ppt - WMO
Research and Development Project Improving the prediction of heavy precipitating systems over La Plata Basin LPB-ReD Presented by Alice M. Grimm Based on the draft of a preliminary RDP proposal written by Celeste Saulo - University of Buenos Aires, CIMA, Argentina Christopher Cunningham - CPTEC-INPE, Brazil Alice Grimm - Federal University of Paraná, Brazil La Plata Basin (Grimm, 2011) - 5th largest basin in the world, 2nd in SA - Area: 3.100.000 km2 - Principal sub-basins: Paraná, Paraguay and Uruguay Rivers. - Covers parts of 5 countries: Argentina, Bolivia, Brazil, Paraguay and Uruguay. - Population > 200 millions. - Produces most of the electricity, food and exports of those 5 countries (~70% GNP). High Impact weather in La Plata Basin Time scales from the very short range to the intraseasonal, and up to a season. Heavy and/or persistent rains (frequently leading to floods and landslides) SACZ (summer) – blocking events (winter) – MCS (spring and summer) – cyclogenesis (autumn and spring) Severe storms (tornado, wind gusts, hail, intense precipitation, lighting, etc) Late Frosts Warm/cold spells Droughts Synoptic and mesoscale features One of the main mechanisms supporting severe storms formation over LPB is the interaction between large-scale cold fronts and moisture advection from the Amazon. This interaction gives rise to explosive convective complexes that are amongst the largest and strongest in the world (Zipser et al., 2006). Heavy precipitating events over LPB, are mostly associated with cyglogenesis, cut off lows, mesoscale convective systems, blocking, stationary South Atlantic Convergence Zone (SACZ). There are major regional differences in the structure, intensity, and diurnal cycle of rainfall systems. The La Plata Basin MCSs (average area: ~5 105 km2; average lifetime: ~12 hours), are larger and more intense than the rainfall MCSs in the Amazon Basin (average area less than 1105 km2 and shorter lifetime: 3-6 hours). MCSs are influenced by mesoscale effects such as jets and other topographically forced circulation and surface atmosphere interactions. The SALLJ is the jet with most extensive influence. MCSs are modulated by the diurnal cycle. Equatorward incursions of frontal systems Vera et al. 2006 The day-to-day variability of rainfall over subtropical South America and western Amazon basin is largely explained by northward incursions of mid-latitude systems to the east of the Andes, even in summer. The deep northward intrusion of midlatitude systems is attributed to the dynamical effect of the Andes topography, which plays a significant role on the structure and evolution of the synoptic systems that cross South America. Equatorward incursions of frontal systems From Garreaud and Wallace 1998 Cold fronts tend to be directed northward to the east of the Andes, fostering the advance of cold air incursions into subtropical/tropical latitudes. In summer there is large impact on the precipitation, through the equatorward propagation (~10 ms-1) of a northwest-southeast oriented band of enhanced convection ahead of the leading edge of the cool air, which tends to be followed by an area of suppressed convection. This synoptic scale banded structure, which maintains its identity for about 5 days, is the dominant mode of the day-to-day variability of deep convection, contributing with ~25% of summer precipitation in the central Amazonia and ~50% over subtropical South America. These bands influence convection in the SACZ. Fontal systems influence on tropical convection Three types of influence: Type 1 - frequent in austral summer and spring, is characterized by the penetration of a cold front in subtropical South America that interacts with tropical convection and moves with it into lower tropical latitudes. Type 2 - also more frequent in austral summer, is characterized by Amazon convection and southward enhancement of convection in a quasistationary northwest–southeast oriented band extending from the Amazon basin to subtropical South America with the passage of a cold front in the subtropics. When this pattern remains longer than 4 days, it often characterizes the SACZ. Type 3 - more frequent in austral winter, is represented by a quasistationary cold front in subtropical South America and midlatitudes, without significant interaction with tropical convection. There are differences in MCSs for the 3 types of frontal system interaction with tropical convection. For instance, Type 2, often evolving into SACZ, has larger horizontal extent but less vertical development than Type 1. Siqueira and Machado 2004; Siqueira et al. 2005 Fontal systems influence on tropical convection Type 1 Three types of frontal system influence on tropical convection: Type 2 Type 3 Mean cold cloud top fractions Siqueira and Machado 2004; Siqueira et al. 2005 Mean 850 hPa wind anomalies The South American Low-Level Jet LLJ Composites NDJF, 00 UTC (Marengo et al. 2004) 06 UTC 12 UTC 18UTC SALLJ spatial structure depicted by NOAA/P-3 