Home |
Search |
Today's Posts |
![]() |
|
uk.sci.weather (UK Weather) (uk.sci.weather) For the discussion of daily weather events, chiefly affecting the UK and adjacent parts of Europe, both past and predicted. The discussion is open to all, but contributions on a practical scientific level are encouraged. |
Reply |
|
|
LinkBack | Thread Tools | Display Modes |
#1
![]() |
|||
|
|||
![]()
I know this forum is populated by various people with meteorology
backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes) http://physics.open.ac.uk/~lsteele/ |
#2
![]() |
|||
|
|||
![]()
On Feb 9, 9:28*pm, Liam Steele wrote:
I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Why not treat the assimilated version as a forecast? You can construct various forecast scores averaged over spatial zones. We often use ROCs but it depends what you want to test. Anything you might do with a time series i.e. sequence of forecasts you can use for a single event over space. You could use this to test the improvement between different assimilation methods and different assimilation starting points. I have no direct experience of working in assimilation schemes but we look for skill in much that we do. There is a good page on verification here. http://www.cawcr.gov.au/projects/verification/ Andy |
#3
![]() |
|||
|
|||
![]()
On Feb 9, 9:28*pm, Liam Steele wrote:
I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ You mention a lat-long plot of the observations as the base to compare the assimilated figures with but surely observations are not made on a lat-long grid but at irregularly spaced points. If one of these observations is grossly in error it can be weeded out before attempting the fit but to do so would require the meteorological expertise acquired through experience. Alternatively, all the observations could be used, whether suspect or not and if the error at one point exceeds the rms error of the entire grid by some specified amount it could be removed at that stage though that would make the comparison invalid though it may well be more use operationally. I can't see any way round using the rms error even if a rogue point or two spoils it. December 2010 "spoilt" the (fairly) smooth winter warming of the last few decades but it happened and can't be ignored. Tudor Hughes, Warlingham, Surrey |
#4
![]() |
|||
|
|||
![]()
On 09/02/12 21:42, Andy M. wrote:
On Feb 9, 9:28 pm, Liam wrote: I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Why not treat the assimilated version as a forecast? You can construct various forecast scores averaged over spatial zones. We often use ROCs but it depends what you want to test. Anything you might do with a time series i.e. sequence of forecasts you can use for a single event over space. You could use this to test the improvement between different assimilation methods and different assimilation starting points. I have no direct experience of working in assimilation schemes but we look for skill in much that we do. There is a good page on verification here. http://www.cawcr.gov.au/projects/verification/ Andy Thanks Andy, that site is very informative. Something like the Displacement and Amplitude Score is what I would like, which compares the magnitude of the model predictions with the spatial miss-match as well. However, actually getting and implementing the code will be something else entirely! -- Liam (Milton Keynes) http://physics.open.ac.uk/~lsteele/ |
#5
![]() |
|||
|
|||
![]()
On 09/02/12 22:26, Tudor Hughes wrote:
On Feb 9, 9:28 pm, Liam wrote: I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ You mention a lat-long plot of the observations as the base to compare the assimilated figures with but surely observations are not made on a lat-long grid but at irregularly spaced points. If one of these observations is grossly in error it can be weeded out before attempting the fit but to do so would require the meteorological expertise acquired through experience. Alternatively, all the observations could be used, whether suspect or not and if the error at one point exceeds the rms error of the entire grid by some specified amount it could be removed at that stage though that would make the comparison invalid though it may well be more use operationally. I can't see any way round using the rms error even if a rogue point or two spoils it. December 2010 "spoilt" the (fairly) smooth winter warming of the last few decades but it happened and can't be ignored. Tudor Hughes, Warlingham, Surrey Hi Tudor. Yes, I thought about restricting the domain that is used for quantifying the analysis, but that seems a bit sneaky, as though I don't trust the model results! What seems to be happening is that the model predicts broadly the right features (e.g. clouds) but these might be shifted slightly so they are in an adjacent grid box to the observations. Thus, when calculating the rms, the model is penalised twice: once in the grid box with too much cloud and again in the adjacent grid box with too little cloud. What I would ideally like is something that can compare magnitudes as well as the spatial offset of model predictions. A look on the link Andy posted suggests I need something like the Displacement and Amplitude Score. How I would implement something like this is not clear though. I imagine it's not simple! -- Liam (Milton Keynes) http://physics.open.ac.uk/~lsteele/ |
