Forums

Back

RE: Humidity (Dew Points) for Fire Weather

RSS (Opens New Window)(Opens New Window)
RL
Ryan Leach, modified 5 Years ago.

Humidity (Dew Points) for Fire Weather

Youngling Posts: 18 Join Date: 1/19/16 Recent Posts

One of our closest core partners is the wildland fire community, and of critical importance to them is wind and humidity. I know there have been other discussions about wind, so I'll just focus on humidity here.

 

We would like to use the NBM as our default guidance for everything, but we've found this prescribed fire season that it falls far short of the SuperBlend when it comes to humidity forecasting. I think the root cause of that is the URMA which the NBM uses for bias correction.  I compared the Obs grid to the URMA analysis at several stations in our area and found pretty big differences between the two analyses. 

 

The first example is KBTM, one of our ASOS sites: 

There's a couple of things here I'd like to point out. First, the error histogram (right) is wide, stretching out to ±10%. Second, I highlighted all the cases where either the Obs or URMA grids were below 40% humidity. This is the dry side where we could actually have prescribed fire; this is the range our fire customers care about the most. Notice in the lower RH area the URMA has a wet bias compared to the Obs grid.  This is also reflected on the MinRH plot below.

MinRH is a part of every Spot forecast we put out, but it is also available through NDFD, so burners have access to it through other means as well, so if we are constantly adjusting it down for the Spots we are sending a mixed message. This is exactly what has been happening this spring.

 

Another case is the Lodge Pole RAWS (5,814 ft elevation).

 

In this case the wet bias of the URMA in low humidity is even bigger. This is generally the case for many RAWS stations in our area. I assume the dew point/ RH from the ASOS's weighted heavier in the analysis hence the better performance at Butte. There are some locations in the low elevations of central Idaho where URMA actually has a dry bias, but that hasn't been as much of an impact because most of our fire weather customers have been in Montana so far this year, but that could also be a problem.

 

I assume the URMA is analyzing dew points and then calculating the humidity from that. I know that's how it is run in our SmartInits for GFE. I know a lot of work has been done to improve the temperature forecasting in the mountains, and I'm hoping some of those changes can easily be translated to dew points so we could see an improvement in the RH analysis/forecasts from URMA/NBM.

 

Our desire is to have a good humidity forecast because that is what is important to our users, and it is what most sensors directly measure (even if they report a dew point like ASOS). So if improved dew points are the way to get there I'm all for it, but if there is another way to get a good humidity analysis and forecast that works just as well for us too.

For now, I'm going to recommend that the office go back to using SuperBlend as the starting for our humidity grids.

 

Thanks,

 

Ryan Leach,

Senior Meteorologist,  IMET, and GFE focal point

WFO Missoula

Attachments:

Attachment

Attachment

Attachment

SL
Steven Levine, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Youngling Posts: 174 Join Date: 11/13/14 Recent Posts
Hi Ryan,

Thanks for putting this together.  Clearly, we've got a problem out there.  One question: Over how long/what period did you take these stats?

For moisture, the RTMA analyzes specific humidity then combines it with pressure and temperature to create a dew point grid; it is logistically much easier to work with an independent variable like specific humidity rather than dew point or RH.  The daily min/max RH from URMA is simply the min or max RH at a given grid point, calculated from the temp/moisture/pressure grids over 06-18Z.  

A relevant point to make here is that for any observation that reports more than once per hour, the URMA grid will only use the ob closest to the top of the hour.  While the min/max T grid will account for extra intra-hour obs, the min/max moisture grids do not.

So my first point of curiosity is whether this RH pattern extends to temperature, specifically whether we are getting too cool in the afternoon and/or with maxT (this is the sort of thing we can check here).  If not, the next place to look would be the moisture background and obs.

Steve

On Sun, May 5, 2019 at 2:55 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

One of our closest core partners is the wildland fire community, and of critical importance to them is wind and humidity. I know there have been other discussions about wind, so I'll just focus on humidity here.

 

We would like to use the NBM as our default guidance for everything, but we've found this prescribed fire season that it falls far short of the SuperBlend when it comes to humidity forecasting. I think the root cause of that is the URMA which the NBM uses for bias correction.  I compared the Obs grid to the URMA analysis at several stations in our area and found pretty big differences between the two analyses. 

