Rain map. Now on the map: how the new Yandex.Weather. Not yet, but technically and theoretically feasible

Weather is one of the essential factors that directly affect our well-being and mood. Good, sunny weather makes us happy and positive, bad weather leads to apathy, melancholy and depression. Trying to predict the weather for the near future, we turn to the help of various network services available online and completely free of charge. In this article, I will talk about the services that display the precipitation map online, describe their capabilities, and also explain how to work with them.

To date, there are a sufficient number of sites that provide users with access to a map of precipitation in real time. Such services are presented both in the Russian-speaking segment (for example, Yandex weather map), and on foreign, mainly English-language resources (Meteoblue, WeatherOnline, Gismeteo, etc.).

The vast majority of them have a Russian-language interface, and when switching to them, the Russian language is automatically activated. All of them are free, have an intuitive interface. The most advanced of them offer the user to take part in the project by joining the community and purchasing special equipment.

These sites are pretty easy to use. You go to the resource, turn on the desired display type (for example, precipitation, temperature, wind, humidity, etc.). Next, use the search (or mouse) to find the region you need, and observe the result.

Let's look at the resources that offer their users in real time.

Meteoblue.com - Displays real-time rainfall maps

The Meteoblue.com online service offers users high quality weather information. It is aimed mainly at weather-dependent business structures and professionals working in meteorology. The project was created by the University of Basel back in 2006. It has gone through several stages of its development, and now offers precipitation forecasts based on the work of NOAA / NCEP GFS models.

1. To work with the resource, go to meteoblue.com.
2. Select the country and type of display (precipitation - precipitation).
3. After that, you can observe on the map how precipitation changes with each subsequent hour.
4. The menu on the left will allow you to see the weather on this moment. As well as a forecast for the future, view meteograms and enjoy webcams online.

Windy.com - shows online the number of thunderstorms on the map

Windy.com is one of the successful projects of the Czech millionaire and philanthropist Ivo Lukakovic. He developed an online map to track rain and snow for home entertainment. Over the years, the service has acquired a professional status, replenished with strong programmers, and received a mobile implementation on Android and iOS OS. At the moment it is one of the most popular weather resources in the world.

1. To work with the service, log in to windy.com;
2. On the right, activate the option "Rain, thunderstorm";
3. In the "location search" line, type in the name of the country (city) whose weather you need to know.

Other weather options are also available on the map - displaying temperature, clouds, waves, wind, dew point et cetera.

Precipitation intensity level at windy.com

Ventusky is a popular Czech weather service.

Czech service Ventusky (from Latin "ventu" - "wind" and English "sky" - "sky") is a worthy national competitor to the aforementioned Windy.com. The Ventusky project was created by the Czech company InMeteo, which specializes in creating online meteorological forecasts around the world. The weather image on the map is constantly changing in real time, the resource has a high-quality Russian-language interface. It also has its own mobile applications, which are available by clicking on the "App" button.

1. Launch ventusky.com.
2. On the left, select the “Precipitation” (or “Thunderstorms”) option and use the mouse (or the search bar) to find the area you are looking for.
3. At the same time, selecting the “Precipitation” option allows you to view the forecast of precipitation accumulation for an hour or three hours.
4. And the activation of the “Thunderstorms” option will allow you to select a number of relevant indices (CAPE, CIN, wind shear, lifted index, etc.).

In addition to displaying thunderstorms and precipitation, the Ventusky map has views of humidity, waves, snow cover, wind speed and gusts, and other features.

Weatheronline.co.uk is the authoritative British weather resource.

The Weatheronline.co.uk map is a British weather resource that displays both the state of the weather and its forecast for the next time period. The project has Russian-language localization, and a fairly wide range of functionality. You can choose both general continental maps and maps of individual countries, and view the dynamics of precipitation changes using them. The resource has a database of archives with which you can see how the weather has changed over the past periods.

1. To work with the resource, go to pogodaonline.ru (Russian version of Weatheronline).
2. On the left, select the desired country (county) to display, and on the right, select the display mode (precipitation).
3. The control panel at the top will allow you to read the weather news (in English).
4. And also view archives, analyze data on radars and satellites, enjoy the picture from webcams.

