Real-Time Midterm - Elizabeth Dahl

Event Background

On August 14, 2020 at 23Z, northern Minnesota experienced an outbreak of severe thunderstorms, producing several tornadoes across the area and reports of wind damage. In total, over 20 tornadoes occurred in Minnesota, according to the NWS Twin Cities office. 

Figure 1: Surface analysis map from August 15, 2020 at 0Z. Form the NOAA Weather Prediction Center.

To get a better idea of the conditions during the event, we can look at the surface conditions. A sounding from the peak of the storms would be ideal, but data from the Minneapolis WFO is unavailable for that time. Figure 1 depicts surface conditions at the peak time of the storm on August 15th at 0Z. Minneapolis was experiencing warm temperatures, with a high of 79 degrees F and a dew point of about 72 degrees F, making for a moist, humid atmosphere. At this location, and near Duluth, where the severe storms took place, there is a meeting of warm and cold fronts with an associated low pressure system in northern Minnesota. Low pressure systems are typically associated with precipitation and severe weather, so the tornadoes and thunderstorms are likely due to that. 

Assessing Our Performance

In total, we issued 11 warnings throughout the duration of the event; five tornado warnings, four severe storm warnings, and two flash flood warnings. Out of these 11 warnings, seven were verified and four were not. The NWS Duluth office issued a total of 12 warnings, with us missing five of those. We had a mix of verified warnings that were earlier and later than the official warnings, and for the most part our verified polygons were in a similar area to the ones drawn by NWS.

Early Flash Flood Warning

At 23:42Z, we issued a flash flood warning for an area about 70 miles northwest of Duluth. The polygon for the warning area is is shown in Figure 2. At the time, there was a high reflectivity in the radar image, which indicated to us that there would be higher rainfall rates in the area. Additionally, spotter reports came up in the area that there was heavy rain around this time. When drawing the polygon, we focused on the areas that had the highest reflectivity value and also had a spotter report.

Figure 2: Our flash flood warning polygon, filled in blue, issued at 23:42Z on August 14th.

When the event actually happened, NWS Duluth issued a flash flood warning at 23:47Z for the dark green rectangular area, shown in Figure 3. Our polygon was in just about the same location as the NWS one, with a small portion to the southwest of the polygon not accounted for. We issued our warning about 5 minutes before NWS issued theirs, which would have allowed residents more time to take action. 

Figure 3: NWS-issued flash flood warning polygon, outlined in a bold green rectangle, from 23:47Z on August 14th.
From Iowa State Local Storm Report Application.

Missed Tornado Warning

One warning that we missed was a tornado near the Solana State Forest at 0:10Z, outlined in bold red in Figure 4. We were focusing more on spotter reports to identify tornadoes because some of the radial velocity imagery was unclear. There were not any storm spotter reports for this tornado, which is one of the reasons why we failed to warn for it. The radial velocity in this area at the time the tornado warning was issued is shown in Figure 5. To identify tornadoes from the radial velocity, we were looking for velocity couplets. At this time, the velocity couplet in the area was very unclear, if at all visible to us. There is an alternating pattern in toward-radar and away-radar winds in the area, but we could not make out a clear couplet, so we decided to not issue a warning. Throughout the lab, we would periodically check the correlation coefficient plots to look for a "blue dot" of tornado debris. We did not see that for this warning either, which is another reason why we failed to issue a warning for it.

Figure 4: NWS-issued tornado warning polygon, outlined in bold red over Solana state forest, from 0:10Z on August 15th.
From Iowa State Local Storm Report Application.

Figure 5: Radial velocity at 0:10Z on August 15th in the area of the tornado.

Narrow Severe Storm Warning

The last warning I am focusing on is our severe storm warning issued at 0:06Z. Our warning polygon is shown in Figure 6. We identified a severe storm occurring about 50 miles southwest of Duluth. We based this warning primarily off of storm spotter reports. Some of the warnings that came in for this storm were fallen trees, peeled off roofs, and uprooted docks. These spotter reports, along with radial velocity images showing high wind speeds, led us to issue a severe storm warning for this area.

NWS also issued a warning in nearby areas around the same time as ours. At 0:10Z, NWS issued a severe storm warning for the area in the bold yellow polygon, as shown in Figure 7. The NWS's polygon covers a much larger surface area than the one we issued. This would have caused issues for us, as not enough people would have been warned about high wind speeds in the area, possibly causing extra damage to property. 

In retrospect, I am glad that we issued some warning during this time for the area, but regret that we did not select a larger area. Something that prevented us from doing this was the radar picking up ground clutter near Duluth. Because of this clutter, we were unsure of whether or not that area was experiencing a severe storm. To better address this, we could have looked more into the radial velocity near Duluth to determine what wind speeds were at that time.

Figure 6: Our severe storm warning polygon, filled in blue, issued at 0:06Z on August 15th.

Figure 7: NWS-issued severe storm warning polygon, outlined in bold yellow, from 0:10Z on August 15th.
From Iowa State Local Storm Report Application.

Forecast Difficulties

Even though we had a mix of successful and failed warnings, the warning process overall went relatively smoothly. Some issues we ran into with the warning process included missing new storm reports and being unsure when to end the warning. With the layout of the website, it was difficult for us to easily identify which storm reports were new and which were old, so sometimes we missed a new report pop up on the map. We could have fallen back on radar for those, but we were relying primarily on storm reports, as mentioned previously. We also had trouble knowing exactly when to end the warning. We knew the storm system was moving quickly, but not sure the "normal" times to issue warnings for, or which warnings required a longer issuance time than others. Checking the radial velocity more often could have helped us with that and allowed us to make more accurate reports.

Real-world storm warnings also face challenges. The NWS has an extensive network of radar data available for the United States, but some areas still are not covered by radar. This can present issues for forecasters trying to issue warnings in those areas, since they would have to rely primarily on storm reports. This could mean forecasters' issuances are delayed, causing possible extra damage to life and property. Another big challenge is deciding exactly when to issue a warning. Some members of the public have a distrust in the NWS because of issues like false warnings. If an office issues too many warnings, people may begin to not pay attention to warnings as much since they have panned out to be false. On the other hand, if an office issues too few warnings, issues arise in protecting life and the property of those who did end up being affected by a storm.

Technical issues also arise in the real-world that prevent forecasters from properly issuing warnings, such as broken radars and internet outages. At one point during the forecasting exercise, my partner lost their internet, so I needed to work on issuing the warnings alone. This was difficult because I wanted someone to bounce ideas off of and help me check the radar and storm reports. During that time, there was an increase in severe weather so I did end up needing to issue a warning without consulting her, which I thought was challenging.