TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Historical weather data API for machine learning, free for non-commercial

196 点作者 meteo-jeff将近 3 年前

11 条评论

meteo-jeff将近 3 年前
Some technical background:<p>Open-Meteo offers free weather APIs for a while now. Archiving data was not an option, because forecast data alone required 300 GB storage.<p>In the past couple of weeks, I started to look for fast and efficient compression algorithms like zstd, brotli or lz4. All of them, performed rather poor with time-series weather data.<p>After a lot of trial and error, I found a couple of pre-processing steps, that improve compression ratio a lot:<p>1) Scaling data to reasonable values. Temperature has an accuracy of 0.1° at best. I simply round everything to 0.05 instead of keeping the highest possible floating point precision.<p>2) A temperature time-series increases and decreases by small values. 0.4° warmer, then 0.2° colder. Only storing deltas improves compression performance.<p>3) Data are highly spatially correlated. If the temperature is rising in one &quot;grid-cell&quot;, it is rising in the neighbouring grid cells as well. Simply subtract the time-series from one grid-cell to the next grid-cell. Especially this yielded a large boost.<p>4) Although zstd performs quite well with this encoded data, other integer compression algorithms have far better compression and decompression speeds. Namely I am using FastPFor.<p>With that compression approach, an archive became possible. One week of weather forecast data should be around 10 GB compressed. With that, I can easily maintain a very long archive.
评论 #31998638 未加载
评论 #31998665 未加载
评论 #32004024 未加载
评论 #31998857 未加载
评论 #31998478 未加载
评论 #32000055 未加载
m0llusk将近 3 年前
This is how we could defeat a rogue AI: Distract it by talking about the weather.
MeteorMarc将近 3 年前
It would also be fun to have the historical weather *<i>forecasts*</i> so that you can compare the forecasts with the eventually measured data.
评论 #31998038 未加载
bernulli将近 3 年前
Hi meteo-jeff, this looks really cool!<p>I have two questions:<p>1) How does the spatial resolution come into this? Is it constant data all across the 2kmx2km (?) parcel with an abrupt change, or is it interpolated in some way? Can I query the coordinates of the mesh?<p>2) How &#x27;historical&#x27; does it get? How far back can I go with this?<p>Thank you!
评论 #31997925 未加载
评论 #31995877 未加载
ricksunny将近 3 年前
Original open-meteo HN thread for background (9 months back) <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=28499910" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=28499910</a>
aarreedd将近 3 年前
Does not seem accurate. This is telling me it snowed 1.33cm on June 17, 2022 in New York City.<p><a href="https:&#x2F;&#x2F;api.open-meteo.com&#x2F;v1&#x2F;forecast?latitude=40.71&amp;longitude=-74.01&amp;daily=rain_sum,snowfall_sum&amp;temperature_unit=fahrenheit&amp;windspeed_unit=mph&amp;timezone=America%2FNew_York&amp;past_days=92" rel="nofollow">https:&#x2F;&#x2F;api.open-meteo.com&#x2F;v1&#x2F;forecast?latitude=40.71&amp;longit...</a>
评论 #32002319 未加载
m3kw9将近 3 年前
Still trying to predict weather using historical is like trying to predict the next number on a roulette using historical numbers
评论 #32000230 未加载
评论 #32003226 未加载
评论 #32002337 未加载
mhalle将近 3 年前
Thanks for offering this service!<p>You explain your API offers historic data using the &quot;past_days&quot; parameter. Could you also offer a &quot;date&quot; parameter for a given day, or are you only keeping a rolling window of data?
评论 #32001388 未加载
Kalanos将近 3 年前
check out <a href="https:&#x2F;&#x2F;docs.aiqc.io" rel="nofollow">https:&#x2F;&#x2F;docs.aiqc.io</a> for easy walk-forward, multivariate deep learning: <a href="https:&#x2F;&#x2F;docs.aiqc.io&#x2F;notebooks&#x2F;gallery&#x2F;tensorflow&#x2F;tab_forecast.html" rel="nofollow">https:&#x2F;&#x2F;docs.aiqc.io&#x2F;notebooks&#x2F;gallery&#x2F;tensorflow&#x2F;tab_foreca...</a><p>excited to play w some of this data
farmin将近 3 年前
Do you have a commercial option? Does anyone know good alternatives?<p>What forecast models do you use for Australia?
评论 #31997995 未加载
评论 #31997211 未加载
评论 #31996322 未加载
melony将近 3 年前
You can get these for free from the government websites.
评论 #31997357 未加载
评论 #31997349 未加载