Rainviewer.com - analyzes and predicts rainfall with radar information

The website rainviewer.com created by Russian enthusiast Aleksey Schastlivy makes it possible to obtain precipitation data both for the current moment and for the future. The resource periodically scans sources of radar information. Processes the received information and places the received data on the site map. You can see both the current state of precipitation in the region you need, and the future forecast, for which it will be enough to click on the "Play" button on top.

Conclusion

The resources listed above display a map illustrating various rainfall online. The services are free and have a Russian-language interface. And can be recommended to everyone who wants to get detailed information about the real-time weather forecast.

In contact with

Weather online.
National Meteorological and Hydrometeorological Services wmo.int » short introduction

Below online maps weather animated maps of weather phenomena maps with animation of current radar observations interactive maps of temperature, pressure, precipitation, cloudiness, etc. Maps with weather information displayed in real time all atmospheric processes, or weather phenomena online. As well as "weather on the maps" with forecasts for tomorrow and 5 days in advance throughout Russia.

» for clarity, click and see examples and comparisons here

Animated maps, animation of weather phenomena.

Animated weather forecasts for Europe and the European part of Russia for 2 days. Animation of the forecast for 48 hours according to the regional model of the Hydrometeorological Center of Russia:

Video (press play) → animation cards precipitation and pressure according to the global spectral model of the Hydrometeorological Center of Russia, forecast for three days. Animation of weather phenomena in the districts of North-West Central South and North Caucasian Privolzhsky Ural Siberian Far East.

Prognostic cards.

Synoptic maps, surface prognostic maps with frontal analysis.

Unlike the forecast maps presented above, these synoptic maps are more informative. The meteorological elements plotted on the map will "tell" you about cold and warm fronts, precipitation zones, centers of cyclones and anticyclones (areas of low and high atmospheric pressure), directions of their movement, atmospheric pressure, wind, all precipitation, adverse weather phenomena, etc. ranges from 24 to 60 hours.

Temperature forecast cards for five days.

Storm cards..

Another "World Wide Lightning Location Network" map. It shows all the "lightning strikes" around the world, i.e. this is a global map. Its peculiarity is that the data is displayed in real time(in the menu you need to select "Real-Time"). Also, click the buttons "Cloud Overlay"(show cloudiness, located at the top left of the map) and "satellite"(top right). Thus, you can see more and other weather phenomena. The map is interactive, full screen mode is possible, move around, zoom in, etc. WWLLN Real Time Lightning Data.
True, there is a "BUT". To view the "new map" map, the authors added this note. "Known to work with Firefox 27, Chrome 33, Safari 7, and Internet Explorer 11. The maps are not supported for non-desktop browsing (i.e. phones or tablets)." Browser versions must be at least specified, plus some browsers for phones and tablets are not supported.
» lightning map - lightning map around the world in real time

Satellite weather, view of the Earth's surface from space.

Speaking about the weather on the map, weather maps, etc., users often mean online viewing, visual display of cloudiness, precipitation, and other atmospheric phenomena. So DO NOT forget to visit the satellite weather page afterwards. The content is useful and interesting and for many visitors who want to watch "weather online on maps" will be quite relevant.
» satellite weather, real-time images

The most important hydrometeorological phenomena.

