Wednesday, April 4, 2018

Intricacies of Atmospheric Rivers

The phrase atmospheric river was first used by researchers in the early 1990s to describe narrow corridors of strong integrated water vapor transport that are frequently found in the mid and high latitudes.  Atmospheric rivers typically achieve high water vapor contents by extracting water vapor from the tropics and/or local moisture convergence.  They are often found along and ahead of cold fronts where flow convergence can concentrate moisture and the strong pre-frontal low-level jet can produce large values of integrated water vapor transport.

Atmospheric rivers are not new phenomenon.  They have played an important role in our weather and climate for ages.  What is new is the name, recognition of their role by atmospheric scientists, and the development of techniques to identify and predict atmospheric rivers.

Two variables are commonly used to identify atmospheric rivers.  The first is the integrated water vapor (IWV), which is the total mass of water vapor in the atmosphere, typically expressed as the depth the water vapor would take if it were all condensed out as rain.  One advantage of IWV is that it can be inferred using satellite and GPS receivers.  The other is the integrated water vapor transport (IVT), which is the total amount of water vapor moving over a location.  IVT is dependent on both the amount of moisture and the strength of the flow and calculating it requires profiles of both moisture and wind.  For this reason, it is most often calculated using three-dimensional gridded analyses or forecasts.

I look at IWV as a measure of moisture availability for storms and IVT as a measure of moisture delivery to storms.  I generally use IVT because it has a stronger correlation with precipitation over the mountain west than IWV, as can be seen in the plots below.  However, in some lowland areas of the western US and in the High Plains, IWV has a higher correlation.   These are areas where high moisture availability plays an important role in fueling convective storms (note, this is an instantaneous correlation – high IWV is often a result of prior moisture transport).
Source: Rutz et al. (2014)
IVT has a strong correlation with precipitation, but the correlation isn't perfect.  This is because precipitation is dependent on many factors, including a mechanism for lift.  The highest correlations between IVT and precipitation are found in the Pacific coastal ranges, Cascade Mountains, and Sierra Nevada (see left-hand panels above).  These mountains are relatively broad and are frequently experience strong cross-barrier moisture transport accompanying atmospheric rivers, with forced lifting by the mountains generating precipitation. 

The correlation is lower, however, over the western interior.  Within that region, the highest values are found in the Idaho Panhandle, central Idaho Mountains, southwest Utah, Mogollon Rim, Tetons, and San Juan Mountains.  These lower correlations reflect several factors, including the greater diversity of processes generating precipitation over the western interior and contrasts in precipitation efficiency between wide and narrow mountain barriers during some storms.

The mountains of northern and central Utah have lower correlations.  This does not mean that we can't get significant precipitation from atmospheric rivers.  Instead, it reflects the fact that we sometimes get significant precipitation from storms that don't feature large cross-barrier moisture fluxes.  An example are cold, post-cold-frontal storms that generate large amounts of snowfall when the water vapor content of the atmosphere is relatively low (in an absolute sense) and the flow weak.

Moisture transport is very important for storm dynamics, but one needs to be cautious about developing AR-myopia.  As shown above, the correlation between IVT and precipitation is not perfect, and is modest or even low in portions of the interior western U.S.  Consideration of the precipitation generation mechanisms, such as frontal or orographic lift, is also important.

As an example, below is the 0600 UTC initialized NAM IVT forecast valid 0000 UTC 7 April.  Most of the eastern Pacific off the California coast is experiencing atmospheric river conditions (IVT>250 kg/m/s), with the highest values along an axis running from about 35˚N, 130˚W to 40˚N, 125˚W.

Source: CW3E

The NAM forecast 3-h accumulated precipitation at this time is, however, nearly non-existent in that area of high IVT and is instead highest over northern California, especially in the coastal ranges, northern Sierra Nevada, and southern Cascades where there is both warm frontal and orographic forcing generating precipitation.  The AR provides the moisture.  The fronts and mountains provide the dynamics.

I hope to look at what will happen as this system moves inland in a future post.  Forecasts currently suggest something that looks like what you get when a November cyclone meets April.


  1. No doubt you are puzzling over the GFS in the LCC guidance. What is concept Mt Baldy temp? Temp on the summit, 11,000 feet? Snowfall is for the Collins snowstake, at 9700 feet? From the 4/4 12z run. Mt Baldy temp rises almost steadily from 10 degrees F at 9am Weds 4/4 to 30 at 10 am Sun 4/8. Rain snow line during the main part of the event, Fri 12am to Sat 8pm looks to stay below 8,000 feet. Seem right? Rain/snow line does rise to 9000 feet, but that appears to be after most precip. Then Baldy temps fall to the teens as the cold front passes Sunday/Monday. Seeds for your future post.

    1. I don't puzzle about anything :-).

      Those numbers in the LCC guidance are pulled from the nearest GFS grid point to Alta. Water amounts are for that grid point. Snow-to-liquid ratio and snow water content (percent) are based on the Alcott and Steenburgh (2010) algorithm applied to data from that grid point. Snow amounts are based on the GFS precip multiplied against that snow-to-liquid ratio.

      The biggest weakness here is the GFS overprediction problem at this location. We're not correcting for that. Our snow-to-liquid ratio algorithm isn't perfect, but it works pretty well.

      Wind speed and temps are for Mt. Baldy (11,000 ft) and involve the use of an algorithm applied to the GFS data.

    2. And, looking further at the numbers, those in the table on Sunday are clearly wrong. Looking into this issue now.

    3. Just noticed the GFS winds, gusting to 76mph Sun 2pm. Gonna be wet and windy wild wild storm if this verifies. I've been out in these types of April storms, very nasty.

      Sunday numbers seemed fine to me given the AR story the models are telling. I'll look forward to your explanation if you get time to put it in a post on how this storm plays out.

      Right now it seems like strange exciting weather batten down the hatches find high elevation low angle as the avy danger seems like it will really spike.

  2. Re the IVT forecast shown above - looks like a huge areal extent of significant IVT, but I don't really see a well-defined river.

    1. The river is more apparent at this site

      As Jim suggests, river may be a bit of hyperbole.

      Pineapple express?

    2. Thanks for link Peter - the PWAT product there shows huge slug of high PW crossing Hawaii and heading northeastward.