#6
![]() |
|||
|
|||
![]()
On 09/02/12 21:42, Andy M. wrote:
On Feb 9, 9:28 pm, Liam wrote: I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Why not treat the assimilated version as a forecast? You can construct various forecast scores averaged over spatial zones. We often use ROCs but it depends what you want to test. Anything you might do with a time series i.e. sequence of forecasts you can use for a single event over space. You could use this to test the improvement between different assimilation methods and different assimilation starting points. I have no direct experience of working in assimilation schemes but we look for skill in much that we do. There is a good page on verification here. http://www.cawcr.gov.au/projects/verification/ Andy Further reading of the web page has brought up the CRA verification method, which compares spatial details as well as magnitudes. And, it even has some IDL code supplied! Looking good... http://www.cawcr.gov.au/projects/ver...ification.html -- Liam (Milton Keynes) http://physics.open.ac.uk/~lsteele/ |
#7
![]() |
|||
|
|||
![]()
On Feb 9, 11:34*pm, Liam Steele wrote:
On 09/02/12 21:42, Andy M. wrote: On Feb 9, 9:28 pm, Liam *wrote: I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Why not treat the assimilated version as a forecast? *You can construct various forecast scores averaged over spatial zones. *We often use ROCs but it depends what you want to test. *Anything you might do with a time series i.e. *sequence of forecasts you can use for a single event over space. You could use this to test the improvement between different assimilation methods and different assimilation starting points. *I have no direct experience of working in assimilation schemes but we look for skill in much that we do. There is a good page on verification here. http://www.cawcr.gov.au/projects/verification/ Andy Further reading of the web page has brought up the CRA verification method, which compares spatial details as well as magnitudes. And, it even has some IDL code supplied! Looking good... http://www.cawcr.gov.au/projects/ver...ification.html -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Yes it is a good site .., I have not used those scores but having some idea of the spatial coherence and the departure from the target is good ... might look myself. You should also think about getting hold of a copy of Jolliffe and Stephenson http://books.google.co.uk/books/abou...d=Qm2MjWVvUywC I am told by Dave that a new edition is on its way. What model are you building the assimilation system for? Good Luck, Andy |
#8
![]() |
|||
|
|||
![]()
On Feb 10, 7:47*am, "Andy M." wrote:
On Feb 9, 11:34*pm, Liam Steele wrote: On 09/02/12 21:42, Andy M. wrote: On Feb 9, 9:28 pm, Liam *wrote: I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Why not treat the assimilated version as a forecast? *You can construct various forecast scores averaged over spatial zones. *We often use ROCs but it depends what you want to test. *Anything you might do with a time series i.e. *sequence of forecasts you can use for a single event over space. You could use this to test the improvement between different assimilation methods and different assimilation starting points. *I have no direct experience of working in assimilation schemes but we look for skill in much that we do. There is a good page on verification here. http://www.cawcr.gov.au/projects/verification/ Andy Further reading of the web page has brought up the CRA verification method, which compares spatial details as well as magnitudes. And, it even has some IDL code supplied! Looking good... http://www.cawcr.gov.au/projects/ver...ification.html -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Yes it is a good site .., I have not used those scores but having some *idea of the spatial coherence and the departure from the target is good ... might look myself. You should also think about getting hold of a copy of Jolliffe and Stephensonhttp://books.google.co.uk/books/about/Forecast_verification.html?id=Q... I am told by Dave that a new edition is on its way. What model are you building the assimilation system for? Good Luck, Andy I am working on a model of the martian atmosphere, and we are currently using the 'Analysis Correction' assimilation scheme that they used at the Met Office in the early 90s. There are more modern assimilation schemes, but apparently they have difficulty when only assimilating a few observations (and for Mars we only have one sounder that gives us observations). I'm looking in particular at water vapour and ice data, and trying to get the model's cloud prediction as accurate as possible so we can work out what role they play in Mars' climate. As an example, I've put three plots up on my website: http://www.physics.open.ac.uk/~lsteele/assimilation.png. The top panel is observation data, the middle panel is an assimilation and the bottom panel is a standard model run. To my eye the assimilation resembles the observations more than the standard model, but it actually has a larger rms error. Hence I was looking for something that weights things spatially as well as simply by magnitude. Liam |