 

The first example is KBTM, one of our ASOS sites: 

There's a couple of things here I'd like to point out. First, the error histogram (right) is wide, stretching out to ±10%. Second, I highlighted all the cases where either the Obs or URMA grids were below 40% humidity. This is the dry side where we could actually have prescribed fire; this is the range our fire customers care about the most. Notice in the lower RH area the URMA has a wet bias compared to the Obs grid.  This is also reflected on the MinRH plot below.

MinRH is a part of every Spot forecast we put out, but it is also available through NDFD, so burners have access to it through other means as well, so if we are constantly adjusting it down for the Spots we are sending a mixed message. This is exactly what has been happening this spring.

 

Another case is the Lodge Pole RAWS (5,814 ft elevation).

 

In this case the wet bias of the URMA in low humidity is even bigger. This is generally the case for many RAWS stations in our area. I assume the dew point/ RH from the ASOS's weighted heavier in the analysis hence the better performance at Butte. There are some locations in the low elevations of central Idaho where URMA actually has a dry bias, but that hasn't been as much of an impact because most of our fire weather customers have been in Montana so far this year, but that could also be a problem.

 

I assume the URMA is analyzing dew points and then calculating the humidity from that. I know that's how it is run in our SmartInits for GFE. I know a lot of work has been done to improve the temperature forecasting in the mountains, and I'm hoping some of those changes can easily be translated to dew points so we could see an improvement in the RH analysis/forecasts from URMA/NBM.

 

Our desire is to have a good humidity forecast because that is what is important to our users, and it is what most sensors directly measure (even if they report a dew point like ASOS). So if improved dew points are the way to get there I'm all for it, but if there is another way to get a good humidity analysis and forecast that works just as well for us too.

For now, I'm going to recommend that the office go back to using SuperBlend as the starting for our humidity grids.

 

Thanks,

 

Ryan Leach,

Senior Meteorologist,  IMET, and GFE focal point

WFO Missoula


--
Ryan Leach RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6537484VLab.Notifications@noaa.gov

Attachments:

Attachment

Attachment

Attachment

RL
Ryan Leach, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Youngling Posts: 18 Join Date: 1/19/16 Recent Posts
Steve,

Sorry, I meant to put it in my original post. This is data since April 15th. My intention was to cover the period of time we have been doing spot forecasts.

While the timescales may affect the MinRH, the hourly RH can be important as well since we sometimes get spot requests with an hourly or 3-hourly breakdown of weather elements. A question we often get is, "When will the RH drop below X?". The burners need this information to know when they can ignite.

We're also getting more requests for hourly out put and spots in the "Tabular" format, so the hourly RH is important too.

The temperatures in URMA compare much better to the Obs than the humidity does, and I haven't noticed a significant bias in the temperature fields. I'm confident most of the error is from the moisture fields. Unfortunately, we don't archive Obs/URMA grids for dew point because we would run out of space in /data/verify.

Ryan N. Leach
Senior Meteorologist, IMET
National Weather Service, Missoula
(406)329-4840


On Mon, May 6, 2019 at 7:07 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Hi Ryan,

Thanks for putting this together.  Clearly, we've got a problem out there.  One question: Over how long/what period did you take these stats?

For moisture, the RTMA analyzes specific humidity then combines it with pressure and temperature to create a dew point grid; it is logistically much easier to work with an independent variable like specific humidity rather than dew point or RH.  The daily min/max RH from URMA is simply the min or max RH at a given grid point, calculated from the temp/moisture/pressure grids over 06-18Z.  

A relevant point to make here is that for any observation that reports more than once per hour, the URMA grid will only use the ob closest to the top of the hour.  While the min/max T grid will account for extra intra-hour obs, the min/max moisture grids do not.

So my first point of curiosity is whether this RH pattern extends to temperature, specifically whether we are getting too cool in the afternoon and/or with maxT (this is the sort of thing we can check here).  If not, the next place to look would be the moisture background and obs.

Steve

On Sun, May 5, 2019 at 2:55 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

One of our closest core partners is the wildland fire community, and of critical importance to them is wind and humidity. I know there have been other discussions about wind, so I'll just focus on humidity here.

 

We would like to use the NBM as our default guidance for everything, but we've found this prescribed fire season that it falls far short of the SuperBlend when it comes to humidity forecasting. I think the root cause of that is the URMA which the NBM uses for bias correction.  I compared the Obs grid to the URMA analysis at several stations in our area and found pretty big differences between the two analyses. 

