The US mainland is shrouded in a slew of weather detection buoys. In my humble opinion, the data from these buoys are greavingly under-used and ignored by much of the weather analysis personel in the US. There are many reasons for this, not far from the top is the validity of the data.
There is only a portion of these buoys that are under the ownership of the US government. Many are under the ownership and responsibility of private companies. The bulk of them are in the petroleum industry. The angst between the petroleum industry and conservation agencies are well documented by other media outlets and I don’t care to re-hash it here. Suffice to say, if the two just got into sync, lots could be done to protect the mainland and outlying areas. But, we are not there.
The main issue with weather detection and analysis (WDA) is the inherently liability if people actually use it for the intended purpose. I’ve witheld many of my ideas due to this. Imagine, all your methods are mathematically sound, but the underlying data is flawed…. People make decisions, launch ships, they flip, lives are lost. Well, the data that helped the captains make those decisions are called into investigation. Well, …. you can see the issue.
However, I’m a permanently disabled mathematician, data analysis and programmer. I no longer see what multi-million dollar lawsuit can do to harm me anymore. So, I’m pretty much saying, “fuck it” and following my ideas for building a buoy-data driven early warning WDA system in hopes that someone in power will see the validity of this proof of concept and run with it. The data is all there, easily scraped from the various locations. It is time to get coding in hopes that the human element can make use of the WDA to make wise choices for their life decisions.
Mainland, or Continental, United States. Phase 2 will include Hawaii and Alaska. Phase 3, if possible, will investigate the possibility of tapping into other countries buoy-generated data. For now, we are looking at Phase 1 as I know this data is available, and is most relevant to Jay’s Cafe’ Community (where I live).
Using RSS XML aggregators, feed the data into a database to be retrieved from weather.jayctheriot.com to build weather maps showing the inherent risks to life and property.
Step 1: A single Data Buoy
- Recall the data from the last month and perform a sinusoidal regression. Find if the current data is more than one standard deviation, SD, from the average of the data from the same time of day (TOD) for the last month. I’m using the SD here as a test. It may be exchanged in the future if I can find a more appropriate test.
- If this test is good, then the analysis moves to the proximal data buoys to compare the analysis of that point. If the next point is valid, then the same data test is executed on all data buoys within the 10 nm, 20 nm and 50 nm range.
Step 2: Looking at neighbors
- Locate nearest neighbors in the database and repeat the calculations performed in Step 1, for neighbors in the 10 nm, 20nm and 50 nm range.
- The results should follow a natural decay pattern. Increasing the granularity could potentially enable forecasters to determine what the anomaly could be. However, that is beyond the scope of this project, at the current time.