WeatherCitizen User Stories

Reply to this topic with to share your experience using WeatherCitizen in the wild.

Ski Touring in Northern New Hampshire

We recently took a trip up to northern New Hampshire to search for some good back-country skiing. This area did not have cell coverage, so we navigated using (old-fashioned) maps and line of sight. Since the area is pretty heavily logged, we knew we would be able to find some good open trees for skiing.

Skiing around the valley, we found a nice open deciduous forest that had not been re-vegetated by striped maple and basswood. Once we found our line, I turned on the background data collection in WeatherCitizen so that we could re-trace our ski line through the forest. I took a few pictures along the way to document the state of the forest and snow conditions.

Map Link

Extracting percent cloud cover from images using machine learning

We have trained a Deep Convolution Neural Network (DCNN) to determine the percent cloud coverage in images taken with the WeatherCitizen App. Ultimately, we hope to have this feature in the App and on our web map but for the time being it is easily available through our API.

If you visit our web map and click on one of the images an info bubble will open up with a larger version of the image:


If you click on the larger image it will bring you to an url that looks like this:

This is the full sized image. If you change this url to replace files with ccpred like this:

It will return JSON that includes the cloud_cover estimate. I’ve included a screen shot of what this looks like in Firefox. I like to use Firefox for querying our API in a web browser because it does a good job of rendering the JSON and because you can hide parts of the JSON if you want to reduce the clutter.


Click on “Collapse All” to remove some of the clutter:

This use case is from Andy DePaola, an oyster farmer in Mobile Bay, AL.

" I am an oyster farmer in Mobile Bay and marine observations are critical to my operations and survival of my oysters.

I’m delighted to participate in a program that collects weather and other observations that improve the accuracy and reliability of National Weather Service forecasts. The producers of Weather Citizen have been very receptive to expanding user inputs to observations that are important to my business, such as salinity. Current salinity forecasts are directed toward navigation, monitoring is far from my farm, and the lack of granularity results in unreliable forecasts. Salinity is an important consideration oyster farming, for flavor especially, but low salinity has also wiped out my entire crop. I’m also a scientist and would like have a network of oyster farmers acting as citizen scientists, inputting salinity and other data. Weather Citizen could also be used as a portal to register adverse events that harm my oysters, threaten human health, and the viability of my business."

Using WeatherCitizen to share weather observations on twitter

Several mornings a week, I use the WeatherCitizen iOS app to make a weather observation from my back porch and submit these to both the WeatherCitizen server and twitter.

Here's an example posting from April 8, 2019 as it appears on the Weathercitizen server

and on twitter.

#mzreport WW 43.5051 -72.156 WW 35F. Flurries

— Jerry Bieszczad (@BieszczadJerry) April 8, 2019
Here's a link to all of my twitter postings. You can also search twitter for WeatherCitizen's #wxreport hashtag (formerly #mzreport) to find tweets by other WeatherCitizen users.

Comparing WeatherCitizen data to a Vantage Pro 2 weather station

We’ve set up a Vantage Pro 2 weather station with a Samsung Galaxy S4 phone running the WeatherCitizen App so we can compare the results from sensors in both devices.

Below is a plot comparing the temperature measurements from the Vantage Pro 2 station with the battery temperature sensor and ambient temperature sensor on the Galaxy S4 phone. The match very well during the nighttime hours, but the phone gives consistently high temperature readings during the day. The battery temperature (red) is also slow to adjust at the beginning of the measurement when the phone is brought from inside, where it’s warm, to outside, where it’s very cold in January in New Hampshire!


If we look at the luminescence data from the phone, we see that the warm temperature readings are from solar heating of the phone on a sunny day:


We hope to use this weather station to benchmark the WeatherCitizen App and to develop corrections to improve data quality.

Tracking hikes in the White Mountains

The weathercitizen app and map can be a fun way to track activities in the outdoors. Besides just recording location, the weather data and geotagged photos make an interesting way to share a hike. Here’s a well-known route in New Hampshire known as the Presidential Traverse [map]:

As one can see from the images, we were in the clouds over the northern Presidential range, but somewhere south of Mount Monroe we dropped below the cloud ceiling and had nice views of the rest of the range for a while.

