Proposal: Natural Intelligence - Bird-driven weather forecasting

Proposals Round 5
#1

bird-storm-desert
bird-storm-desert800Ă—561 20.8 KB

Name of Project: Natural Intelligence - Bird-driven weather forecasting

Proposal in one sentence:

Birds can predict the weather months in advance to avoid extreme weather events, how might we through technology partner with this natural intelligence to enhance human understandable weather forecasting.

Description of the project and what problem is it solving:

Inspiration: Can Birds Predict Storms (and leads for data research & collaborators).

Round 1 - 1 month Gather & Prepare backtest

  1. Gather & scrub appropriate bird migration data and weather data
  2. Begin backtesting to prove viability of a forecasting model

Further iterations

  1. Research and determine what ongoing data & resources is needed to construct a forward prediction model
  2. Co-design and evolve this with real-world stakeholders that would value this service.

Grant Deliverables:

  • Scrubbed & ready bird migration & relevant weather data
  • Create the model & plan to backtest viability for a weather forecasting model. The actual work to be complete in subsequent rounds.

Spreading the Love

Looking to hire a data scientist to help explore existing datasets and prior work on ML, main datasets are bird migrations & extreme weather events - to be gathered, scrubbed & analysed.

This role will be advertised on the Algovera opportunities board and DeWork.

Squad

Klaus Bravenboer. Twitter @klausbrave
Sebastian Wallace. Twitter @sebjwallace

  • Looking for a data scientist
3 Likes
#2

This is a really intriguing project. It’s always a good idea to take some base knowledge off nature and transform it humanely. However, I don’t really see what bird parameters would you do a data study on. I suspect it’s not neurological but behavioural, example how does a flock of birds moves as a collective before and after a storm.
Btw, I’m a Machine Learning Researcher/Engineer who has done research and publications in collective intelligence. Would be happy to assist on the project. Peace :dove:

1 Like
#3

While its plausible that birds have senses for storms, their behavior is not only optimized for storms, but also for food and mating. Whereas we have many excellent satellites that can directly quantify the amounts of gasses, temperature, humidity, etc, and there is a tremendous infrastructure behind it. While its interesting that some predictions are possible, I do not see how the predictions can be better than the state of the art.

1 Like
#4

This sounds interesting.

While my prior is that it would be hard to beat the state-of-the-art as @endomorphosis says, I wonder if it is possible to figure out what amount of time it takes for meteorological forecasts of longer-term climate events over a region to converge (since they often change, to my knowledge) and if bird migration is a quicker signal for such events.

Another angle could be to understand if bird migration patterns around unexpected weather events reveal insights that meteorologists failed to capture with their climate models.

If I were you I’d explicitly highlight cases where meteorology is slow or somewhat inaccurate in making predictions to really highlight the utility of alternate data like bird migration and push this proposal through.

1 Like
#5

I was in the shower, and i was thinking what might be a better idea, is to try to predict the mortality or fecundity of the birds protected by the migratory bird treaties, based on the environmental conditions.

1 Like
#6

Maybe that’s an interesting use case for this project; reducing the significant infrastructure/compute costs of weather prediction by using natural proxies (e.g. bird migration data). Not sure how realistic this claim is but it’s at least a fun concept to entertain!

1 Like
#7

Yes this is the idea.
Birds have this natural intelligence, how might we partner with what already exists using technology to grow our literacy and translation rather than needing to mimic or make our own version.

#8

I envisage looking at migration data as is referenced in the article.
Looking at deviation from normal migration paths to avoid extreme weather events.
Would love to speak with you more about your past experience and how it could be applied here.

#9

Birds are forecasting and avoiding extreme weather events up to 3-months in advance according to the research.

Our best efforts are 5-10 days.

This is worth the research to discover what’s at the source of this ability and can we partner with it or mimic this ability. How value would it be for us to have this ability as a species?

“A seven-day forecast can accurately predict the weather about 80 percent of the time and a five-day forecast can accurately predict the weather approximately 90 percent of the time. However, a 10-day—or longer—forecast is only right about half the time.”

#10

I think you are conflating “weather” and “climate”.

Weather is forecasting a hurricane, a rain storm, on a specific day.
Climate is predicting whether there will be an “el nino” year.

The paper claims that the birds are able to predict climate, better than NOAA, however there are caveats.

Only for selected years “1998 to 2016 (excluding 2008)” (this paper was published 2018),

Only for selected months. “Therefore, May and June songbird phenology and clutch size were equal or more accurate long-range predictors of hurricane activity than the early season”

Which indicates that they are cherry picking the data, or otherwise known as “p hacking”.

Also, i should note that nothing in this paper uses any machine learning whatsoever, it does use statistical predictions however, to correlate the number of hurricanes in a season with the number of nests.

"Statistical analyses were accomplished using the R statistical package version 3.4.4 "

1 Like
#11

Hi here. Thank you for your insightful response.

This is the kind of rigour I would love applied to validating, invalidating or discovering if there is something here worth investigating and building upon further.

Whether it’s weather or climate, let’s investigate what piece of the puzzle birds natural intelligence is holding, what data about their movement or their nest numbers or nest height could be used to help us make predictions about future events. Or collate & discover what gaps there is in our current knowledge, models, data strategies that could built over time.

#12

Like i mentioned before, there are treaties regarding migratory birds, and that a better use of the data, is to predict bird mortality and fecundity based upon climate, such that nation states can comply with their obligations.