• ✔ August 6th In the Omsk region, rain, thunderstorm, wind 15-20 m/s are expected.
• ✔ August 5-6. In Tatarstan on August 5-6 at night heavy rain, thunderstorm, wind 18-23 m/s. In the Magadan region heavy rain (August 6 on the coast is very heavy).
• ✔ 5th of August. AT Astrakhan region, Kalmykia wind 23-28 m/s. In the Stavropol Territory, Kabardino-Balkaria, North Ossetia, heavy rain, thunderstorm, hail, wind 15-20 m/s (up to 25 m/s in the mountains of Kabardino-Balkaria).
• ✔ August 4-5. in the Rostov Volgograd regions, Krasnodar Territory, August 4 in the mountains of Adygea heavy rain, heavy rain, thunderstorm, hail, wind 23-28 m/s. In Primorsky Krai, heavy rain, thunderstorm.
• ✔ August 4th. In the Murmansk region, rain, wind 15-20 m/s (up to 25 m/s on the coast). In the Arkhangelsk region, heavy rain, wind 18-23 m/s. Heavy rain in the Nenets Autonomous Okrug, wind on the coast 22-27 m/s. Heavy rain in Komi. In the Belgorod region heavy rain, thunderstorm, hail. In the Voronezh region, heavy rain, heavy rain, thunderstorm, hail, wind 18-23 m/s are expected. In Crimea, heavy rainfall, thunderstorm, large hail, wind 20-25 m/s are expected. In the Krasnoyarsk Territory in the Turukhansk municipal district, heavy rain, thunderstorm, wind up to 18 m/s, intense heat in the central and southern regions (daytime temperature up to 35 degrees). Extreme heat in Khakassia (temperature during the day up to 35 degrees). For Tuva, rain, thunderstorm, hail, wind 15-20 m/s. In the Tomsk region, heavy rain, thunderstorm, hail, wind up to 18 m/s are expected. AT Irkutsk region heavy rain (very heavy in mudflow-prone areas), torrential rain, thunderstorm. For Buryatia and Trans-Baikal Territory, rain, heavy rain, thunderstorm, hail, wind up to 18 m/s are expected. In the Amur Region, heavy rain, heavy rain, thunderstorm, wind 15-20 m/s are expected. Heavy rain in the northern regions of the Khabarovsk Territory. Heavy rain, thunderstorm in the central regions of Yakutia at night (long rain in Yakutsk). On the coast of Chukotka autonomous region wind 22-27 m/s.
• ✔ August 3rd. For Dagestan, heavy rain, heavy rain, thunderstorm, hail, wind up to 25 m/s. AT Perm region, Sverdlovsk region heavy to very heavy rain, wind 15-20 m/s. In the west of the Khanty-Mansiysk, Yamalo-Nenets autonomous regions heavy rain, thunderstorm...

Interview with the head of the most popular weather service in Russia.

To bookmarks

On November 30, Yandex launched a new version of its weather service. Now forecasts for winds, temperature and pressure can be viewed not only in the form of tabular data, but also in the form of an interactive map of the whole world. In this form, the forecast is given for the next 36 hours, and you can also look at data about the past at the same depth (for example, in Moscow).

TJ spoke with the head of Yandex.Weather Dmitry Solomentsev about new version service, methods for calculating forecasts, problems with weather applications on smartphones and notifications from the Ministry of Emergency Situations.

Dmitry Solomentsev

Dmitry Solomentsev: Two years ago, we launched our own weather forecast in Yandex.Weather - it was. Then they launched a precipitation map, which, to the nearest minute, helps to understand where it will rain. Now we present a global forecast that we make using another tool.

Previously, the weather in Russia was always consumed in tabular form. In Europe and America - on the contrary: they always have weather maps. It is much more convenient for them to look on the maps, where, for example, Hurricane Irma will go and which regions it will affect. We went in this direction and decided to try to give people the opportunity to consume the weather on the map.

The most popular layers - precipitation, air temperature, wind direction and pressure - will be the main part of this release. Air temperature, wind direction and pressure maps are now global due to the changes we made to this service. These worms that move on the map show which way the wind is blowing. The color scale shows how strong the wind is at any point.

When you point at one of the worms, does it show a vector?

This is a vector reduced to the main directions. That is, this vector cannot have an angle of 357 °, it has only eight gradations: for example, western or northwestern.

That is, you do not have enough data to show the exact direction - or are you just taking other people's data, and such information is not there?

No, we form the data ourselves, but people are so accustomed to consuming information - they are also used to the fact that the wind direction usually has the format “wind northwest”. And in the new version of Yandex.Weather, you can click on the map and see a pop-up icon with the current weather for any point.

The point of this launch is to give people a different look at the weather. Maps are a great way to find out what's going on in other regions. For example, “weather in the Moscow region” is a fairly large and frequent layer of queries. The Moscow region is large, and people are interested in what the weather is like in Moscow or Solnechnogorsk. A great solution is to see the map and see these pop-up icons around the places people are most interested in.

