Atmospheric teleconnections bird migration and implications for probabilistic

Atmospheric teleconnections, bird migration, and implications for probabilistic forecasts of bird strikes Steven B. Feldstein Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania, U. S. A. Presented at Tel Aviv University, Department of Zoology, Tel Aviv, Israel on May 12, 2010

PREDICTABILITY l Three time scales associated with atmospheric predictability l Deterministic Predictability (Weather) (Useful 3 -5 day numerical model forecasts) Numerous studies show a linkage between weather, i. e. , storms, precipitation, fronts, etc. , and bird migration. Van Belle et al. (2007) 3 -day forecast of bird migration intensity. l Extended-Range Predictability (1 week to 1 month timescale) Predictability mostly poor (because of chaos), except perhaps when teleconnction patterns are excited. Linkage to bird migration associated with storms that accompany teleconnection patterns (Elkins (2008). l Monthly and Seasonal Predictability (Climate, > 1 month, average of weather) (Closely linked to seasonal cycle and boundary forcing) (Useful ensemble forecasts. ) Linkage between El Nino/Southern Oscillation and bird migration (Hameed et al. 2009).

El Nino/Southern Oscillation and Bird Migration at Attu, Alaska l From 1980 -2000, each year, about 100 birders visited Attu, Island to add Asian bird species to the North American bird lists Island, Alaska to add Asian birds to their North American bird l Common Sandpiper Long-toed Stint Rustic Bunting l AIPD = Asian Individuals Per Day (North American species excluded from total) (Hameed et al. 2009)

El Nino/Southern Oscillation

ENSO, storms, and bird migration

ENSO affects bird migration through its influence on the latitude of the jet and the storms that follow the jet, i. e. ENSO alters the environment through which storms propagate.

The dominant Northern Hemisphere teleconnection patterns North Atlantic Oscillation Pacific/North American pattern Teleconnections evolve on a 7 -10 day time scale (longer than weather time scale). They alter storm path and storm intensity. Climate Prediction Center

Siberian Vagrants and the PNA Teleconnection Pattern l Records of all observations (locations and dates in California) of Dusky Warbler, White Wagtail, Red-throated Pipit and other Siberian vagrants l Composite PNA index Dusky Warbler (PNA=- 0. 42, n=15, p<0. 1) White Wagtail (PNA=-0. 26, n=22, p<0. 02) Red-throated Pipit (PNA=-0. 38, n=79, p<0. 01) (Steven Feldstein, Peter Pyle, Steve Mlodinow, Richard Erickson, Jim Tietz)

Anomalous wind vectors associated with Dusky Warbler records in California Lag -3 Days Lag -1 Days Lag -2 Days Lag 0 Days

Nearctic Vagrants and the North Atlantic Oscillation (Elkins 2008, British Birds)

Circumglobal Teleconnection Pattern EOF 1 Wave packets associated with SL precip wet dry Time-averaged V 300 over persistent event (lag -6 to lag +9 days) Correlation with EOF 1 =0. 83 Correlation with EOF 1 =-0. 72

300 -h. Pa geopotential height evolution - Middle Eastern precipitation -6 days -4 days -2 days Composite analysis Evolution of 300 -h. Pa height field determines the wind, T, P, P, and rainfall, variables which influence bird migration. 0 days +5 days +2 days +7 days Wave packet first seen +4 days over NE Pacific. Wave field persists for 2 weeks. This slow evolution +9 days may allow for a 7 -day weather forecast for Israel? Feldstein and Dayan (2008)

Israeli Migrants and the Circumglobal Teleconnection Pattern (CTP)? QUESTIONS: Does the CTP influence bird migration over Israel? l Can the CTP be used to forecast the bird migration intensity or bird strike frequency with a 17 day lead time (fall season)? l Beyond 3 -4 days, is a forecast of bird migration intensity with a probabilistic model based upon the CTP better than that based upon a deterministic linear regression model (e. g. , Van Belle et al. 2007)? l

16 North Pacific sea level pressure cluster patterns Example of cluster analysis

Tropical Convection Associated with the Madden-Julian Oscillation (MJO) l Phase 1 Dominant intraseaonal oscillation in the tropics l MJO cycle: 30 -60 days l Shading OLR l Phase 2 Time between phases ~ 6 days Phase 3 Phase 4 Phase 5 Phase 6 Time between Phases ~ 6 days Phase 7 From Wheeler and Hendon (2004) 20°E Phase 8 180° From Wheeler and Hendon (2004) 60°W

Frequency of occurrence for each cluster pattern and MJO phase

1 -7 day Forecast of Anomalous Bird Migration intensity in Israel Phase Number = location in Israel Lag = 1 to 7 days (Feldstein and Dayan 2008) Pattern Number = cluster pattern Color denotes anomalous bird migration intensity determined from composites of the daily bird migration intensity for each pattern number

1 -7 Day Probabilistic Bird Migration Intensity Forecast l Cluster analysis with 300 -h. Pa meridional wind (1 -7 day variability (CTP) dominated by the upper tropospheric flow and it also determines the lower tropospheric flow where birds are observed) (Cluster patterns represent slowly-evolving component of flow, small number of patterns with large spatial scales) Analysis performed at separate locations (airports, radar stations, etc. ) in Israel l Analysis can be performed separately for soaring (raptors) and powered flight (shorebirds) migrants l Conditional probabilities based on the accumulation of migrants during previous days (bad weather) l

Seasonal Forecast of Anomalous Bird Migration Intensity in Israel mean is the anomalous seasonal meridional wind is cluster pattern c is the frequency of cluster pattern c Obtain forecast of seasonal mean meridional wind Determine the cluster pattern which has the largest projection onto Estimate seasonal mean bird migration intensity (above average, below average) in terms of the most frequently occurring cluster pattern.

Combined Probabilistic/Deterministic Bird Migration Intensity Forecast F F = (w 1 F 1 + w 2 F 2)/(w 1 + w 2) F 1 = probabilistic forecast based on cluster patterns F 2 = linear regression forecast based on deterministic weather forecast model (e. g. , Van Belle et al. 2007). W = P(B|A) A= forecasted bird migration intensity B= observed bird migration intensity Presumably F 1 (F 2 ) forecast is better for longer (shorter) lead times. Presumably Bayesian forecast also possible using multimodel ensemble approach

Conclusions Bird migration related to (a) weather, (b) teleconnections (PNA, NAO), and (c) climate (ENSO) Forecast of bird migration intensity based upon cluster analysis of the 300 -h. Pa meridional wind Using weights, can combine probabilistic forecast with deterministic linear regression forecast The technique can be extended to seasonal bird migration intensity forecasts.