missions in SALLJEX SALLJ diurnal cycle at 700 hPa depicted by SALLJEX observations (Nicolini et al. 2004) From C. Vera Synoptic Variability X SALLJ The SALLJ events are conditioned by synoptic variability, and can be separated into two groups: (1) events in which the LLJ extends farther south, at least to 25S, (2) events in which the jet leading edge is north of this threshold. (1) Moisture flux (2) Moisture flux Precipitation Precipitation Nicolini et al 2002 Mesoscale Variability LPB exhibits the most extensive “hot spot” of the most intense thunderstorms on Earth, according to the TRMM data (1 January 1998 - 31 December 2004). Locations of intense convective events according to different proxies for convection intensity, using the color code matching their rarity. The parameter limits for each category are indicated above each color bar. For example, of the 12.8 million PFs, only about 0.001% (128) have more than 314.7 lightning flashes per minute. (Zipser et al., BAMS, 2006) Mesoscale Variability In La Plata Basin Mesoscale Convective Complexes (MCCs) occur frequently during October-April. Summer Autumn Winter Spring • Average area: ~5105 km2 • Average lifetime: ~12 hours • Cycle: preferentially initiate in late afternoon and mature during nighttime. • Develop east of the Andes, and move preferentially southeastward. • Intensification related with the subtropical jet and SALLJ. • More than 80% of MCCs occur during SALLJ events that penetrate farther south. Compilation of the MCCs location as given by several works (J. C. Conforte) Diurnal Cycle The peak is observed at afternoon/early evening in most of the monsoon region, consistent with the more suitable thermodynamic conditions during this part of the day. From R. H. Johnson Along the northeastern coast of South America there is frequently afternoon genesis and subsequent inland propagation of coastal squall lines forced by onshore low-level flow (e.g. Cohen et al. 1995). The coastal band of convective cloudiness increases to a maximum in the late afternoon and weakens during nighttime. After inland propagation, convection is reactivated in the afternoon of the next day (Garreaud and Wallace 1997). Nocturnal convection is prevalent over the subtropical plains (LPB), where it can be ascribed to the diurnal cycle of the SALLJ (Berbery and Collini 2000), to the nocturnal convergence into the valley of the Parana River basin, and to the decrease of the compensating subsidence associated with the SACZ (Silva Dias et al. 1987). Contribution of synoptic and intraseasonal timescales to total variance of summer rainfall Synoptic variability Ferraz and Grimm 2004 Intraseasonal variability Intraseasonal Variability in Summer (10-100 day bands) EOF1 – 16,3% (Ferraz, 2004) REOF1 – 10,0% REOF2 – 7,3% 10-20 day band REOF1 – 7,8 % 20-30 day band REOF1 – 11,5 % Separating into different frequencies: 30-70 day band REOF1 – 10,6% Intraseasonal Variability (30-70 day band) Composites of rainfall anomalies and vertically integrated moisture flux for wet and dry phases of the first and second rotated EOFs (“dipole” modes), in the 30/70 day band. “Westerly regime” Wet phase MJO - Phase 4 Precip. anomalies Grimm 2011 (in preparation) “Easterly regime” 1st rotated mode Ferraz and Grimm 2004 Dry phase 2nd rotated mode The Madden-Julian Oscillation allows some predictability, but there are other intraseasonal oscillations not well understood. Intraseasonal Variability Origin Not yet well understood. Spectral analyses of precipitation and OLR show distinct peaks in the 10-20 day band, 20-25 day band, and 30-70 day band bands. Remote influence? The first rotated mode seem to be related to the MJO via tropical teleconnection, while the second rotated mode seems to be related to wave-trains propagating southeastward from West or Central Pacific, rounding the southern tip of South America and turning toward the northeast, probably originated from MJO-related convection. Regional circulation, like the SALLJ, may be modulated by intraseasonal fluctuations of zonal flow above the Andes and consequent fluctuations of the orographically bound cyclone east of the Andes. Local forcing? Simulations with a regional climate model by Grimm et al. (2007) show that there are links between soil moisture, surface temperature and regional circulation that can modulate intraseasonal oscillation thus producing interannual variability with pattern similar to the first intraseasonal patterns. Moreover, the rough topography in Southeast Brazil is important in shaping the associated circulation and precipitation patterns. REGIONAL FORCING: The intraseasonal variability, might be favored or hampered, according to its phase, by local circulation anomaly set up by processes triggered by soil moisture conditions in spring, so that the first mode of interannual variability of summer precipitation is not just the rectification of intraseasonal variability or product of random sampling of intraseasonal events. November REOF1 