That is, the map always shows up-to-date data?

Yes, all information is real. Previously, the weather was calculated on request: a person came to the service, we showed the weather for him, calculated in Matrixnet. Now we store this in a precalculated form and are able to update the data due to the fact that we have built a chain that allows us to collect all the factors together.

It's like the Matrixnet search index: one factor changes and you recalculate?

Yes. All relevant tables are updated in a timely manner, and then the service works, which can search for these tables, issue point queries and cut tiles from them, from which the weather map is built.

What kind of gradation from below comes in increments of three hours?

This is the time scale. We have 36 hours in the past and 36 hours in the future, and we can see how the day follows the night, where it gets colder or warmer.

We worked on this scenario mainly because we see how people use the precipitation map: they often look into the past. For some reason, they are interested in how events developed. Therefore, we decided to do the same for all other parameters.

Any idea why people are interested in the past on a precipitation map?

We have a forecast, but everyone has their own intuition. Users can scroll back to see if we believe our prediction or not.

The data for the past is what the weather actually was, or is it your forecast? Do you keep information for the archive about what the weather was like in the end "actually"?

Not really. What you see in Yandex.Weather is data on a continuous regular grid. But weather observations are a bunch of heterogeneous sparse data. These are weather stations located approximately in the places where icons stick out on the service map balloons, and here the satellite flew by and took something off. All this data is sent to us with a delay. We begin to calculate the forecast not from the current moment, but from the past.

That is, at some point on the tape from the past, complete data may not yet be received?

Yes. Any data that is displayed on the service is the best combination of all data and forecasts at the moment. You need a model to spread this data around the world. Where there is a weather station, we show the fact, that is, most likely, how it really was.

All stations have a different update period. Some station may be late for half an hour - this is a good option. Sometimes it can be 3-5 hours late.

Do you have your own stations?

No, we don't have our own. We buy all data from different providers. Observation is not an easy thing, and we do not have such expertise. We cannot correctly install the station, or rather, we know how to install it according to GOST, but for this we need to hire a person who will service the station and take observations correctly. This is not our area of ​​expertise.

We collect all possible data that can be obtained from open sources or bought under a contract, such as satellite data. Based on our meteorological expertise, we make atmospheric models that predict the weather, and then these models get into the machine learning algorithm. This chain is our added value.

We buy radar data. On the precipitation map, you see circles, and in the center of this circle is a radar. This is a large construction site the size of a four-story house. He looks around him into the atmosphere for 250 kilometers. The distance depends on the terrain. We buy this data. Next, we train the convolutional neural network by showing it this data, that is, several consecutive shots. She predicts what will happen in the future.

Why does the radar look into the atmosphere? Studying whether there are clouds or not?

It detects the presence of droplets larger than 100 microns. Then we have to do the analytics ourselves - there are clouds or something is dripping from the clouds. The most trivial way: if the radar recorded a lot of droplets, then, most likely, precipitation falls there. A weather radar works exactly like an aviation radar: it scans in a circle and understands from the reflected signal whether there are drops or not.

And how many such radars are in Moscow?

There are two in Moscow: one in Sheremetyevo, the other in Vnukovo. This is understandable, because aviation needs accurate precipitation data most of all. We have made a service for people.

The precipitation map has been wildly popular in the summer during times of heavy rain and hurricanes. By the way, the Yandex.Weather service is evenly popular, taking into account seasonality, and in general the most popular weather service in Russia.

Yes, but not the only one.

It's good that it's not the only one. Precipitation map is what people pay a lot of attention to in summer. Maybe they will pay attention in winter, if there are heavy snowfalls.

This is an example of how we take other people's data, buy it and do something good with it that benefits people.

As far as I understand, now Yandex.Weather will show forecasts for the whole world - and these maps are primarily tailored for foreign users. What is their forecast consumption pattern and will it somehow be relevant for us?

First of all, we are rolling it out for Russian users. Why did I say about foreign? Because there is a significant difference in how they watch the weather and how we do it.