 

The first example is KBTM, one of our ASOS sites: 

There's a couple of things here I'd like to point out. First, the error histogram (right) is wide, stretching out to ±10%. Second, I highlighted all the cases where either the Obs or URMA grids were below 40% humidity. This is the dry side where we could actually have prescribed fire; this is the range our fire customers care about the most. Notice in the lower RH area the URMA has a wet bias compared to the Obs grid.  This is also reflected on the MinRH plot below.

MinRH is a part of every Spot forecast we put out, but it is also available through NDFD, so burners have access to it through other means as well, so if we are constantly adjusting it down for the Spots we are sending a mixed message. This is exactly what has been happening this spring.

 

Another case is the Lodge Pole RAWS (5,814 ft elevation).

 

In this case the wet bias of the URMA in low humidity is even bigger. This is generally the case for many RAWS stations in our area. I assume the dew point/ RH from the ASOS's weighted heavier in the analysis hence the better performance at Butte. There are some locations in the low elevations of central Idaho where URMA actually has a dry bias, but that hasn't been as much of an impact because most of our fire weather customers have been in Montana so far this year, but that could also be a problem.

 

I assume the URMA is analyzing dew points and then calculating the humidity from that. I know that's how it is run in our SmartInits for GFE. I know a lot of work has been done to improve the temperature forecasting in the mountains, and I'm hoping some of those changes can easily be translated to dew points so we could see an improvement in the RH analysis/forecasts from URMA/NBM.

 

Our desire is to have a good humidity forecast because that is what is important to our users, and it is what most sensors directly measure (even if they report a dew point like ASOS). So if improved dew points are the way to get there I'm all for it, but if there is another way to get a good humidity analysis and forecast that works just as well for us too.

For now, I'm going to recommend that the office go back to using SuperBlend as the starting for our humidity grids.

 

Thanks,

 

Ryan Leach,

Senior Meteorologist,  IMET, and GFE focal point

WFO Missoula


--
Ryan Leach RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6537484VLab.Notifications@noaa.gov

--
Steven Levine RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6542606VLab.Notifications@noaa.gov

Attachments:

Attachment

Attachment

Attachment

RL
Ryan Leach, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Youngling Posts: 18 Join Date: 1/19/16 Recent Posts

Red is the cases where either Obs or URMA were at or below 40% RH, I made reliability plots and histograms for T, Td (calculated), and RH. Those graphics are below, but my general observations are that the overall range of the differences of temperature are smaller than for dew point and not as strongly biased. There are sites where the bias is different, but most show a bias similar to these.

My main concern is whether or not we can get the URMA Td/RH closer to the observations, similar to the way temperatures were improved.

I think it is interesting that pressure is used in the moisture analysis, since the stations don't report pressure from the RAWS network (or DOT) in our area. Where do you get the pressure to turn the measured T/Td/RH into a specific humidity for analysis?   Are there corrections made for differences in the grid point elevation and the actual station elevation?

 

Attachments:

Attachment

Attachment

SL
Steven Levine, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Youngling Posts: 174 Join Date: 11/13/14 Recent Posts
So we're clearly looking at a moisture issue here, not a temperature issue.  That gives us a good place to start.

The assigned pressure for an ob (if needed) comes from it's reported elevation; this happens before anything happens with any grid.

The moisture/pressure issue with the grids is a bit more complicated and is better described in the context of a case.  I'll try to do that when I have time to look at this one.

Steve

On Thu, May 9, 2019 at 7:24 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Red is the cases where either Obs or URMA were at or below 40% RH, I made reliability plots and histograms for T, Td (calculated), and RH. Those graphics are below, but my general observations are that the overall range of the differences of temperature are smaller than for dew point and not as strongly biased. There are sites where the bias is different, but most show a bias similar to these.

My main concern is whether or not we can get the URMA Td/RH closer to the observations, similar to the way temperatures were improved.

I think it is interesting that pressure is used in the moisture analysis, since the stations don't report pressure from the RAWS network (or DOT) in our area. Where do you get the pressure to turn the measured T/Td/RH into a specific humidity for analysis?   Are there corrections made for differences in the grid point elevation and the actual station elevation?

 


--
Ryan Leach RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6568762 VLab.Notifications@noaa.gov
U
Anonymous, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Steve,

 

Is there any way we could get the equation used for the RH? Assuming RAWS observations are getting into the URMA I think a manual calculation at a few locations to see what is produced may help. I know coming up with a pressure for a site based on elevation can be problematic so that may be part of it.

