2020 Indicators: SST's / SAL / MSLP / Steering / Shear / Instability / Sat Images

[AutoPenalti]

A suppressed MDR until late August, which is usually the start of peak but will be very short-lived.

[MarioProtVI]

crickets in the long-range models. If this were a 2005 or 1933, surely we would be seeing some kind of activity in the models. That said, of course the season could still end up active especially since most seasons are quiet in July anyway.

There was an MDR cyclone showing up around mid-to-late July on the CFS yesterday.

[CyclonicFury]

I'm not sure if Eric Webb's plot uses a different climatology, but August doesn't look suppressed at all on the NMME plot from NOAA.
 https://twitter.com/CyclonicFury/status/1280630203098836992

[USTropics]

I'm not sure if Eric Webb's plot uses a different climatology, but August doesn't look suppressed at all on the NMME plot from NOAA.
https://twitter.com/CyclonicFury/status/1280630203098836992

I just plotted the raw data, and I'm pretty sure I know what the problem is. The netcdf data format for NMME have lead time groups 0-12. Most seasonal forecast models use a 1-12 format (1 being August, 2 being September, etc.) but NMME has a 0 for remainder of July (so 0 for July, 1 for August, 2 for September). If you don't set initial from 0 to 1 when calling the data, July will be initial. So his plots are off by a month. Below are the July and August forecasts I just plotted using the netcdf files:

Lead time of 0 (remainder of July):


Lead time 1 (August):


Comparison to official product for August:

[aspen]

Given all of this, it seems like there’s a chance of MDR season starting in late July with a more active August than the last few years, correct?

[CyclonicFury]

At this rate we will surpass 2014, the last time this naming list was used, in named storms before the "real" season begins.

[SFLcane]

I'm not sure if Eric Webb's plot uses a different climatology, but August doesn't look suppressed at all on the NMME plot from NOAA.
https://twitter.com/CyclonicFury/status/1280630203098836992

I just plotted the raw data, and I'm pretty sure I know what the problem is. The netcdf data format for NMME have lead time groups 0-12. Most seasonal forecast models use a 1-12 format (1 being August, 2 being September, etc.) but NMME has a 0 for remainder of July (so 0 for July, 1 for August, 2 for September). If you don't set initial from 0 to 1 when calling the data, July will be initial. So his plots are off by a month. Below are the July and August forecasts I just plotted using the netcdf files:

Lead time of 0 (remainder of July):
https://i.imgur.com/MFlHOre.png

Lead time 1 (August):
https://i.imgur.com/7Hh0H5H.png

Comparison to official product for August:
https://i.imgur.com/WIkaTKI.png

Quiet August I don’t think so. What program are you using?

[SFLcane]

 https://twitter.com/BenNollWeather/status/1280856664879411202

[SFLcane]

From the new NMME my take is there might be a few recurving storms this season but not many.

[Steve]

https://twitter.com/BenNollWeather/status/1280856664879411202

That's ****ed. I have been wondering where the concentrations of rain were since they all seemed to point in certain directions but ended in the water or in Mexico or the islands. This paints a different picture. I don't usually follow the monthly forecasts of the climate or longer range models, but this is a bad sign for the Gulf from Texas to Florida. I think most of us assumed there would be 5 or more named storms in the Gulf this year. But this update shows the threat for South Texas and also the NC Gulf. Definitely trending toward dangerous here.

Also congrats to Philippe Papin on his job at the NOAA/NWS. Hopefully he won't tail off on his posts.

[SFLcane]

This is such a wild crazy comparison regarding this season. If this won't get you to be ready this season nothing will. Can't stress enough enjoy this slow time.

 https://twitter.com/webberweather/status/1280909663710650374

[cycloneye]

This is such a wild crazy comparison regarding this season. If this won't get you to be ready this season nothing will. Can't stress enough enjoy this slow time.

https://twitter.com/webberweather/status/1280909663710650374

I have seen many comparisons with 2017 but this one that Webb has is the most omminous one yet. Hopefully PR gets spared.

[Shell Mound]

So to be sure that 2020 will be hyperactive, in addition to the next few formations in the subtropics, we need at least one or two MDR formations before July ends.

Why? Not having a go, but this seems entirely arbitrary. Why not one or two in the Gulf or BoC? Or in the subtropics? Is this because you figure MDR storms would accumulate more ACE than the usual locations where July storms would form?

It doesn’t have to be in the MDR, but there is a bit of a correlation between the presence of AEW-based July systems and an above-average Atlantic hurricane season. Seasons with a July system formed from a tropical wave tend to be more active than those with a system formed from other methods.

Also, despite what some people have been saying, credible people do use seasonal dynamical models to predict likely TC tracks:
 https://twitter.com/AndyHazelton/status/1280910192864038917



 https://twitter.com/AndyHazelton/status/1280910974837555201



 https://twitter.com/NMBBuckeye/status/1280911313611489287



 https://twitter.com/AndyHazelton/status/1280919404935557121



These models have even proven reliable, as the tweets and images illustrate.

[Hammy]

Doesn't having a lot of green also mean that you have multiple (often strong) storms going over the same area? I assume a single storm or two wouldn't show up in the models like that.

