Enhancing objective short-range forecasts of the pre-convective environment using SEVIRI data

Ralph A. Petersen1, Robert Aune2 and Thomas Rink1 1 Cooperative Institute for Meteorological Satellite Studies (CIMSS), University of Wisconsin – Madison 2 NOAA/NESDIS/STAR, Advanced Satellite Products Team, Madison, Wisconsin

ABSTRACT:

This paper describes development efforts and evaluation results made with the CIMSS NearCasting system during the last year. Tests of the 1-9 hours forecasts of GOES products were made at US National Weather Servicer (NWS) Forecast Offices, CIMMS, and both the Storm Prediction Center (SPC) and the Aviation Weather Center (AWC) divisions of the National Centers for Environmental Prediction (NCEP). The tests at the two centers focused both on where/when severe will occur, as well as all forms of deep convection will and will not occur. The results showed that the prediction period could be successfully increased from 6 to 9 hours, that the analysis improved by increasing the number of observations projected forward from previous, that the NearCast fields can provide information about low-level triggering mechanisms and storm severity, that the NearCasts enhanced NWP guidance by isolating which forecasts areas were and were not likely to experience convection and when, and that successful use of the NearCasting tools requires increased forecaster training and education - both about the NearCast system itself and interpreting satellite observations and derived products.

INTRODUCTION:

This paper provides an update to Petersen et al., 2010. During the past year, primary emphasis of CIMSS NearCasting activities has focused on testing and evaluation of the short-range forecasting system internally at CIMSS, at local US National Weather Servicer (NWS) Forecast Offices, and both the Storm Prediction Center (SPC) and the Aviation Weather Center (AWC) divisions of the National Centers for Environmental Prediction (NCEP). The tests at the two centers had different objectives according to their missions: at SPC, they focused on identifying areas where severe convection would occur, while at AWC, they attempted to identify areas where all forms of deep convection would and would not occur. However, all groups wanted to: a) Increase lead-time, b) Reduce false alarms and increase probability of detection, c) Provide updates / detail to NWP guidance for next several hours, and d) Increase use of satellite products (they currently rely heavily on Numerical Weather Prediction (NWP) models and radar observations).

FINDINGS:

Summary of Forecaster Evaluations:

NearCast products were evaluated from periods ranging from several months to several weeks. Forecaster appraisals of the current and potential value of existing and future products included the following observations/recommendation:

1. Provide information about dynamic triggering, 2. Extend forecast length 3. NearCast fields (especially tendencies) were most useful when used to diagnose initial growth and coverage 4. limit the usefulness of product at times 5. Nearcasts were most valuable when used in conjunction with observations and other model data (both where convection will and will not occur) - Especially useful in updating / verifying mesoscale NWP and Ensemble guidance, especially during summer when only ~15% of NWP forecasts are correct 6. More experience was needed using the product to help with interpreting the observed fields and combined parameters.

The NearCast model developments and future actions that resulted from these recommendations are described in the following sections

Extending NearCast Length and Areal Coverage:

A number of research and applications activities have been undertaken to address the issues listed above. The first of these addressed both the issues that forecasters wanted the NearCast guidance extended beyond 6 hours into the future and wanted to have fewer areas blocked by cover. Before evaluation activities were initiated at SPC and AWC, the length of the NearCasts was successfully increased from 6 to 9 hours into the future. Since the NearCast products are generally available within less than one hour after the GOES observing sequence is begun, this effectively expands the maximum useful time of the NearCast products by 60% (from 5 to 8 hours).

Extending the length of the NearCast predictions also benefitted the data coverage in both the initial analyses and the NearCast predictions. The NearCasting system is unique in that it uses Lagrangian trajectory techniques to project geostationary satellite observations forward in time to future geographical locations. This is done once every hour as new observations become available, with the information about trajectory location and parameters stored every 15 to 30 minutes. Graphical NearCast products are then produced by ‘binning’ all of the trajectory data from all previous observations valid at the same time onto a regularly spaced grid. As such, data from up to 9 previous observation times can impact the graphical NearCast analyses and predictions.

Figure 1: Comparisons of NearCast low-level Equivalent Potential (θE) analyses for 24 May 2011showing increased analysis coverage when using one hour of on-time observations only (left), on-time observations plus forward projections of data from 4 previous hours (center) and on-time observations plus forward projections of data from 9 previous hours (right).

