Ideas for improving the CTC / OT algorithm for aviation forecasters

During the later half of this week in the aviation group at HWT, a couple of ideas were discussed for increasing the applicability of the OT, and maybe even more so CTC algorithms in short-term and nowcast aviation forecasting. I relayed the initial idea of incorporating a color scale and approximate value for cloud top height to the OT algorithm, which Steve Silberberg from AWC was very interested in, to Kris Bedka. Kris briefed me on similar products that exist or are in development that might be able to be implemented in AWIPS for use at the AWC. This would be used jointly with the UWCI algorithm to help identify "threat" convection.

Perhaps an even more applicable tool would be to include an approximate cloud top height or change in cloud top height between satellite scans with the CTC algorithm to quantify the actual vertical growth of individual convective updrafts. This could be extremely valuable to aviation forecasters in a short term / nowcast timescale for echo top forecast updates. Seeing a specific value for height change might be an even quicker assessment of cloud growth when combined with the CTC color scheme. Also, en route air traffic would be able to use such a tool to avoid rapidly growing updrafts prior to tops reaching 34000+ ft. This preventative measure would increase air safety by decreasing the probability of moderate to severe turbulence encountered by aircraft flying over cells with tops just below 340 due to the little knowledge pilots currently have of cumulus development outside of VFR. Both aviation forecasters this week expressed great enthusiasm for the advancements that such a product would bring to the aviation forecasting community including Brad Sherman from the FAA. To help familiarize the group with FAA operations, Brad explained how commercial aircraft usually fly over tops under 340 and "wing-wag" around those with 340+ tops. Learning about how flight delays and cancellations during the warm season rapidly stem from not only ongoing and developing convection, but also "snapshot" 21 and 23z 250+ convective top coverage forecasts made 4 hours in advance, was engaging to say the least.

-Dan Hartung

Unique role of simulated satellite imagery in severe weather forecasting

This week has been very challenging for all forms of 6-12 hour probabilistic forecasting, primarily for QPF and convective initiation. The general mesoscale setup at 12z each day has been at least one (and in a couple instances multiple) remnant MCS complexes over the CONUS domain from the prior evening, each of which had its own unique outflow boundary that was not captured in both determinate and multi-member ensemble runs initialized at 00z. Since the CIMSS simulated ABI imagery from the NSSL-WRF is initialized at 00z, it has predominately missed the development of these MCS systems and attendant boundaries as well. That being said, the simulated imagery has not been found useful when trying to determine where and when convection will initiate during the subsequent forecast period primarily due to a lack of skill in predicting the existence and evolution of remnant mesoscale boundaries.

On a more positive note, Chris and I convinced the aviation and severe teams to use the simulated low to mid-level water vapor imagery to identify the structure and location of the remnant surface meso-vortex in TX shortly after 13z 02 June. A closer look at bands 8 and 9 also nicely showed the evolution of the upper level trough / ridge pattern and wind fields that were not immediately identifiable on the standard 250 hPa chart at 12z 03 June. Probably the most important positive outcome of the week thus far with the simulated satellite imagery is that forecasters have found the simulated WV imagery (B10) to be a helpful tool in narrowing down the noisy 500 hPa vorticity field to identify important vorticity signatures and regions of wind shear that could be sufficient to dynamicaly force / support convective initiation during the present forecast period.

They also said it would be helpful if the CIMSS simulated imagery were available from 12z onward instead of the current 17z time period to eliminate having to toggle back and forth with the CIRA simulated imagery currently being used to verify any morning boundaries that the NSSL-WRF 00z run captures.

-Dan Hartung

CI applications in model initialization

Discussions ongoing in the aviation forecast group are focusing on the utility of the cloud-top cooling rate and CI products in helping to initiate rapid-refresh models, such as the RUC of HRRR. Current efforts being tested this year include pushing radar reflectivity information into the models to help initialize the location of storms. A couple limitations of doing this are that the current radar reflectivities do not provide good information about updraft strength or growth in the future. The aviation forecast group is discussing the possibilities of using the CI products as a proxy for vertical development to help the models identify areas of new growth, and thus improve their short term forecasts. We all know that the current models have issues with their short term forecasts as they generally take some time to "spin up". This should be something we should look into in the future for collaborations within the model community.

Convective initiation over central TX on June 2

Figure 1 - 24-hour SPC severe weather reports from 2 June 2010
During this morning's EFP briefing, Steve Weiss asked me to show how the convective initiation (CI) products performed during the previous day over central TX associated with some sporadic severe weather after 2000 UTC (Fig. 1). An MCV existed over central TX during the morning on the 2 June 2010 following an MCS from the previous day. Based on the 00Z NSSL-WRF simulated satellite imagery guidance this MCV was expected to initiate widespread convection around 1900 UTC (Fig. 2). I examined the output from SATCAST as well as UWCI and its associated cloud-top cooling (CTC) rate product during this time period to determine the utility of CI nowcasts in this case.


