We think of climate change as a looming disaster. Yet historic shifts in the way rain falls, explained in shocking clarity through USA TODAY’s reporting, make clear the danger is already here. “Downpour” features an analysis of a century of precipitation records from the National Oceanic and Atmospheric Administration and a unique collection of snow and rain extremes computed by a private climate researcher. These revealed stunning increases in intense rainfall in vast sections of the country over recent decades. Through brilliant visualizations and innovative musical auralizations of rainfall data, “Downpour” provided deep contextual grounding to understand this year’s weather.
The general rise in temperature over the past century has become fairly well known in the collective mindset. The increase in rainfall is a phenomenon less widely understood. This explanatory investigation empowered readers to make sense of weather changes that have tangibly altered people’s lives. Our approach was unique in the types of data points we assembled in one place, and scientific sources told us it was both sound and compelling.
We achieved this kind of impact through the careful selection of numbers:
- At some point over the past three years, 27 states – all east of the Rocky Mountains – hit their highest 30-year precipitation average since recordkeeping began in 1895.
- A dozen states, including Iowa, Ohio and Rhode Island, experienced five of their 10 wettest years in history over the past two decades.
- Of 285 weather stations in the continental United States, 44% get at least one more top rainfall event per year now than they did three decades ago. That means what used to count as their top three wettest rainfalls of the year now happen at least four times a year.
- Intense rainfall events can cause three times as much fertilizer runoff as other precipitation, contributing an outsized share of pollution associated with algae blooms in the Mississippi Basin and Gulf of Mexico, based on our analysis of one Illinois watershed.
- Among cities with sewer systems designed to discharge untreated waste into streams during heavy rains, 97% have experienced an uptick in both annual precipitation and extreme rainfall over the past 30 years.
- These cities are ill equipped to pay for sewer upgrades. The median household income in the 728 cities and towns with these vulnerable sewer systems is $45,520, compared with $67,520 nationally, and the poverty rate is 50% higher than the national average.
This project was powered by analyses in R, QGIS and ArcGIS and came to life via innovative visualizations and auralizations.
Rainfall changes. We used R and NOAA annual data to isolate record-setting precipitation years and average precipitation. For changes in high-intensity rainfall, we used statistics from climatologist Brian Brettschneider. With each U.S. weather station, he identified rainfall thresholds exceeded, on average, just three days a year from 1951 to 1990. He then calculated how often each station recorded that amount 1991-2020. There were widespread increases in downpours.
Vulnerable sewers, residents. Some sewer systems overflow into rivers when storm drains fill. We used QGIS’ inverse distance weighting algorithm to interpolate Brettschneider’s weather station data into polygons, joining them on coordinates of overflow systems. Nearly all such systems had rising rainfall. The American Community Survey showed these cities averaged higher poverty rates and lower incomes, making infrastructure upgrades a struggle.
Increased pollution. We identified one Illinois watershed with continuous, long-term USGS nitrate data, then scraped watershed precipitation levels from Daymet, an online dataset from Oakridge National Laboratory. About 10% of rain events contributed to 33% of pollution.
Hearing rainfall. Composers from Full Sail University produced musical representations of state-level data. They combined traditional instruments, sampling and generative audio to embody annual rainfall. We matched each piece to an animated bar chart of underlying rainfall data with a custom player that lets users experience tones and bars simultaneously.
Changes near you. We used Datasette, Svelte, the Mapbox geocoder and Turf.js to take a user’s location, find the nearest weather station and climate division, and present a century of local precipitation highs, lows and trends.
Guessing trends. We presented a half-drawn line chart showing rainfall over time and let readers draw the remainder for recent years. The interactive then revealed the actual trendline.
What was the hardest part of this project?
The range of separate data analyses and visualizations this project called for was daunting. We set out not only to establish that climate-driven rainfall extremes are real, but also to quantify impacts on humans and ecosystems. We wanted data unassailable among climate researchers yet approachable by any audience.
Choosing the right weather measurements was key. The United States has many distinct climates, and accounting for that is a challenge. Reporters read numerous academic studies and consulted with top climate scientists who specialize in precipitation, formulating with their guidance a unique analysis that would measure how many record wet years and dry years each state had since 2000. The analysis also measured changes in average annual precipitation from 1895 to 2020 by state and in NOAA’s 344 climate divisions.
No database lists extreme precipitation events, and there is legitimate scientific debate about what constitutes “extreme.” To break through that obstacle, USA TODAY worked with a dataset assembled by climatologist Brian Brettschneider. Each U.S. weather station was measured against itself for two time periods, 1951-1990 and 1991-2020. This revealed numerous statistically significant increases in days with heavy downpours.
Another hurdle was missing data. Capturing the impact of heavy rainfall on sewer overflows demanded nationwide coverage. Yet while rain and snowfall amounts are collected each day, hour and minute at weather stations in hundreds of U.S. cities, big swaths of the country have no station. To fill data voids, reporters used an inverse distance weighting algorithm in QGIS to spatially interpolate Brettschneider’s weather station data. It enabled us to say that almost all cities with combined sewer systems are in areas that are having more days with heavy rainfall. Similarly, our analysis of how heavy rains wash polluting fertilizer into streams called for watershed-level rain data interpolated by researchers at Oakridge National Laboratory.
What can others learn from this project?
This project affirmed how crucial it is to enlist experts in the field to develop a sound methodology and choose appropriate data, especially when approaching a topic as complex and controversial as climate science. Your stories will have far more resonance, credibility, accuracy and nuance if you leverage these sources from the beginning.
The scientists we consulted guided us toward the right data source for the right level of geography. We learned that NOAA’s annual average rainfall figures are considered the most accurate for portraying change by state and at the level of climate divisions, which are sub-state regions drawn by NOAA scientists. For high-intensity rainfall events, climatologists encouraged us to talk with Brian Brettschneider, an Alaska-based researcher who shared with us his unique, apples-to-apples calculations on rainfall amounts that previously qualified as extreme for a given location — and that have become more common in recent decades. His findings are rooted in U.S. weather stations, which, unlike climated divisions, are observation points at a single latitude and longitude.
To correlate fertilizer runoff pollution with downpours, we started out by examining data for the weather station closest to each waterborne U.S. Geological Survey nitrate gauge. But experts noted that even if a station is nearby, it might not actually be within the same watershed. We instead turned to (and scraped) the specialized Daymet dataset, compiled by Oak Ridge National Laboratory, which interpolates weather readings onto a uniform grid of North America. These data enabled us to connect rainfall levels to each square kilometer of the watershed we analyzed.