Figure 1 shows the impact of including projections of previous observations in improving the spatial coverage of the NearCast analyses. In this case, the left panel shows an analysis based only on date from 1300 UTC 24 May 2011. The center right panel shows the effect of including both observation at 1700 UTC and projections of data from the previous four observing times (1300, 1400, 1500 and 1600 UTC), effectively tripling the number of ‘data points’ available to the analysis. By including data from nine successive observing times at 2200 UTC, the ‘missing data’ area (black regions over the central and eastern US on the left panel) has been reduced by between 60-80% compared to the analysis made using only on-time data.

It should also be noted that forecasters at SPC were especially impressed by the ability of the GOES data and NearCast system to detect and monitor the progression of the ‘dry line’, an important factor in initiating severe convection, eastward from the Texas panhandle into western Oklahoma during the morning and early afternoon.

Providing information about dynamic triggering:

Although previous examples have shown the ability to identify areas where convective storms are likely to develop, many of the areas characterized as being convectively unstable never experience convection due to either the presence of a strong (as described in Petersen et al, 2010) or the lack of a sufficiently strong low-level lifting mechanism. In addition, both thermodynamic and momentum-based parameters are needed to differentiate the probable severity of the ensuing storms.

Investigation of the lower- and mid-level wind fields showed that the NearCast models had information and sufficient detail about the low-level convergence and atmospheric shear to isolate areas of substantial lifting and deep-layer shear. The tornado event from 2007 over Poland described in Petersen et al. (2010) was again used for testing. In this case, trajectory forecasts were made using initial and geopotential gradients from the NCEP Global Forecast System interpolated to SEVIRI observations locations at 840 and 480 hPa.

Figure 2: Example of evolution of NearCast low-level wind (arrows) and divergence/convergence (blue/red) preceding tornadic storm development over Poland on 20 July 2007. Note the formation of a southwesterly low-level jet upstream of the tornado location (white marker) in the 6 hours period to the event, along with the growth and progression of a band of convergence (and corresponding lifting) across the area of the storm immediately prior to the tornado sighting.

The results in Figure 2 show the development of a low-level jet during the 6-hour NearCast period with a region of strong convergence (divergence) couplet forming near (upwind) and slightly before the tornado formation. This information about dynamical lifting, combined with the information about convective instability and change in capping inversion presented in the previous report, help to identify where within the areas of convective instability storms are and are not likely to occur.

Figure 3: Example of evolution of NearCast low-level moisture flux (white arrows), divergence/convergence (blue/red) and 840- 480 hPa Vectors preceding tornadic severe storm development over Poland on 20 July 2007. Note the continuing low-level moisture supply associated with the south-westerly low-level jet provide energy for sustained convection toward the tornado location (white marker) in the 6 hours period to the event, along with the growth of wind shear immediately prior to the tornado sighting.

Figure 3 shows the development of a low-level moisture flux and wind shear during the 6-hour NearCast period before the tornado formation. The development of strong moisture transport and intensifying wind shear in the vicinity (southwest) of the tornadic convection points to the ability of the NearCast trajectories to provide dynamical information about both the thermodynamic and wind fields immediate prior to storm genesis, even when using wind data obtained from relatively low resolution global NWP models.

Providing help interpreting the NearCast fields and combined parameters:

SPC and AWC forecasts noted that the NearCast fields (especially lower-level transport and stability tendencies) were most useful when used to diagnose initial storm growth and coverage. The NearCasts also added significantly to the value of NWP guidance in helping determine where convection would and would not occur. They were especially useful in updating / verifying NWP/Ensemble guidance during summer, when only ~15% of short-range NWP precipitation forecasts are correct in the US. Forecasters also noted that much more experience was needed using the product to help them interpret the observed and NearCast GOES fields and combined parameters. Specifically, users need:

• Extended Training on use of NearCasts products, • Enhanced Education on both the Satellite Observations and Products, as well as the NearCasting Process, and • To ‘Break the Bond’ to NWP output • Especially the heavy reliance of forecasters on NWP precipitation guidance as a primary tool in predicting the location and timing of severe convection.

This can be accomplished through a combination of 1) improved image display capabilities in operations to help forecasters understand physical/dynamical processes occurring in the , and 2) new displays to improve forecaster understanding and reduce information overload.