Figure 2 - UW-CIMSS NSSL-WRF simulated GOES-R band 9 (6.9 micron) water vapor imagery for 1800 (left) and 1900 UTC (right)
Figure 3 - Radar base reflectivity for 1725 UTC on 2 June 2010
An isolated cell of convection developed on the eastern edge of the MCV at about 1725 UTC (Fig. 3). None of the GOES-R Proving Ground CI products detected this convection because cirrus was contaminating the scene at the time. The NSSL-WRF simulated satellite imagery hinted this during the 1800 UTC forecast with an isolated cell forming on WV imagery in the same area (Fig. 2). There was no additional CI until after 1900 UTC, as the simulated satellite imagery expected. At 1832 UTC, SATCAST nowcasted CI over the southern edge of the MCV (Fig. 4). At this time, no additional convection had developed over the MCV based on radar base reflectivity data (Fig. 4), but at 1909 UTC this area eventually showed reflectivity exceeding 35 dBZ (Fig. 7), suggesting a lead-time of just over 30 minutes of the product in predicting CI.


Figure 4 - Base radar reflectivity at 1831 UTC (left) and SATCAST CI nowcast at 1832 UTC (right) on 2 June 2010
At 1845 UTC, SATCAST continued to nowcast CI further west, still with no development shown on radar (Fig. 5). This area eventually initiated at 1914 UTC, as shown by radar base reflectivities exceeding 35 dBZ (Fig. 8)... maintaining the CI lead-time of around 30 minutes seen previously. At 1902 UTC the first CTC signals were seen with no associated UWCI nowcasts made during that time (Fig. 6). This is likely due to the cooling rates not equaling the minimum threshold to flag CI within UWCI. We display CTC rates starting at -2 K / 15 min, whereas UWCI requires CTC rates exceeding -4 K / 15 min to make a nowcast. Also at 1902 UTC SATCAST showed some CI nowcasts near the areas highlighted by the CTC product, but close examination of the areas flagged for CI appeared to be thin cirrus overrunning small Cu, and thus they were determined to be false CI nowcasts (Fig. 6).


Figure 5 - Base radar reflectivity at 1846 UTC (left) and SATCAST CI nowcast at 1845 UTC (right) on 2 June 2010


Figure 6 - Base radar reflectivity at 1900 UTC (top left), CTC at 1902 UTC (top right) and SATCAST CI nowcast at 1902 UTC (bottom) on 2 June 2010
At 1915 UTC UWCI showed CI nowcasts and strong CTC rates over developing storms on the southeastern edge of the MCV (Fig. 8). However, examining the visible imagery underneath suggests that these storms are already mature with expanding anvil clouds beginning to form. The 1914 UTC radar base reflectivity also shows these areas in excess of 50 dBZ (Fig. 8). SATCAST also nowcasts CI near the area, but these mature Cu are excluded (Fig. 8). This is most likely due to the differences in the cloud typing algorithms used by the two products and would require deeper investigation to draw any detailed conclusions. Radar base reflectivity at 1933 UTC shows that these areas flagged by SATCAST eventually did initiate (Fig. 9), but with much shorter lead-times (~10-15 minutes) than previous seen by nowcasts made 45 minutes to a half hour earlier.

Figure 7 - Base radar reflectivity at 1909 UTC on 2 June 2010


Figure 8 - Base radar reflectivity at 1914 UTC (top left), UWCI nowcast at 1915 UTC (top right), CTC rate at 1915 UTC (bottom left) and SATCAST CI nowcast at 1915 UTC (bottom right) on 2 June 2010
Figure 9 - Base radar reflectivity at 1933 UTC on 2 June 2010

Using simulated ABI satellite imagery and UWCI nowcast product to produce 21z and 23z >40 dbz echo probability snapshot forecasts

Simulated band 13 forecasted ABI satellite imagery for 21z (bottom) and 23z (top) 02 June 2010.(Apologies for the backwards times)

This morning in the aviation focus group at the HWT, I encouraged the use of simulated ABI satellite data to guage an idea of the location of > 40 dbz echos initiated after 18z by the NSSL-WRF. Comparison of 14z B13 simulated ABI data to realtime GOES infrared imagery showed a fairly accurate representation of both intensity and location of the remnants of last night's MCS that moved through IA/northern IL, at the time extending from central MI through northern OH. However, second MCS remnants over northern MO were absent in the simulated data (See WV imagery in previous post). Therefore the outflow boundary and cold pool from the second MCS was not accurately captured by the NSSL-WRF and resulted in the suppression of convection along the weak cold frontal boundary that extended from east-central IL southwestward through MO / KS / western OK until 23z. One large positive result was that the NSSL-WRF did suggest initiation of convection along the outflow boundary from the then MI MCS along the KY/IN border by 21z.