In Russia, there are frequent search queries, for example, “weather in the Tver region”, “where it is warm now”, “where it will not rain on the weekend”. People ask these questions in order to make decisions, such as where to go for the weekend. There is a request to find out the weather in several places at once. If you use a text interface, then first you go to one page, then you turn on another city, you go to another page. It's long and terribly inconvenient. The map immediately gives a visual answer to this request. It immediately becomes clear and simple.

If we talk about situations like "I want to decide what to do on the weekend" - this is planning with a horizon of three or four days. At such a horizon, the accuracy of the forecast is already low.

This is true. Low forecast accuracy is offset by product requirements. If we are talking about a precipitation map, then its horizon is two hours, but it gives an accuracy of up to 10 minutes. The user hardly needs to know what the weather will be in Tver at 14:10 on Saturday, he just needs to understand the picture for the weekend as a whole.

Predicting whether it will rain over the weekend in Tver is much easier than predicting whether it will rain in Tver at 2:10 pm four days later. This is a roughening of the conditions of the problem.

That is, if you have a request general, then you do not need exact data.

Not that they are not accurate, rather, not such detailed figures. But for the next day we have a forecast for every hour.

But this is not a 100% accurate prediction. How accurate is the precipitation map?

If we talk about spatial resolution, there is almost 1 kilometer.

But after all, from the precipitation map, and from the temperature forecast - I'm talking about the product as a whole - it's not clear at all that the service has a certain error, that this weather will not come with 100% probability.

Indeed, in order to manage expectations, we plan to add quantities such as the probability of precipitation to the forecast. The onset of precipitation in our service is now shown by an icon. We will add a probability to the precipitation icon - a number, for example, 40%. A person understands that it is possible to rain, he receives more information for decision making.

It is visually difficult to show that the forecast is becoming less accurate, and besides, it seems that our users already know about it. And, of course, a lot of resources are thrown into making an accurate prediction.

Returning to patterns. There are frequency requests and there is a demand for visual maps, while there is a pattern to go to the “Moscow. Weather for today” and be content with this information. Perhaps you want to make this pattern change - the same requests indicate the need for such a product with a card.

Yes, definitely. But the point is not that the audience is changing, but that we are starting to reach new audience segments that our usual forecast is not enough for.

70% of our audience is enough to see the current weather, the weather for the evening or morning. For the people I spoke about a little earlier, this is not enough. We are improving the script for them. It's not like we're fundamentally changing how we use the weather service, although that would be a cool ambitious goal.

We should always start small, so we simply suggest that people who are uncomfortable right now make it a little more comfortable. Athletes whose sport is associated with the wind fall into this category. These are windsurfing, sailing, skydiving, paragliding. For these athletes, a wind map is essential and they are now finding other specialized resources to use. And we have an up-to-date weather forecast. This is a good chance to show them that we can do it too, and they can stay on our service.

Why do I need to know where lightning strikes now?

On the one hand, lightning is a danger, and it would be nice to know if they strike somewhere far away from you, on the other hand, it's cool to see such things on the map. (shows) This blue zone has very heavy rainfall. When people look at the forecast, they can understand whether this front is carrying thunderstorm activity or not. This allows you to make decisions, although I will not claim that this is a 100% utilitarian thing.

How are lightning strikes recorded? Radar?

A very similar thing to a radar, but not quite. These are much cheaper sensors that are installed on Earth. These sensors register the electromagnetic flash that occurs at the moment of a lightning discharge. A network of these sensors is spread across the Earth - they determine, by means of triangulation, the approximate location of the impact.

What else is in the new service?

For people whose well-being depends on the weather, we made a pressure map so that they can understand when an area of ​​low or high pressure will come to them, and how long it will stay there. With the help of maps, you can also look at the pressure that is imposed on the direction of the wind - cyclones and anticyclones. But this is a completely different story.

Meteorologically dependent are, as I understand it, older people who like to measure pressure in millimeters of mercury, and for them information in numbers is more understandable.

Yandex employee Zalina Bogazova: In general, evidence-based medicine denies the existence of such a diagnosis. There are different categories of people and conditions that can really be associated with the weather, but these are different things, and not some kind of big meteorological dependence. This service is made for these people.

Dmitry Solomentsev: Astrology, moon in Capricorn, everything.