r=-0.32 January REOF1 Grimm et al. 2007; Grimm and Zilli 2008 Mechanisms of the spring-summer relationship Grimm, Pal and Giorgi 2007 Soil moisture * 0.5 Soil moisture * 0.5 + SST Soil moisture * 0.5 - topo SALLJ Areas with significant variation of extreme events: Increase Decrease (Grimm and Tedeschi 2009) Apr (+) SALLJ SESA La Plata Basin floods in May 1998 (from C. Saulo) Apr (+) El Niño Jan (+) Nov (0) La Niña Modulation of extreme precipitation events by climate variability – ENSO SESA Central-East Brazil Central-East Brazil Nov (0) Jan (+) Modulation of extreme precipitation events by climate variability – ENSO REG. MONTH AVERAGE OF EXTREME EVENTS a (1956-2002) a NOV (0) EN (11) LN (9) NEUTRAL (27) 6.3 1.2 2.5 ENSO-related significant changes in the frequency of extreme rainfall events are much more extensive than changes in monthly rainfall, because ENSO influence is stronger on the categories of more intense daily precipitation. (Grimm and Tedeschi 2009) RDP in the La Plata Basin 1. Motivation Main motivation: lack of comprehensive understanding about the processes that determine severe weather events in the La Plata Basin, and our limitation in providing skillful forecasts that would contribute to minimize their impacts. The rainfall patterns in the La Plata Basin result from the interaction of various meteorological systems and scales. Therefore, the region is a suitable “test bed” for the assessment of model performance and the development of specific tools for high impact weather forecasting. Even if some of the systems that cause heavy precipitation can be reasonably well forecasted, predicted location and timing are frequently not accurate and the associated amount of precipitation is usually under predicted. Experiments using an enhanced network of observations obtained during SALLJEX (Vera et al., 2006) suggest that when low level moisture flux is better represented, precipitation forecasts over LPB can be substantially improved (Herdies et al., 2007; 2011). Besides improving the representation of the initial state in order to provide better forecasts over the region, we also need to improve our knowledge of the physics underlying heavy precipitating events . 2. Current state Each operational center uses its own models to forecast at diverse ranges, but there is a lack of documentation about their actual skill for predicting extreme events. Many heavy precipitating events are not captured by state-of the art models. Analyses of ensemble systems’ skills in the region, show that despite the combination of different forecasts, there is very low skill for predicting precipitation thresholds above 25 mm, even with only 1 day in advance. Although available, ensemble prediction systems (EPS) are not fully used by the operational community and even less by other potential users. Most of the products derived from these systems do not concentrate on intense precipitation. The advantage of ensemble forecasting applied to severe weather has to be demonstrated through the development of specific products. Besides dynamical models, there are diverse techniques suitable for heavy rainfall precipitation, mainly related with the use of remote sensing. A radar network exists over Brazil and Argentina, but does not share common algorithms/systems to exchange timely information that could help providing specific alerts. The coordination of the Brazilian and Argentinean network of radars is one of the tasks of the CHUVA project. Thus, involvement in nowcasting techniques could be desirable. 2. Current state Operational radar network over LPB. (Courtesy of Paola Salio) 2. Current state Weather prediction for the monsoon season Skill for predicting daily precipitation during the monsoon season in South America is quite variable depending on the precipitation threshold. There are high probabilities of detection for low values of daily precipitation but they are overestimated, as shown by the bias. For larger values there is low probability of detection and they are underestimated. (Period: monsoon season 2007-2008. Courtesy of Maria Assunção Silva Dias) 2. Current state Forecast of extreme event: South Brazil, 18-19 Aug 2011 Left: 24h precipitation accumulated in 19 Aug 2011. Right: satellite image valid for 201108190600. Precipitation forecast 48 h in advance from Ensemble Prediction System. This extreme event in western Parana State and other locations in the Southern Region affected more than 17,000 people. The amounts of rainfall exceeded 50 mm in 24 hours (1/2 to 1/3 of the mean monthly total). Those excessive rainfalls caused floods that affected more than 300 homes. Strong winds associated with the thunderstorms ruined seven towers used for energy transmission. (Cunningham and Escobar, 2011) 2. Current state Forecast of extreme event: South Brazil, 18-19Aug 2011 Left: surface analysis; Right: synoptic chart at 850 hPa. There was an intense cold front across South Brazil, especially over southern Paraná State. This frontal system was caused by a cyclogenesis process, with the extratropical cyclone located on the Atlantic Ocean, east of Buenos Aires. The northerly moisture transport was enhanced by the low level jet, from the Amazon towards South Brazil. 