Analytics for Weather Patterns
In this article, we will take a closer look at how weather analytics helps us contextualize weather patterns. In a previous post we outlined our approach to systematically detect anomalies. While important local context is provided for the anomalies detected, many scenarios require more data to contextualize them. For instance, is the current heat-wave part of a longer trend of a hot and dry summer this year? Statistically what does normal look like? How has it changed over the years? We will outline some of the functionality that enables us to answer these questions with data and visualizations.
Trends and Comparisons
In early March 2023 our anomalies detection picked up a very obvious heat-wave affecting parts of Argentina and Uruguay. Clicking through to the ‘View Details’ of one of the locations, Buenos Aires Airport, we have a few options that will provide more context to the heatwave.
Clicking on ‘Year-over-Year’, we observe the anomaly as a departure from normal in terms of raw temperature units (weekly means are a staggering ~5C above normal). With a few tweaks to the interactive page, we also form a snapshot of this summer, where we see that the temperatures have been well above normal almost every week between Dec 2022 and Mar 2023. It is also easy to observe the accompanying much lower than normal rainfall every week, hinting that drought conditions could have contributed significantly to the late summer heatwave as well.
How are summers themselves changing over the years? There are a few ways of answering these questions. Altering some of the parameters on the page, we will observe the week-over-week change across two twenty year periods across the last 40 years. It is interesting to see that early summer appears to be warmer but late summer seems to have cooler nights. Yet another way is by analyzing ‘time series’ data. Clicking on the ‘Time Series’ link , we observe how parameters have changed over time for summer (Dec-Mar). There is clearly an increase in maximum temperature but not a noticeable change in the minimum temperature.
Historical Trends
In the previous section, we were able to contextualize current weather patterns for a given location with respect to the latest climatology (past 40 years). However, we are also interested in understanding longer term changes happening to the climatology itself - what are the climate shifts that we have observed? For most parts of the earth, many of our building codes, emergency response plans and infrastructure have been built around the climate that existed several decades ago. How have they changed?
Continuing with the same example of the Buenos Aires area (instead of just a single airport location) we can see some of the changes to typical summer temperatures. We plugged in a value of 33C (p90 summer time high during (1961-1990) and noticed that for the period 1991-2020, this value is likely to be exceeded for an additional 1.3 days during Dec-Mar. It’s easy to see that for 2001-2022, the corresponding number is +4.7 days. For the minimum temperature, the data suggests that not much has changed.
Another example is the increase in ‘tropical nights’ in Paris.
In all our analyses, we remain committed to showing the quality of our underlying data and here we share the ‘coverage’ metric for each time period. While US locations often have very high coverage, non-US data has gaps. We aim to address these gaps soon by providing an option to use ‘’reanalysis’ data.
We are still very early in trying to productionise our understanding of shifts in rainfall distribution and rainfall rates and we are looking for help. If you are a climatologist or data scientist and can think of relevant methods or find glaring gaps please reach out to us.
Analytics as Building Blocks
Most of the functionality that are described above were powered by APIs . /timeline API was used to build the interactive visualizations. It is also possible to download the raw data (in csv / json) of the interactive analyses. We believe that this makes it possible for customers to build custom visualizations or directly plug in the data to their systems , whether it is a spreadsheet or a program. We expect the APIs to evolve as we get more feedback from potential customers. If you want to try our API or sample our data please get in touch.
Interactive, Referencable and Usable
We have observed that it is possible to build more context around weather trends and events using the interactive, referenceable (‘Permalink’ buttons) and readily usable ('Download' buttons) visualizations. In addition to such capability being directly useful in communicating climate change, raw data about trends and comparisons could prove to be important in various planning and operational scenarios as well. We believe that making weather analytics accessible and interactive will enable builders and planners to uncover weather related insights more efficiently.
Extremes
We have described a few ways to systematically detect and understand anomalous weather trends at a local level. These events exert a ‘chronic’ stress on our society and have medium term implications. Then there are the extreme weather events - the one in a hundred year rain that causes roadways to be washed away or a heat-wave so intense that causes trains to be stopped. These have short-term implications and can be life-threatening or very disruptive to our normal lives. With climate change, we know that the nature and frequency of the extreme events are changing. In a future post we will outline our efforts in trying to understand the impact of extreme events on a regional basis. Stay tuned!