Yeah, that is, Yandex.Horoscopes. (laughs)

You really laugh. Once there was feedback: we temporarily removed the moon phase in the Yandex.Weather application, after which several girls wrote to us: “How do we determine when to go to get a haircut?” Then I learned that the phase of the moon determines how fast hair grows.

Back to the subject of weather maps, the main thing we want to change with the current release is to give a different way of consuming the weather to those who have not yet had such an opportunity. By the way, these maps were made possible thanks to the technology of “digesting” the weather before users come to us, namely, by the data update trigger.

Now I don't understand anything. Do users always come to a ready-made forecast?

Previously, we calculated the forecast directly for each specific user. How did it happen? Coming to the service, a person brings with him user data, for example, latitude and longitude using GPS. We combine other data with this data, such as a weather forecast using a 1-2-3-4 model, and make a long vector. This vector comes at the input to Matrixnet, and Matrixnet gives the forecast as a search query.

Now we keep offline factors pre-calculated. This allows us to cut global map. Previously, this would have been too costly in terms of resources. The short moment of using Matrixnet, which happens very often, heavily uses the processors on the servers. Now, thanks to offline weather forecasting, we got rid of it and now we can give maps and calculate a global forecast.

What is offline weather calculated on? This is a large amount of data to keep in the cache. Should it be constantly calculated?

This is a difficult technical task. We have several data providers: they give global weather forecasts made by different providers - European, American, Canadian meteorological agencies. We have our own model, which we consider on top of these models. This allows you to increase the spatial resolution. Their cell size is from 20-30 kilometers, our size is 2 by 2 kilometers: in densely populated areas, we always use our model. And all this data comes with some intervals.

How it happens at a low level: a volume of data has accumulated that will help increase accuracy, new satellites and stations have arrived, these models have been recalculated. We uploaded them to ourselves - we calculated our model. Next, you need to update the weather condition. This all falls into a long process that prepares the factors for the machine learning formula and then applies this formula. All this allows us to divide the region into regions, to carry out calculations that we could not carry out before. Then, as I said, everything adds up to a service with the most up-to-date data. And thanks to the way this service is implemented, the weather forecast is given almost without any delay.

You made a reservation that you do not make stations, you are not engaged in production, you are not aware of how it works. That is, you don’t consider b2b here at all?

We are aware of how it works, we just don’t do production: b2b in the weather can be quite different. We already have several requests for how this data can be used. The first request comes from Yandex itself: it is weather and hyperlocal data for advertising campaigns. Obviously, some banners click better in the heat, and some in the cold.

This can be approached from two sides. On the one hand, display banners correctly, taking into account the current weather in the ranking. on the other hand, you can, for example, give customers the opportunity to customize their advertising campaigns for certain weather conditions.

But "Yandex.Direct" does not yet allow you to do this?

We are currently in a closed test mode with several companies and agencies in order to determine the final functionality of customizing advertising campaigns based on the weather. There is hardcore b2b - for example, aviation: this is no longer about us. But agriculture is a sector closer to us.

Aviation is still connected with safety: you will not give a 100% guarantee.

Nobody gives a 100% guarantee.

Aviation radars give a 100% guarantee.

No. Firstly, radars, like all other pieces of iron, cannot predict. Secondly, they have observational errors, there are no ideal instruments. Here something could be reflected from a cloud, or it could be from a tall building ...

I rather mean 100% responsibility. Data providers, on which people's lives depend, have a high responsibility. Based on them, conclusions are drawn whether the plane should fly or not. You will not take on such responsibility.

Not true. This is a subtle story. Now we turned to those data that affect the lives of people on earth, but aviation needs slightly different data. In particular, all aircraft carry weather instruments. This is the largest network of observations that measure weather parameters. We haven't looked at this data, and right off the bat our technology cannot be adapted to predict these parameters.

And to use data from such aircraft?

You can use it, but how relevant it is for people who walk the earth is not very clear. And aviation is a very closed market, where there are players of the level, for example, Roshydromet, which has its own radars.