#9
![]() |
|||
|
|||
![]()
Liam Steele wrote:
How I would implement something like this is not clear though. I imagine it's not simple! ----------------------------------- Something about me is glad it's still not "simple" to get a PhD ;-) I can't help but it sounds interesting, good luck with it Liam, Dave |
#10
![]() |
|||
|
|||
![]()
On Feb 10, 9:07*am, Sir Loin Steak wrote:
On Feb 10, 7:47*am, "Andy M." wrote: On Feb 9, 11:34*pm, Liam Steele wrote: On 09/02/12 21:42, Andy M. wrote: On Feb 9, 9:28 pm, Liam *wrote: I know this forum is populated by various people with meteorology backgrounds, so thought I'd ask here first... For my PhD I am assimilating data into a general circulation model.. Now what I want to do is compare the assimilation result (a simple lat-lon plot of whatever) with a lat-lon plot of the observations. I can simply subtract one from the other to get a difference plot, but I was wondering if there was a statistical way to compare the 'fit' between two 2D plots? The model calculates the average rms error over all lat-lon points, but this isn't very helpful, as one particularly bad gridpoint can skew the whole result. If you can give any helpful advice I shall endeavor to mention you in the acknowledgements of my thesis. ![]() Many thanks, -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Why not treat the assimilated version as a forecast? *You can construct various forecast scores averaged over spatial zones. *We often use ROCs but it depends what you want to test. *Anything you might do with a time series i.e. *sequence of forecasts you can use for a single event over space. You could use this to test the improvement between different assimilation methods and different assimilation starting points. *I have no direct experience of working in assimilation schemes but we look for skill in much that we do. There is a good page on verification here. http://www.cawcr.gov.au/projects/verification/ Andy Further reading of the web page has brought up the CRA verification method, which compares spatial details as well as magnitudes. And, it even has some IDL code supplied! Looking good... http://www.cawcr.gov.au/projects/ver...ification.html -- Liam (Milton Keynes)http://physics.open.ac.uk/~lsteele/ Yes it is a good site .., I have not used those scores but having some *idea of the spatial coherence and the departure from the target is good ... might look myself. You should also think about getting hold of a copy of Jolliffe and Stephensonhttp://books.google.co.uk/books/about/Forecast_verification.html?id=Q... I am told by Dave that a new edition is on its way. What model are you building the assimilation system for? Good Luck, Andy I am working on a model of the martian atmosphere, and we are currently using the 'Analysis Correction' assimilation scheme that they used at the Met Office in the early 90s. There are more modern assimilation schemes, but apparently they have difficulty when only assimilating a few observations (and for Mars we only have one sounder that gives us observations). I'm looking in particular at water vapour and ice data, and trying to get the model's cloud prediction as accurate as possible so we can work out what role they play in Mars' climate. As an example, I've put three plots up on my website:http://www.physics.open.ac.uk/~lsteele/assimilation.png. The top panel is observation data, the middle panel is an assimilation and the bottom panel is a standard model run. To my eye the assimilation resembles the observations more than the standard model, but it actually has a larger rms error. Hence I was looking for something that weights things spatially as well as simply by magnitude. Liam Thanks for this If you ever get chance I would love to see a surface energy budget (for somewhere) from the simulation for summer time. I am currently teaching this with my second years for the earth of course. I once set an exam question based around a radiation budget for the Martian surface with quite a few hints and values given. No one tried it so I did not put it on ever again. I must have had data from one of the landers in the late 1990s If you have the 4 way radiation streams (simulated) at the surface (or near surface) that would be great if you have the heat fluxes as well even better. Thanks, Andy google Andy Morse Liverpool if you want to contact me at work. |
Reply |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|
![]() |
||||
Thread | Forum | |||
The Most Important Single Question Before Us! — No Question At All | sci.geo.meteorology (Meteorology) | |||
SERIOUS question about CO2 ( Sincere Question. Please Help if you can) | sci.geo.meteorology (Meteorology) | |||
metcheck says sorry.. well done | uk.sci.weather (UK Weather) | |||
Sorry started new post! - See Contrails | uk.sci.weather (UK Weather) | |||
I'm Sorry I Haven't a Clue | uk.sci.weather (UK Weather) |