[USTropics]

I'm not sure if Eric Webb's plot uses a different climatology, but August doesn't look suppressed at all on the NMME plot from NOAA.
https://twitter.com/CyclonicFury/status/1280630203098836992

I just plotted the raw data, and I'm pretty sure I know what the problem is. The netcdf data format for NMME have lead time groups 0-12. Most seasonal forecast models use a 1-12 format (1 being August, 2 being September, etc.) but NMME has a 0 for remainder of July (so 0 for July, 1 for August, 2 for September). If you don't set initial from 0 to 1 when calling the data, July will be initial. So his plots are off by a month. Below are the July and August forecasts I just plotted using the netcdf files:

Lead time of 0 (remainder of July):
https://i.imgur.com/MFlHOre.png

Lead time 1 (August):
https://i.imgur.com/7Hh0H5H.png

Comparison to official product for August:
https://i.imgur.com/WIkaTKI.png

Quiet August I don’t think so. What program are you using?

It's a process, and I'll try and explain as much of it as possible. The TL:DR is this is generated through a python script in a Juypter environment. More specifically, I'm using these python modules (all can be installed through pip or conda):

Plotly - lets us visualize data in graphs, charts, maps, etc.
Basemap - simple basemap file for the background countries/borders/etc.
Numpy - handles the math functions of our vectors
Xarray - to help us merge multiple subsets of data
NetCDF - python has it's own module to help us read netcdf files

First we need to understand what the data structure is. If you're new to what a netcdf file is, it's important to understand the structure. Essentially we're dealing with some intro calculus concepts, more specifically an x,y, z vector (X is longitude, Y is latitude, Z is the data we're plotting). A typical structure for one month of data (ssta, precip, whatever value we're looking at):

Lon (x vector): 1,2,3,4,5... all the way to 360.
Lat (y vector): 1,2,3,4,5... all the way to 180.
Data (z vector):
// fcst (1,0,0-359).......a bunch of data for each corresponding x/y vector point.
// fcst (1,1,0-359).......a bunch of data for each corresponding x/y vector point.
...........................................
..........................................
// fcst (1,179,0-359).......a bunch of data for each corresponding x/y vector point.
// fcst (1,180,0-359).......a bunch of data for each corresponding x/y vector point.

If you understand what a dict is, or a json string, that's essentially what we're dealing with. I'll use the data files I was working with last night, the latest NMME forecast can be found here - ftp://ftp.cpc.ncep.noaa.gov/NMME/realtime_anom/ENSMEAN/2020070800/. If you download one of those netcdf file, you can use a shell environment to look at the textual output. I'm using Anaconda for this (https://www.anaconda.com/). Since we can't visually see netcdf data, we need to convert it to a .cdl file. You can do a simple ncdump command (ncdump -b c filname.nc > filename.cdl) to achieve this. We can now view this data in a text editor and mess around with this raw data, and then reconvert back to a netcdf file later (ncgen -b filename.cdl - will create/overwrite a file called filename.nc).

So now that we know the structure, we can start to grab what we need to visually show this data on a map. There's a lot of python coding involved here. A simplistic overview of the process is to:


- Import list modules above, download a netcdf file. We're going to use that NetCDF python module to call our filename.nc file and then we're going to import the X,Y,Z data (for the NMME file it's called lon, lat, and fcst).
- There's some work needed to be done to get the proper global view, so we'll shift some arrays around to achieve this.
- Next we will establish our trace (add contours, color, and stylize our data).
- We import the basemap to draw over with Plotly, and depending on what trace function you're using (I'm using scatter objects), we'll draw out our outlines and borders with plotly.
- Last step is to setup the output for Plotly. Basically we'll import the trace we made earlier, the basemap project, and some additional things (like the legend and axis).
- Finally we tell Plotly how to handle the output (plot figure offline in this instance), and we now have our image!

Obviously this is a conceptual outline, and only deals with one month of data. If anyone is interested in the actual coding aspect, I'll have a project on github soon (or message me). There's some work I need to do with xarray to merge monthly data (to get accurate tri-monthly means) before I release anything.

[AutoPenalti]

Obviously this is a conceptual outline, and only deals with one month of data. If anyone is interested in the actual coding aspect, I'll have a project on github soon (or message me). There's some work I need to do with xarray to merge monthly data (to get accurate tri-monthly means) before I release anything.

You just pique’d my interest.

[CyclonicFury]

Although I wouldn't be surprised to see a somewhat stronger tropical origin TC later this month, I have skepticism we will see the "lid come off" until around the 3rd or 4th week of August. 2017 did not take off in the deep tropics until very late August. 2018 and 2019 were practically dead in most of August outside the high latitudes.

Atlantic climatology is rarely favorable enough for a late July/early August activity burst. However, it has happened on rare occasions, such as 1990 and 2004. SAL outbreaks are still fairly common through the third week of August.

After August 20 or so, however, I expect an intense burst of activity. I do think we have more potential for tropical activity in late July/early August than the past 2 years, but I don't expect to see major ACE produced until late August.

[cainjamin]

Even though the GFS isn't showing any MDR development in the recent runs, it definitely is showing a moister central Atlantic by the mid-end of July:

Today:


384hr:

[gatorcane]

All of these models and fancy graphics, tweets, etc might look ominous but let’s face it, there is just little skill in long-range forecasting of tropical cyclones. That’s why all of the “indicators” should be taken with a grain of salt. The main signal it could be active in the Atlantic is because of the above normal SSTs across the Atlantic in the MDR and lack of El Niño. It should be active and above normal but just wouldn’t be too quick to say that it equates to active landfall season. The other thing is that there is no sign in the models for any WPAC activity. It is early but as we head into July would expect to see something. There could easily be something else we can’t see yet that helps suppress global cyclone activity more than expected including the Atlantic.

[Steve]

^^

Skill has been going up over the last couple decades. Maybe they can’t pinpoint a specific threat a month out. But no one should think we haven’t gotten to the point where we know a lot about global weather patterns and implications. Certainly computer models get that stuff too to varying degrees of success.