The large stable cold pool that was left over central MO led us to forecast a moderate probability (50-74%) of thunderstorm initiation along the outflow boundary from the morning MO and MI MCSs along the MO/AR and MO/IL/IN/KY borders in agreement with the simulated ABI satellite data from the NSSL-WRF. Even though the simulated imagery wasn't particularly accurate in its evolution of the MO MCS event, the simulated mid-level water vapor channel (B10) was useful in identifying the location of mid-level vorticies over the southern plains and gulf coast during the forecast period. It was one of many tools that was taken into consideration when issuing a slight and moderate probability of convective initiation (> 40 dbz tops) over west-central TX and also over the LA/AL gulf coast for both the 21 and 23z 02 June snapshot forecast periods. As I'll make mention of below, convection in TX was initiated by 20z in the simulated satellite imagery which was at approximately the same time as realtime initiation. The complex appears to be evolving into an MCS as the evening progresses.

During the afternoon updates, the aviation team relied fairly heavily on developing trends in ongoing initiation over western TX and the southeast US as a whole for forecast modifications. We looked extensively at the performance trends of the UWCI nowcast between 18 and 20z. The algorithm performed very well and captured explosive convection over western TX with ~15 minute lead time. The convection was associated with a low-level meso-vortex dynamically supported by a sharp shortwave ejected from the AZ/NM border out over the southern plains shortly after 18z. Real-time water vapor and visible satellite imagery was critical in locating and tracking this feature from 13z onward.

A final thought, Aviation Weather Center (AWC) forecasters Steve Silberberg and Bruce Entwistle discussed with me their strong interest in installing the UWCI, CTC/OT, and simulated ABI imagery products in NAWIPS at their facility and I said I would relay their request to Wayne, Jordan and others.

-Dan Hartung

Using simulated satellite to evaluate model performance

Observed (left) and simulated (right) WV imagery at 1300 UTC on 2 June 2010
During today's morning forecast discussion in the EFP the participants talked about how the 00Z deterministic model runs were of little help to them in making their early morning forecasts. One way that they discussed how this was discovered was by using the simulated satellite imagery provided by the 00Z NSSL-WRF and comparing the simulated imagery to that of reality. They noticed that the model imagery did not accurately describe the atmosphere at the current time see images above) and thus they then had to rely heavily on observed data of ongoing storms and extrapolate what could happen in the future to make their forecasts. This brings up a good point that should be remembered. When the models fail... where do you go?

EFP week 3 begins...

Today begins a new week of Spring Experiment activities within the HWT. There were no operations on Monday, hence the lack of posting. Also, the EWP is not operating due to the short week, so there will be a heavier interaction within the EFP this week. This week's visitors include Bill McCaul from the University of Alabama in Huntsville, and Dan Hartung and Chris Rozoff from UW-CIMSS. We will be focusing more on the simulated lightning threat and satellite imagery from the NSSL-WRF this week as those products are more relevant to EFP operations. Given the opportunity, we will also be evaluating the convective initiation and overshooting top / thermal couplet detection products during the afternoon time frames.

Today's focus for the severe and QPF groups in a mesoscale area centered over Omaha, NE. Bill McCaul and Chris Rozoff will be participating in the aviation group today who are focused on an area covering much of the eastern half of the US. Dan Hartung will be working with the QPF group today. The groups will rotate throughout the week. We are already utilizing the CTC product in the aviation group to highlight areas on satellite imagery of future convective development. We have talked about how the current satellite provides some limitations in detection due to the coarse resolution of the IR channels. Steve Silberberg from the Aviation Weather Center (AWC) discussed with Chris Rozoff and I about the possibilities of providing these products at the AWC and I told him we would be happy to help them getting the data into their NAWIPS systems since we have much experience doing this at the SPC.

The simulated satellite imagery was examined by the QPF group, comparing it to actual satellite imagery from the same time. We were able to determine that the NSSL-WRF was showing a good ability at capturing larger scale features such as short-waves and jet streaks. The smaller, more convective scale features were slightly misplaced, especially over the KS/NE area. We discussed the future of our efforts with the simulated model imagery, including providing more model runs and producing GOES-R-like products from the simulated imagery, such as band differences.