 

Mark

SL
Steven Levine, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Youngling Posts: 174 Join Date: 11/13/14 Recent Posts
When a RAWS observation gets here, it only has a dew point.  RAWS stations typically do not include any sort of pressure information, so it is derived via reported temperature and elevation according to standard atmosphere.  The formula for this estimated pressure is:

PR(PMSL,TEMP,ELEV) =PMSL * (((TEMP - (.0065 * ELEV))/TEMP)**5.256)

When PMSL is not available (usually the case with RAWS), it is assumed to be 1013.25 mb.  It should also be mentioned that this calculation takes place before the background is introduced, so this is before anything would be tossed by the gross error check or something like that.

From this pressure, specific humidity is calculated from the reported dew point:

ES(T) = 6.1078 * EXP((17.269 * (T - 273.16))/((T - 273.16)+237.3))   !saturation vapor pressure
QFRMTP(T,P) = (0.622 * ES(T))/(P - (0.378 * ES(T)))      !specific humidity

Specific humidity is pulled from the background field, too, but using the background pressure at that grid point.  Specific humidity is then analyzed independently, much like temperature and pressure.

Now, it's worth noting that the RTMA/URMA analyzes surface pressure too.  However the derived pressure ob from a RAWS ob would not be used in the analysis, while the moisture ob would be.  So if the assumed/derived pressure is off, that could have a significant impact on the derived dew point that would not show up in the temperature or pressure analysis.

After the variables are analyzed, a processing step converts the temperature, specific humidity and pressure grids to create a dew point grid using the attached fortran subroutine.

I believe any hourly RH grid you might see is produced within AWIPS, and I don't have access to the code for that.

Steve

On Wed, May 15, 2019 at 10:38 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Steve,

 

Is there any way we could get the equation used for the RH? Assuming RAWS observations are getting into the URMA I think a manual calculation at a few locations to see what is produced may help. I know coming up with a pressure for a site based on elevation can be problematic so that may be part of it.

 

Mark


--
Anonymous RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6613635VLab.Notifications@noaa.gov

Attachments:

RL
Ryan Leach, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Youngling Posts: 18 Join Date: 1/19/16 Recent Posts
Steve,

Yes, I can confirm the RH is being calculated in AWIPS (at least for GFE).

It seems like there is a LOT of approximations going on in this process, which raises a couple of questions.

1) Is there somewhere I can see which Td observations are being thrown out? I'm curious how much of the RAWS data is getting in there at all. It seems the moisture field analysis was designed specifically for observations with pressure (e.g. ASOS) and the multitude of other observations in mind. Out west I suspect a large portion of observations don't have pressure.

2) Is it feasible to analyze Td instead of Specific Humidity? Td (in the terrain) behaves similarly to T. Pooling in valleys, thermal belts, etc. If this is possible it might be easy to then take the efforts applied to T and also directly apply them to Td with relatively little effort.

Thanks,
Ryan N. Leach
Senior Meteorologist, IMET
National Weather Service, Missoula
(406)329-4840


On Tue, May 21, 2019 at 9:22 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
When a RAWS observation gets here, it only has a dew point.  RAWS stations typically do not include any sort of pressure information, so it is derived via reported temperature and elevation according to standard atmosphere.  The formula for this estimated pressure is:

PR(PMSL,TEMP,ELEV) =PMSL * (((TEMP - (.0065 * ELEV))/TEMP)**5.256)

When PMSL is not available (usually the case with RAWS), it is assumed to be 1013.25 mb.  It should also be mentioned that this calculation takes place before the background is introduced, so this is before anything would be tossed by the gross error check or something like that.

From this pressure, specific humidity is calculated from the reported dew point:

ES(T) = 6.1078 * EXP((17.269 * (T - 273.16))/((T - 273.16)+237.3))   !saturation vapor pressure
QFRMTP(T,P) = (0.622 * ES(T))/(P - (0.378 * ES(T)))      !specific humidity

Specific humidity is pulled from the background field, too, but using the background pressure at that grid point.  Specific humidity is then analyzed independently, much like temperature and pressure.

Now, it's worth noting that the RTMA/URMA analyzes surface pressure too.  However the derived pressure ob from a RAWS ob would not be used in the analysis, while the moisture ob would be.  So if the assumed/derived pressure is off, that could have a significant impact on the derived dew point that would not show up in the temperature or pressure analysis.

After the variables are analyzed, a processing step converts the temperature, specific humidity and pressure grids to create a dew point grid using the attached fortran subroutine.

I believe any hourly RH grid you might see is produced within AWIPS, and I don't have access to the code for that.