And the guys from agriculture have already contacted us. There are several companies that provide services for agricultural producers and large agricultural holdings. They are very interested in the weather forecast: they are especially interested in the fact that we have radars, that we can predict the accumulated number of millimeters of precipitation, that we can take the stations that they buy and install for themselves, and refine the forecast for these stations. This is generally a good thing, but so far we have not spent any resources on entering this market.

What are these companies willing to pay for? Here they have a weather station: let's say they want you to be able to predict the weather for the week ahead based on their data using your model. And you have, say, 18 sources from the largest companies and more data - why don't they just look at your weather for free?

It does not mean a replacement, but an addition. Often agricultural fields are located in areas where the frequency of data is not very high, so adding one weather station to good location can be of great benefit. Second, they want a ready-made service. Yes, this is how they work now: they open websites and watch the weather. But to make the process more convenient for them, you need to introduce certain triggers: wind strength, lack of precipitation and the level of the desired temperature. These are favorable conditions, for example, for fertilizers. Conventionally, a person needs to color the calendar - when it is possible to plant, when it is impossible - and periodically notify him that something is changing.

Yandex.Weather widget on Android, standard weather app on Apple Watch, precipitation map in Yandex.Weather app on iOS

Let's talk about your application. I now have a pain due to the fact that on iOS the weather is displayed by default, which they take from the American The Weather Channel. It seems to be a normal source, but obviously not relevant for Russia. I see strong discrepancies between what The Weather Channel and Yandex.Weather gives. On Android, I replaced the weather application with Yandex.Weather, there is even a widget in notifications, but on iOS, Yandex.Weather does not integrate with the clock.

Many users tell us about watches. This is exactly what we are working on right now. We calculated internally the costs of developing a good quality application for wearable devices and realized that the costs may not pay off. On the other hand, there are push notifications, which are much cheaper to develop, will be broadcast everywhere - on Apple Watch, Android Wear. It was on push notifications that we made our bet: this is how we can directly contact the user, but we try to maintain a balance without annoying people.

And what kind of notifications do you mostly send?

There are two main classes. The first is the Ministry of Emergency Situations. We broadcast dangers from the Ministry of Emergency Situations, which after that began to come every day. The second is notifications of a sudden change in the weather, for example, the beginning of precipitation after a long absence.

It probably doesn't happen in manual mode, does it? We here at the newsroom often send notifications by hand when something unusual happens.

The only thing that we send manually is alerts from the Ministry of Emergency Situations, with which the IT system could not be adjusted. We just look at their website and notify further. Everything else is automatic.

We're currently testing notifications with a rain map, like "It's going to rain in 10 minutes" or "It's going to rain in 30 minutes". Around Wednesday, we start to announce the weather for the weekend, when we are sure of the forecast.

There is a delicate balance here, you need to understand when to send and not send a notification. This is a machine with a lot of settings, and we are experimenting with it carefully. We try to involve people, but rather look at some long-term metrics, because a good notification in our understanding should not force a person to open the application. The notification should contain approximately all the information, this increases loyalty. A person remembers that Yandex.Weather turned out to be useful and warned in time.

You said that your service is the most visited in Russia. But these are desktop statistics. What about mobile apps?

AT mobile application we have fewer people than on the desktop. Mobile Internet younger in terms of the audience and, as a result, there are people whose life depends little on the weather.

Our desktop audience is older. They need to know the weather for the weekend, they all have dachas. We have a stable and established audience on desktop. AT mobile version The site has a similar situation - it is one-third of the desktop audience in size. The audience in our applications has been growing since we updated them two years ago with the launch of Meteum. Grow organically. Right now in Google Play in terms of weather applications, we are top 1 in Russia, but I could be wrong.

These applications have become popular in large part because we have an hourly forecast and rainfall map that no one else has in this form. We noticed this on the metrics at the beginning of this summer. Now we are trying to visually improve the application, because before that it was very simple. Become more likable.

Why does the Ministry of Emergency Situations screw up so often? Whose data are they working on?

They use Roshydromet data. Why they often have a false positive of the warning system - because they have the task of overdoing it. This is the right strategy for them. The value of warnings is certainly declining, but they cannot fail to warn. Why they often make mistakes is not a question for me. Hazard forecasting and normal weather forecasting are two different types of science.