2. Current state Probability of detection of a precipitation event for several thresholds. All curves display 48h forecast. Each curve displays one model of the CPTEC suite. T126_L28 is the AGCM with 100km of horizontal resolution approximately and 28 layers in the vertical. T213_L42 is the AGCM with 63km of horizontal resolution approximately and 42 layers in the vertical. T299_L64 is the AGCM with 40km of horizontal resolution approximately and 64 layers in the vertical. Acoplado is the ocean-atmosphere coupled model; the atmosphere is integrated at a T126L28 resolution. RPSAS_40 is the limited area model integrated at 40km of horizontal resolution and initiated with CPTEC’s regional analysis. GPSAS213 is the CPTEC AGCM integrated at 40km of horizontal resolution and initiated with CPTEC’s global analysis. ENS_GLOB is the Ensemble Prediction System integrated at T126L28 resolution. 2. Current state Forecast of extreme event: North Argentina, Feb 2010 Accumulated precipitation forecasts obtained with different models compared against observations (first row). (a) 2/Feb/2010, (b) 5/Feb/2010 and (c) 19/Feb/2010 (Suaya et al. 2010). Different models were unable to predict the amount of precipitation associated with diverse events occurring during February 2010 (the event on February 5 reached 240 mm in one day). 3. Some scientific questions How early can we predict extreme precipitating events, in a probabilistic sense? Which factors determine such predictability? (some potential factors: time of the year, soil moisture, low frequency variability (such as ENOS), wave interactions; How well are the mechanisms leading to heavy precipitation understood / represented by current models? How much of the forecast skill can be attributed to the resolution and convective parameterizations used? How much can an enhanced network of observations improve the forecast of extreme events? What is the most convenient way to combine the information from different ensemble members, so that it becomes useful for predicting extreme precipitating events? What is the role of wetlands areas (e.g. Pantanal) in the occurrence/maintenance of extreme precipitation events and extensive flooding risk? What is the role of biomass burning and megacity emissions in the precipitation? What and how can we learn from model errors diagnostics? Diagnostic techniques help enhancing our understanding of the climate system, identifying problems and improving the models. 4. Main objectives of the proposed RDP To improve the assessment of state-of-art models ability to anticipate extreme weather conditions and to quantify their skill; To analyze the sources of model errors in order to feedback the development of models itself; To evaluate the impact of extra data in the quality of forecasts; To use TIGGE outputs and other available ensemble systems, to develop products/ tools adequate for extreme weather forecast in the region and evaluate them in the operational context; To foster the use of TIGGE products by the organization of training courses and workshops. Extended prediction? Influence of climate variability? 5. Societal impacts It is expected that an RDP like the one proposed here, will have immediate impacts through the implementation of state-of-the art prediction techniques at the operational services. Given that the focus will be on heavy precipitating events, and through the involvement of water resources/water administration/alert systems agencies, it is also expected that this project will help to articulate and facilitate the communication between weather forecasts community and the users of forecasts. A regional virtual center for monitoring and forecasting severe weather for Southeastern South America could be established and maintained in the future, so as to strengthen and optimize regional networks. 6. How the RDP will undertake the research and development? Before describing how the RDP will undertake the research and development (including methodology and time line), it is necessary to carry out a workshop for discussion of the research that should be undertaken in the RDP. Some issues: - use of ensemble forecast information for mitigation of severe weather impacts; - techniques of ensemble forecasting applied to severe weather; - experiments using an enhanced network of observations; - other research and methods to achieve the objectives. Before this workshop, a summer school for operational forecasters and end-users of the meteorological information is proposed in order to disseminate the basic concepts behind ensemble prediction and evaluation, in order to homogenize the level of knowledge of the users of information and hence motivate them. FORECAST DEMONSTRATION PROJECT A FDP should happen within the proposed RDP.