The elevation of the hike can be plotted to show our progress over the course of the long day [map]:

The Signature of a Hurricane

On September 14, 2018, Hurricane Florence made landfall at Wrightsville Beach, NC. The National Data Buoy Center (NDBC) has a number of buoys just off the coast at this location. Data from these buoys are included on the WeatherCitizen map:

The plunge in barometric pressure at 6am shows the eye of the storm passing over Wrightsville Beach. Because the storm rotates around its eye, the wind direction flips by 180 degrees as the eye passes. There are also short-lived drops in wind speed and wave height corresponding to the calm in the eye of the storm. The data can be explored on the WeatherCitizen map.

Field Study on Long Island Sound

On June 20,2019, we performed a large-scale weather data collection day in Long Island sound. The purpose of this effort was to collect data that could be assimilated into forecasting models, and then see if the data from WeatherCitizen improved the forecast. We colloborated with SUNY Stony Brook and the CT Department of Agriculture, for a total of 12 observers. The observers sailed on various ferries and research vessels, as well as recording observations from beaches and docks. The screenshot below shows where data was collected:

Unfortunately, the weather was somewhat uncooperative; while the seas were exceptionally calm, there was thick fog and poor visibility all day, forcing some observers to stay in port. Despite that setback, however, the day was still a great success. We recorded close to 2600 data points in about 8 hours!

Since our data collection day, we have processed the data into CSV files, and it being formatted for assimilation into forecasting models. In an effort to get more more even coverage of the area of interest, we plan to perform another data collection day in August, hopefully with better weather!

Current and Historic Radar Data

The latest version of the WeatherCitizen map includes a radar layer. As a follow-up to a previous post, here’s an image from Hurricane Florence making landfall in North Carolina last year:

AIS-Weather Integration

The US Army Corps of Engineers Coastal and Hydraulics Laboratory, in cooperation with the Maritime Administration and supported by Serco Inc., has developed a technology which uses shipboard Automatic Identification System (AIS) for telemetry of weather observations from vessels at sea. The AIS-Weather system automatically collects weather data from a shipboard weather station, and broadcasts these data which can be received by AIS stations ashore or receivers aboard satellites. This technology has the potential to provide much-needed weather observations at locations far from land for assimilation in weather models. (US Army Point of Contact: Brian Tetreault)

As a proof-of-concept integration of AIS-Weather with WeatherCitizen, we used the WeatherCitizen API to upload AIS-Weather data to our server for display on our web map. The map below shows data from three vessels in the Pacific on September 23, 2019.

Buoy Ride-Along

One of the features for WeatherCitizen we’ve been working towards is a method for estimating sea state (Significant Wave Height and Wave Period). The idea is that a user would place the phone on a flat surface in a boat or kayak, and then record data for a couple of minutes without disturbing the phone. The app would then process the data immediately and report the sea state.

To test our algorithm for calculating the sea state, we convinced the fine researchers at the University of New Hampshire to allow us to piggy-back onto one of their CO2 monitoring buoys. These buoys are put out for about 10 months at a time off the coast of NH’s Isles of Shoals (location is marked with the red dot below).


The idea for the test was to put a phone on the buoy and record a really long session, hopefully over about two weeks. The data would then be transmitted in real time via 4G data, and stored on the WeatherCitizen server. After the two weeks of data recording, we’d then download the data and process it in MATLAB, using our algorithm.

To protect our phone from the North Atlantic harshness, we purchased a weatherproof box from PolyCase and arranged a phone and large capacity charger inside. We used blueboard to insulate and secure the phone and battery.

Once we had our payload completed, we traveled to UNH in Durham and bolted it to the frame of the buoy.

A couple weeks later on 12/20/2019 we met them at their pier so we could start up the app and begin collecting data! They deployed the buoy the next morning.

We were able to monitor the battery charge and temperature throughout the test using the web map feature on our website, and the battery lasted just over 12 days. We thought this was pretty impressive given the cold temperatures and 24/7 recording.

We then downloaded the data, applied our algorithm, and compared the results to sea state data reported by another buoy about 10 miles further north along the coast.


Overall, we were able to measure SWH within ~25cm of the “ground truth” buoy. Pretty good for a $250 kit strapped to a buoy! It’s also important to note that our buoy was 10 miles away from the ground truth buoy, and so some variation in the data is to be expected.


Unfortunately, we were not as good measuring the wave period. We mostly picked up the 0.5Hz pendulum oscillation of the buoy.

Next, we’re going to continue tweaking our algorithm to better measure period, and hopefully run a similar test in the spring.