time.

Extra Work

...

if You Have Time

Add more fields to the stamp plot

The stamp plot only shows the perturbed ENS members but there is still space left to display additional fields, as well. Try to add the control forecast (from ENS) and the operational forecast to it. HintsSome hints:

• plot the control forecast into the 51st map

  51

  (dw[51]). The control forecast is stored in the same file as the perturbed forecast members: fc_ens.grib.

  To access it (via the read() function) you need to omit the number parameter and set type to "cf".

  Read it in with this code:

  Code Block
  f = read(data: g, type: "cf", step: step)

• plot the operational forecast into the 52nd map

  52

  (dw[52]). The operational forecast is stored in fc_oper.grib.

  For the read() command you just need to specify the step.

...

• Read it in with this code:

  Code Block
  f =read(source: "fc_oper.grib", step: step)

NOTE: While setting up these extra plots it is a good idea to temporarily comment out the

...

loop processing the perturbed forecast members.

Add more fields to the spaghetti plot

The spaghetti plot only shows the perturbed ENS members. Try to add the control forecast (from ENS) and the operational forecast to it as well. You should use different isoline colours for itHintsthem. Some hints:

• plot the control forecast with

  use a thick

  green isoline

  red contour line. The control forecast is stored in the same file as the perturbed forecast members: spag_ens.grib.

  To access it (via the read() function) you need to omit the number parameter and set type to "cf".

  Read it in with this code:

  Code Block
  f = read(data: g, type: "cf")

• use thick green contour line

  plot the operational  forecast with thick red isoline

  . The operational forecast is stored in spag_oper.grib.

  You just need to read

  Read it in with

  the read() command. 

...

• this code:

  Code Block
  f = read("spag_oper.grib")

NOTE: While setting up these extra

...

plots it is a good idea to temporarily comment out the

...

loop processing the perturbed

...

forecast members.