The coronavirus pandemic has been changing the way we live. In a scenario of social distancing, the value of supply chains grows significantly. Even with reduced mobility, food, inputs, and raw materials need to reach their destination. Weather forecasting is a great ally to guarantee these logistics. LABS reached out to two Brazilian startups that offer accurate weather forecasts to safeguard transportation and food production: Pluvi.on and Agrosmart.
Agrosmart is a digital agriculture platform that provides irrigation intelligence and weather forecasts for farmers. The company saw a growing need for weather forecasting in the countryside during the pandemic. “In the field, with social distancing, there are less available workers. It is more important to understand the weather on the farm to make better decisions,” says CEO and co-founder of Agrosmart, Mariana Vasconcelos.
And this demand is not only due to the pandemic but also because adverse climatic events that Brazil has been experiencing in recent days: a “bomb cyclone” hit the southern region of the country (a major food producer) where it left victims and destruction in late June. The effects of the phenomenon reached the southeastern region of Brazil, where Pluvi.on managed to alert about them four days in advance on the south coast of São Paulo, Brazil’s most populous state, where it brought rains that exceeded 100 mm (almost 4 inches) in 24 hours.
According to the World Meteorological Organization (WMO), the reduction in the number of flights due to COVID-19 also affected weather forecast data, which are in part fed by aviation. According to the organization, the impact is already noticeable mainly in data about oceans. “With the reduction in air traffic that facilitates observation, in regions like Africa, it has become impossible to have complete data on certain areas,” says WMO, in a report on the website UN News. Places where classical observation instruments are used are the most affected, although the analysis of this impact is not yet accurate.
Diogo Tolezano is the CEO and co-founder of the Brazilian weather forecast startup Pluvi.on. He believes that the pandemic might have some impact on the loss of data provided by commercial aviation. “We are talking about mathematical models and all data with quality are very important for the response and accuracy of the forecast”, he explains. Tolezano points out that in developing countries, such as Brazil, much of the observation network is manual and, with confinement measures, these instruments are no longer monitored, which can also affect the work.
To make predictions about the weather in Latin America is a daunting task. In traditional services, errors are common, largely because the southern hemisphere has fewer satellites than the northern hemisphere. As Tolezano explains, “the further you look, the less accurate it will be”. It is a problem in the southern hemisphere and especially in Brazil, as there is no satellite for meteorology in the country. “We have a very small radar network for a country of its continental size. The United States has at least four times more radars than Brazil “, he says.
Although federal and municipal authorities have data, access to them is bureaucratic and interpretation by ordinary citizens is even more complicated, according to the CEO. “At the end of the day, precisely because of the lack of these data, the response here for Brazil ends up being less precise”. In the United States, the average accuracy of rain forecasts is around 90%, according to Tolezano. In Brazil, the rate drops to 70%, following the global weather forecast model that the National Oceanic and Atmospheric Administration (NOAA) runs.
Startups search for data and use different software for a more accurate forecast
As data were missing, Pluvi.on went after them and developed its own weather station in Brazil, since the cost of an imported weather station would make the acquisition in large quantities unfeasible (the Brazilian station costs a fifth of an imported station, Tolezano said).
With the station, the startup can collect data in real-time. The company installs a network of stations in the region where it operates and uses these data to correct and improve projections. Based on better predictability, it is possible to deliver more customized alerts with the longest advance time, according to Tolezano. “We are managing to anticipate these extreme events at least three days in advance. We deliver more predictability, more security, both for companies that have assets at risk and for the population”.
The station, however, does not forecast the weather. The forecast is a mathematical model. What the station does is help the startup to confirm and adjust this model to what happened in a place, on a street, in a company. With that, the company develops its algorithms. And the difference is in the accuracy of the data.
“Forecasting models in Brazil usually project a different weather prediction for every 20 kilometers (12.4 miles), that is, you sometimes have even a group of towns with a single weather forecast. And these models do not consider much terrain effects, physical features that change the weather forecast in the city a lot”, he explains. Pluvi.on runs a different forecast model every 2.4 km (1.4 miles), which roughly refers to a different forecast for each neighborhood in large cities.
Pluvi.on is present in 12 states in Brazil with more than 300 pieces of equipment. For each client, the startup has a different alert system. The company has as a customer one of the largest railway companies in Brazil, Rumo. In addition to rain and wind alerts, it gets a specific landslide alert in railroad areas.
Another major customer is Concessionária Tamoios, a highway concessions company responsible for a stretch of road which connects São Paulo to the coast, for which the startup created an alert mechanism also looking at landslides. For the population of the east side of São Paulo city, however, the company uses only a warning of heavy rain.
Tolezano had to readjust the expansion path planned for this year due to the pandemic. “As part of our service takes place in the field, we had a natural impact on the speed of growth that we had planned for 2020”, he comments. The company took advantage of the period to advance the development of roadmap and test other, more digital business models.
“Soon we will launch our app for the population, with a lot of focus on rain forecasting and monitoring and an API for developers and businesses to automatically plug-in our services, with the possibility of upgrading if they want more personalized and accurate information, with their network of stations”, he explains.
Agrosmart, besides Brazil, where the company works over 680,000 hectares (1,680 acres), is present in 9 countries, mainly in Latin America: Argentina, Colombia, Peru, Guatemala, Costa Rica, Mexico, Bolivia. The company also has operations in the United States and Israel.
The agritech provides algorithms for farmers, but also allows them to use data and create any alerts to what they already know, “to take advantage of the consolidated knowledge of years of farming together with the agronomist”, she explains. The startup also allows other companies to exchange data, generating integrated intelligence with farmers.
According to Vasconcelos, Agrosmart works throughout the agribusiness chain, from seed, pesticides, and R&D companies, from experimental fields in the best use of products to monitoring supply chains. “When there are food companies that buy from generally small producers, we help to understand this chain, help the producer in that chain to create climate resilience”, she explains. The agritech also operates in the financial sector, with data that help banks and insurance companies to measure risks.
Agrosmart’s weather forecast brings a 7-day advance solution with a 7-kilometer (4.35 -mile) grid on the farm. The forecast is applied to several products, such as irrigation. “We use the weather forecast to be able to forecast demand for the following days, whether there will be demand for irrigation or not, to find out if conditions are going to be favorable for applying inputs or entering machinery, or starting planting.”
With foreign investors concerned with deforestation in the Brazilian Amazon, the weather station is also able to predict the risk of bushfires, which, according to her, was highly demanded in the region. It is still possible to predict crop pests such as coffee rust and fungal diseases in potatoes, for example. “We predict how the conditions to develop the disease will be in the coming days to assess the risk of the disease. For coffee plantations, for example, we launched a model through which we can assertively know whether there will be rust or not 15 days earlier. It is a model that is derived from the weather forecast”, she says.
For other crops, Agrosmart has customers who use weather forecasting to coordinate logistics. To find out if it is possible to harvest and fetch the grain, for example. Since, if it rains, the grain will not be dry to be removed. Thus, it is possible to plan a logistics route to collect raw material.
With two investment rounds, the startup raised $9.8 million in seed and Series A rounds and seeks to be the leading Latin American digital agriculture platforms. “We were able to position ourselves outside the country, which brings a strong message of sustainability and production and we managed to do that from Brazil. We are sending our Brazilian technology abroad,” says Vasconcelos.