Presentation is loading. Please wait.

Presentation is loading. Please wait.

Object-based precipitation analysis: application to tropical cyclones and the Slovenian radar data Mini Workshop on NWP Modelling Research in Slovenia.

Similar presentations


Presentation on theme: "Object-based precipitation analysis: application to tropical cyclones and the Slovenian radar data Mini Workshop on NWP Modelling Research in Slovenia."— Presentation transcript:

1 Object-based precipitation analysis: application to tropical cyclones and the Slovenian radar data Mini Workshop on NWP Modelling Research in Slovenia 15.December 2011 Gregor Skok Julio Bacmeister, Joe Tribbia, Benedikt Strajnar, Jože Rakovec, Anton Zgonc, Mark Žagar

2 Overview Object based analysis Analysis of tropical cyclone precipitation using satellite data Hail-area tracking algorithm using radar data

3 Object based analysis Doctoral thesis: “Object-Based Analysis And Verification Of Precipitation Over Low- And Mid-Latitudes”

4 TRMM 3B42 3 hourly precipitation accumulations, 0.25 deg Motivation

5 Object identification method Based on MODE - Method for Object-based Diagnostic Evaluation developed by Devis et al (2006a,b) Part of Model Evaluation Tools (MET) verification package developed by NCAR 3 steps: – Smoothing – Thresholding – Identification of self-enclosed areas as objects The method tries to simulate what a human forecaster or analyst might infer by a more subjective visual evaluation of a field => (Objective) simulation of a subjective evaluation

6 Original MODE method Thresholding only Smoothing Thresholding after smoothing

7 Methodology - Extended method Time evolution of objects “tri-dimensional” objects Enables study of properties: location, size, shape movement, lifespan, total object precipiation, ….

8 Doctoral thesis: Pacific, 6-years of 3-hourly TRMM 3B42 precipitation data

9 Trajectories for 2001 BLUE – short lifespan, RED – long lifespan highest density of objects with a longer lifespan (red) is in the ICTZ and in the low-latitudes in the west eastern tip of the ITCZ – mainly objects with short lifespan Central America – mostly shortlived objects …….

10 Trajectories for 2001 ORANGE – eastward, GREEN – westward Movement in the northern and southern parts of domain is predominantly eastward In the ITCZ region, movement in both directions is present although westward movement (green) is more frequent In the eastern and western part of the ITCZ the westward movement is clearly dominant.

11 Number of objects vs. lifespan Straight in a Log-Log graph = Power law

12 Analysis of tropical cyclone precipitation using satellite data Gregor Skok, Julio Bacmeister, Joe Tribbia TRMM 3B42 precipitation data The IBTrACS tropical cyclones track database 11 years

13 FiT - “Forward in Time” object identification

14

15

16 -Merging! -Forced to check into the past and also perform merger there

17 FiT - “Forward in Time” object identification

18

19 -Don’t allow merging! -Larger “wins” -No need to check into the past -> only forward in time -Side benefit: faster and less memory consuming

20 FiT - “Forward in Time” object identification

21 The problem of “missed” precipitation Inside objects (threshold 7mm/3h) there is only 50 % of all precipitation. The other 50 % is located in a dislocated self- enclosed areas of low-intensity precipitation or just outside the borders of objects. We want to include nearby low-intensity precipitation for TC analysis

22 Estimation of object precipitation by “grown” objects Precipitation threshold

23 Secondary thresholdPrecipitation threshold Estimation of object precipitation by “grown” objects

24 Sequentially grow objects: 1 iteration

25 Estimation of object precipitation by “grown” objects Sequentially grow objects: 4 iterations

26 Estimation of object precipitation by “grown” objects Sequentially grow objects: 9 iterations -> end

27 Estimation of object precipitation by “grown” objects Unattributed precipitation Might be more unattributed low intensity precipitation below secondary threshold In GROWN objects (to 1 mm/3h) now 75 % of all precipitation

28 IBTrACS database

29 Identification of TC objects Object IBTrACS TC center MATCH? YES Distance smaller than 2.5 deg MATCH? NO Distance larger than 2.5 deg

30 Animation

31 TC precipitation [mm/day]

32 Contribution of TC precipitation to all precipitation [%]

33 Zonal means of TC precipitation GLOBAL SEA LAND

34 Regions

35 TCs contribute about 4 % (on average 40 km 3 /day) This percentage is on average higher for oceans than for land (4.8 % vs. 1.4 %). NH the TCs contribute around 5.1 % and in SH about 2.8 % precipitation Compared to the oceans, the land sub-regions have much smaller TC precipitation volumes. some land regions get over 3 %: Australia, Maritime continent with E Asian islands and E Asia some seasons TCs contribute more precipitation; i.e. N America (6 % in SON), Australia (4 and 5.5 % in DJF and MAM), Maritime continent with E Asian islands (5,5 % in JJA and SON), E Asia (3 and 6 % in JJA and SON) and S Asia (4 % in SON)

36 Yearly global TC precipitation

37 Hail-area tracking algorithm using radar data Gregor Skok, Benedikt Strajnar, Jože Rakovec, Anton Zgonc, Mark Žagar Using volumetric radar data from Lisca – 8 years Areas with hail precipitation identified using a combination of two methods: Waldwogel et al. (1979) and Gmoser et al.(2006). This produces a 2D binary field – hail yes/no. Radar scan is performed every 10 minutes. A sequence of 2D binary “hail” fields is fed into the object identification algorithm The movement of objects represent the movement of areas with hail precipitation

38 Animation

39

40 Hail area tracking No smoothing/thresholding possible since the field is binary The hail areas are relatively small and move fast – they often do not overlap in 10 minute intervals To overcome this problem the objects are artificially grown in all directions This improves the overlap but can merge nearby objects The value of parameter describing the “extent” of growth has to be selected carefully – sensitivity analysis

41 not grown

42 grown by 1 km

43 grown by 2 km

44 grown by 3 km

45 Number of objects Not an exponential distribution Hail events not a random shortlived process

46 Results Trajectories longer than 150 min. Red = eastward, blue = westward Direction of movement by azimuth (regardless of lifespan)

47 Thank you

48 ORIGINAL

49 SMOOTHING

50 OBJECTS AFTER THRESHOLDING

51 Success of TC matching Total IBTrACS TCs = 1144 Total IBTrACS TC trajectory points = Matched TCs = 1141 Matched TC trajectory points = Matching success of trajectory points = 81.5 % Non-matching happens when precipitation amounts in the TC are very low (usually below threshold so that no object is identified) and therefore a smaller amount of global TC precipitation is “missed” because of this reason.


Download ppt "Object-based precipitation analysis: application to tropical cyclones and the Slovenian radar data Mini Workshop on NWP Modelling Research in Slovenia."

Similar presentations


Ads by Google