Steve

On Wed, May 15, 2019 at 10:38 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Steve,

 

Is there any way we could get the equation used for the RH? Assuming RAWS observations are getting into the URMA I think a manual calculation at a few locations to see what is produced may help. I know coming up with a pressure for a site based on elevation can be problematic so that may be part of it.

 

Mark


--
Anonymous RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6613635VLab.Notifications@noaa.gov

--
Steven Levine RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6664453 VLab.Notifications@noaa.gov
SL
Steven Levine, modified 5 Years ago.

RE: Humidity (Dew Points) for Fire Weather

Youngling Posts: 174 Join Date: 11/13/14 Recent Posts
Ryan and all,

We make KML files of moisture obs used/not used and keep those available here.  Just look for the files ending in _moist_nr.kml.  They work just like the temperature and wind plots many of you are already familiar with.  The only difference is that the ob/background fields will contain specific humidity values rather than dew points.  I do have similar files that convert to dew point, I will start uploading those as well.

We can easily go back and make these for April 15th as well.

As for the straight dew point issue: that's been discussed but never implemented.  At this point, the plate for our next upgrade is pretty full, but we will look to see what we can do.

Hopefully the upcoming 3D system (still 2 or so years away) will be better able to accommodate this dew point issue.

Steve

On Tue, May 21, 2019 at 4:02 PM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
Steve,

Yes, I can confirm the RH is being calculated in AWIPS (at least for GFE).

It seems like there is a LOT of approximations going on in this process, which raises a couple of questions.

1) Is there somewhere I can see which Td observations are being thrown out? I'm curious how much of the RAWS data is getting in there at all. It seems the moisture field analysis was designed specifically for observations with pressure (e.g. ASOS) and the multitude of other observations in mind. Out west I suspect a large portion of observations don't have pressure.

2) Is it feasible to analyze Td instead of Specific Humidity? Td (in the terrain) behaves similarly to T. Pooling in valleys, thermal belts, etc. If this is possible it might be easy to then take the efforts applied to T and also directly apply them to Td with relatively little effort.

Thanks,
Ryan N. Leach
Senior Meteorologist, IMET
National Weather Service, Missoula
(406)329-4840


On Tue, May 21, 2019 at 9:22 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:
When a RAWS observation gets here, it only has a dew point.  RAWS stations typically do not include any sort of pressure information, so it is derived via reported temperature and elevation according to standard atmosphere.  The formula for this estimated pressure is:

PR(PMSL,TEMP,ELEV) =PMSL * (((TEMP - (.0065 * ELEV))/TEMP)**5.256)

When PMSL is not available (usually the case with RAWS), it is assumed to be 1013.25 mb.  It should also be mentioned that this calculation takes place before the background is introduced, so this is before anything would be tossed by the gross error check or something like that.

From this pressure, specific humidity is calculated from the reported dew point:

ES(T) = 6.1078 * EXP((17.269 * (T - 273.16))/((T - 273.16)+237.3))   !saturation vapor pressure
QFRMTP(T,P) = (0.622 * ES(T))/(P - (0.378 * ES(T)))      !specific humidity

Specific humidity is pulled from the background field, too, but using the background pressure at that grid point.  Specific humidity is then analyzed independently, much like temperature and pressure.

Now, it's worth noting that the RTMA/URMA analyzes surface pressure too.  However the derived pressure ob from a RAWS ob would not be used in the analysis, while the moisture ob would be.  So if the assumed/derived pressure is off, that could have a significant impact on the derived dew point that would not show up in the temperature or pressure analysis.

After the variables are analyzed, a processing step converts the temperature, specific humidity and pressure grids to create a dew point grid using the attached fortran subroutine.

I believe any hourly RH grid you might see is produced within AWIPS, and I don't have access to the code for that.

Steve

On Wed, May 15, 2019 at 10:38 AM VLab Notifications <VLab.Notifications@noaa.gov> wrote:

Steve,

 

Is there any way we could get the equation used for the RH? Assuming RAWS observations are getting into the URMA I think a manual calculation at a few locations to see what is produced may help. I know coming up with a pressure for a site based on elevation can be problematic so that may be part of it.

 

Mark


--
Anonymous RTMA/URMA Discussion Group Virtual Lab Forum https://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6613635VLab.Notifications@noaa.gov

--
Steven Levine RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6664453 VLab.Notifications@noaa.gov

--
Ryan Leach RTMA/URMA Discussion Group Virtual Lab Forum http://vlab.noaa.gov/web/715073/home/-/message_boards/view_message/6666076 VLab.Notifications@noaa.gov