The global death toll from the coronavirus pandemic could triple by year’s end, with an additional 1.9 million deaths, while a fall wave of infections could drive fatalities in the United States to 410,000, according to a new forecast from the Institute for Health Metrics and Evaluation at the University of Washington.

The estimate reinforces warnings by many experts that cooler, drier weather and increased time spent indoors could boost viral transmission in the Northern Hemisphere surge this fall and winter — something typically seen with other respiratory viruses.

The institute’s forecasts were influential earlier in the pandemic in guiding policies developed by the White House coronavirus task force, but they have been criticized by some experts as projecting further into the future than can be done reliably.

The U.S. death toll from covid-19, the disease caused by the coronavirus, now stands at 183,000, according to health data analyzed by The Washington Post. The IHME model projects that under the most likely scenario, 410,451 people in the United States will have died by Jan. 1.

The best-case scenario is 288,381 deaths and worst-case is 620,029, that model forecasts.

The scenarios pivot on human behavior and public policy. The best-case scenario would result from near-universal mask-wearing and the maintenance of social distancing and government mandates limiting the size of indoor gatherings. The worst-case scenario assumes that people and their communities stop taking precautions.

“It’s easy given the summer lull to think the epidemic is going away,” Christopher Murray, director of IHME, said Friday on a conference call. But there are “bleak times ahead in the Northern Hemisphere winter, and unfortunately we are not collectively doing everything we can to learn from the last five months.”

In recent weeks, the daily numbers of U.S. infections and deaths have gradually tapered, following an early-summer surge driven largely by community transmission in the Sun Belt. But experts warn that viral infections typically spike in the weeks after school resumes and when colder, drier weather can help viruses stay viable longer and spread more easily.

One key insight in recent months is that the coronavirus has a much harder time transmitting outdoors. But as temperatures drop, people will spend more time inside.

Few models forecast as far into the future as IHME does. Many do not project further than four to six weeks ahead.

“Beyond that, it’s all conjecture and guesswork because there are so many factors we just can’t predict and factors about transmission that truthfully scientists don’t understand very well yet,” said Jeffrey Shaman, an infectious-disease expert who leads the modeling team at Columbia University. “What happens the next few months really depends on what we do as a society the next few weeks.”

Murray defended the longer-range projections, however, saying they are intended to help government leaders and the public make decisions now to avoid the worst-case scenario. The new forecasts represent the institute’s first attempt to model the global spread of the virus and show the daily death toll peaking in mid-December at about 30,000, a more than fivefold increase from the current daily fatality numbers.

The specific timing of the second wave matters less, he said, than the fact that it is likely to happen — and that many deaths could be avoided if precautions are taken now. The IHME projects that with rigid adherence to social distancing and mask-wearing, 770,000 lives could be saved globally by year’s end.

One of the most useful tools for public health officials following different projections has been “ensemble models” — an aggregation of several of the country’s leading models to make a more reliable forecast. One of the most heavily used ensemble models — assembled by Nicholas Reich, a biostatistician at the University of Massachusetts — shows deaths declining slightly but maintaining a fairly high rate of more than 5,000 a week for the next four weeks, with a likely total of 205,093 before the end of September.

But the recent declines in cases and deaths appear to be prompting a relaxation in behavior that could quickly drive those numbers back up. One model that takes into account the extent of social distancing, using cellphone data, forecasts the daily number of deaths increasing over the course of this month from roughly 1,000 deaths a day to almost 2,000 by end of September.

“We’ve been seeing mobility increase the last few weeks,” said Lauren Ancel Meyers of the University of Texas at Austin, who developed the model that factors in mobility.

In Texas, for example, mobility data shows more people staying home in June when hospitalizations were going up at alarming rates. But since August, that mobility has accelerated.

“One of the challenges is as soon as things look good in the community, it’s tempting to say the virus is gone,” Meyers said. “What we really should be thinking is: How did we get to this better place? By being cautious and vigilant. And instead of relaxing, we should focus on the things that worked.”

About 910,000 people globally are known to have died of the virus, according to IHME — a tally higher than the 865,000 used by the World Health Organization. But most experts agree the true death toll is probably higher, with many people dying at home or without having been tested for the virus.

The “most likely” scenario produced by IHME is for 2.8 million global deaths by the end of December. The best-case scenario would about 2 million and the worst-case about 4 million.