Improving Models Through the Use of Backyard Weather Station Data

[BornFromTheVoid]

I wasn't sure where to put this, hopefully the model area is ok.

 

After seeing this paper abstract today, I wondered was there the possibility to improve models in general by creating a more dense network of observations using backyard weather stations and other amateur data sources..

 

Improvement to Mesoscale Analyses and Forecasts by Assimilating Dense Pressure Observations

 

Surface pressure observations have been shown to provide valuable information to synoptic-scale forecasts and analyses, but their ability to describe mesoscale phenomena, many of which exhibit distinct pressure signatures, remains underexplored.  To capture these features, a very dense network of observations was sought by exploring novel, yet extant observation platforms.  Pressure observations from citizen observers and backyard meteorologists are readily obtainable and were found to increase the density of observations by an order of magnitude over the ASOS network in the Pacific Northwest.  Quality control and bias correction methods for these pressure observations, including the option of using pressure tendency observations, were developed.  A month-long series of experiments using the University of Washington Real-Time Ensemble Kalman Filter system examined the impact of assimilating these additional pressure observations on mesoscale analyses and short-term forecasts. The assimilation of these additional pressure observations made localized adjustments to the surface pressure, wind and temperature fields surrounding various mesoscale phenomena. Short-term forecasts following analyses produced using the dense pressure observations had statistically significant reductions in errors throughout the lower troposphere.  In addition, assimilating these observations also yielded improved forecasts of frontal passage timing and convective development.

 

 

Obviously enough, manner errors and uncertainties would need to be taken into account, but between smart phones, home weather stations, vehicle weather data and more, surely a dense network of observations could be derived and utilised to improve models and forecasts?

Thoughts?

[Recretos]

In my humble opinion, 5 good 99% accurate observations are much better than 20 half accurate or even false observations (just an example). 

 

When it comes to models, you want only the best data assimilation in the input data for initial conditions.  

 

You wouldn't put sand and dirt in your cars gas tank, now would you? Your car might perhaps still run, but the performance of the engine would be highly questionable, if any at all.  

 

Besides, the data assimilated by the GDAS (Assimilation system for GFS/GEFS) for example, is not really thrown directly into the model, but is checked and QA-ed.

So for the usage in short term local models, I guess it would be worth experimenting with. But I am very sceptical about the usage of this additional data in the large scale models.

 

Best regards.