Figure 4 illustrates a static view of one such display option. Although these display combination are best viewed in time loops, the single-time display of a 6-hour NearCast provides a good example for discussion. When loops of only the right panel were presented to forecasters, they had difficulty determining where to focus their attention. Although in retrospect forecasters noted that convection tended to occur in areas where the convective instability was increasing, their attention was initially drawn to the larger areas of convective instability, which often persisted in areas of strong capping inversions or minimal low-level lifting.

Figure 4: Example of multiple panel display of a 6 hour NearCast of derived Convective Instability (right panel) and component parameter fields (left panels) from 0000 UTC 25 May 2011. Reds and yellows in left panels indicate areas of lower θE (cooler, drier air) and greens and blues indicate higher θE (warmer, moister air). Scales change from lower to upper panels. Red areas in right panel are convectively stable (θE increases with height) and blue/purple/white regions indicate greatest convective instability (θE increases with height).

By including synchronized, but smaller-sized loops of the two component parameters predicted by the NearCasting system on the left of the same display, forecasters were much more able to visualize and understand 1) where the low-level thermal/moisture patterns were most favorable, 2) how the differential transport of drier/cooler air aloft could increase the Convective Instability, 3) where stability tendencies (both positive and negative) were largest (e.g., near Dallas, Texas, where sustained convection was observed at this time, over NE Colorado, where low-top supper cells formed as very cold/dry air moved over moderately warm/moist air at lower-levels, and over northern Pennsylvania, where the lack of warm/moist air at low-levels inhibited convection) and 4) how all of this Nearcasts related to the dryness patterns observed in GOES moisture imagery.

Another aspect of the displays that drew comment from forecasters was use of Convective Instability as a primary GOES NearCast guidance parameter. The NearCast products were developed to take advantage of the parameters that GOES observes best (e.g., 2-3 deep layers of moisture). Adding temperature data to produce θE improved further upon the moisture-only products and provided the basis for calculating Convective Instability as the difference in θE for a deep layer between the lower- and middle-troposphere. The choice of this parameter was consistent with many years of observations that convection tended to form along the trailing and especially leading edges of dry bands observed in water-vapor imagery. Questions arose as to how this deep-layer Convective Instability (CI) related to more common indices such as the (LI) or CAPE. In particular, many people thought of Convective Instability as an indicator of instability over a very thin layer. Figure 5 shows the results of tests comparing LI and deep-layer CI for a variety of temperature/moisture conditions at 850 and 500hPa typical of convective events. The results show a linear relationship, with the stable/unstable threshold point (where LI crosses 0o) shifted slightly in CI to about -3o. The result show that deep-layer CI can indeed be treated as a substitute for LI, but also retains important information about mid-level dryness that is observed well by GOES but this is not included in the LI.

Figure 5: Comparisons of LI and deep-layer Convective Stability for a variety of temperature and moisture conditions at 850 and 500hPa.

FINDINGS / RECOMMENDATIONS:

Results of development and evaluation efforts with the Geostationary satellite sounding based NearCasting system developed at CIMSS during the past year include the following findings and recommendations:

• Based on success of Proving Grounds experiences, move the NearCast products from the SPC Proving Grounds to SPC forecast desks for greater exposure and more systematic evaluation • Expand Proving Grounds evaluation beyond SPC and AWC to include NCEP’s Hydrometeorological Prediction Center (HPC) and Ocean Prediction Center (OPC) • Enhance Training for both the NearCasting systems and products and overall Geostationary satellite products and their uses • Test the inclusion of POES retrievals in the NearCasting system as a means of removing biases in GOES soundings • Convert NearCast model to Isentropic Coordinates to better follow moisture/dryness in clear air (and therefore adiabatic flow) conditions – Alternatively, test assigning parcels to level of maximum retrieval weighting function for each retrieval • Expand the use of objective validations of the pre-storm environment using a variety of asynoptic data sets – GPSMET, AMDAR, Wind Profilers, . . . • Exploring the possibility of working with EUMETSAT to provide real-time SEVIRI NearCasts for evaluation over Europe, east-center and southern Africa as a further expansion of the GOES-R Risk Reduction.

REFERENCES:

Pajek, M., R. Iwanski, M. Konig and P. Struzik, 2007: Extreme Convective Cases - The use of satellite products for storm nowcasting and monitoring, EUMETSAT Conference, Darmstadt, Germany

Petersen, R. A., R. Aune, T. Rink, 2010: Objective short-range forecasts of the pre-convective environment using SEVIRI data, EUMETSAT Conference, Cordoba, Spain