We may not see a hurricane in our Meteum, because we have never been trained in such a thing. They occur rarely, especially in the central part of Russia, and we do not specifically single them out from the whole mass of weather events.

It's strange: you highlight lightning, but not hurricanes.

We define lightning by measurements, but hurricanes or tornadoes need to be predicted. We try to give the user information for making everyday decisions. Roshydromet and the Ministry of Emergency Situations have the infrastructure that we do not have: airplanes, radars, exit points. Some of this data we receive, but not all. Having only all this data, and maybe even more, one can begin to predict dangerous phenomena well. But this is not the task that we set ourselves as a service.

That is, you do not set such a task and do not have the technical ability to warn users about dangerous phenomena? Can you say that the rain will end in 10 minutes, but you can't predict strong winds?

A strong wind can, but a hurricane in the direct sense of the word is not.

If the situation that was in the summer of 2017 repeats, will you be able to warn users about this?

We predicted a very strong wind that day, up to 28 m/s - this is approximately what was recorded. But it is impossible to say something definite about this, for example, “A hurricane is coming,” because we cannot track such formations in our model. We accurately predicted a strong wind, but we could not say what it was connected with, and that day we did not send warnings about it. We are working on sending pushes using more parameters than now.

You would like to warn users about such things, but you can't. There is a noticeable discrepancy here: you do not want to take responsibility?

The field of warnings about emergencies is very overregulated - only the Ministry of Emergency Situations can submit them. Everything that the Ministry of Emergency Situations gives, we broadcast verbatim and without changes. But if we have a strong wind with gusts of 28 m/s in our forecast, then why not warn about it.

The situation with more frequent notifications from the Ministry of Emergency Situations will grow into that there will be less trust in them over the next months, and by the next rainy season, trust will drop to a level where no one will respond to the alert. Perhaps they will draw conclusions and stop sending so often, and at some point a second situation may occur. In this case, I want someone to still send this warning. Will this source be "Yandex" - that is the question.

All predictions are wrong. No matter how perfect technologies are, and no matter how many means of change there are, all forecasts are wrong. If you need to understand what will happen in Moscow within a few hours, then you need to go to our map, including precipitation, look and understand what will happen.

OK then. Now, if you are not successful with the Ministry of Emergency Situations ... I know that Yandex works well with Moscow city services. If weather data is sent through such structures, then you can warn them, for example, that city service cars need to leave and carry out work to prevent floods.

I can't say anything about this. Yes, we would like to do this, and there is interest from a certain side, but so far there is no concrete implementation.

Not yet, but technically and theoretically feasible?

Quite. The utilities are the ones we can help. We can predict heavy snowfall. Or black ice is business logic on top of weather data. Ice is determined by the presence of precipitation during the previous day, humidity, wind, and temperature. All this is predicted, collected in a formula, and she says whether there will be black ice or not.

Do other services use this weather data?

Yes, for example, they are in Yandex.Navigator.

Do they predict traffic jams?

No, it is not used there yet, as far as I know. We are planning to implement weather in many places. It seems to be relevant on Yandex.Maps, but now the main effort is focused on incorporating the weather into advertising products.

If you are going to relax in nature, or even planning a trip abroad, it will be useful to know the weather forecast in the region where you are going to go. This application will help you display the weather around the world on Windy Maps, see the temperature and weather conditions online anywhere in the world very easily.

Detailed satellite and vector map with weather forecast for Russia — Windy and OpenWeatherMap. Weather vector map of Russia and the world. Weather forecast for Russia on the map. World weather on the map. Weather in all regions and cities of Russia on the map of Windy. Interactive weather map in Russia and the world.

Description

Change the scale of the image and select the required region or city and study on the map the current weather for a given region at the moment. The symbols indicate: rain, sun, clouds, etc. Thus, we get complete information about the weather in the place of interest to us.
But the most amazing thing, using this application, and selecting the satellite mode, you can see the clouds online, their position at the moment right on the map!
If you select a specific city and click on special character, then in the window that opens you will see detailed information with windiness, humidity and even a weather forecast for several days.