An operational phenological model for numerical pollen model for - - PowerPoint PPT Presentation

▶
an operational phenological model for numerical pollen
SMART_READER_LITE
LIVE PREVIEW

An operational phenological model for numerical pollen model for - - PowerPoint PPT Presentation

Dec 29, 2022 39 likes •770 views

An operational phenological model for numerical pollen model for numerical pollen prediction Helfried

slide-1
SLIDE 1
  • An operational phenological

model for numerical pollen model for numerical pollen prediction

Helfried Scheifinger and Elisabeth Koch ZAMG Siegfried Jäger and Uwe Berger SciCon Science Pharma Consulting GmbH Robert Neumcke mindshape freiburg

slide-2
SLIDE 2

Zentralanstalt für Meteorologie und Geodynamik

slide-3
SLIDE 3

24.06.2010 Scheifinger

Overview

  • 1. Motivation
  • 2. Selection of 14 phases
  • 3. TSM problems
  • 4. Probabilistic representation

Zentralanstalt für Meteorologie und Geodynamik

  • 4. Probabilistic representation
  • 5. Spatial resolution
  • 6. Representation on the Web
slide-4
SLIDE 4

24.06.2010 Scheifinger

Overview

  • 1. Motivation
  • 2. Selection of 14 phases
  • 3. TSM problems
  • 4. Probabilistic representation

Zentralanstalt für Meteorologie und Geodynamik

  • 4. Probabilistic representation
  • 5. Spatial resolution
  • 6. Representation on the Web
slide-5
SLIDE 5

COST ES0602, COST ES0603 and WMO Joint Workshop Chemical and Biological Weather Forecasting: State of the art and future perspectives – Conclusions –

Natural allergenic species, in particular, pollen, are an

Zentralanstalt für Meteorologie und Geodynamik

Natural allergenic species, in particular, pollen, are an air quality issue of major concern. Allergies are increasing globally and the projected health impacts are alarming. Still, harmful allergenic species are currently not controlled by regulatory policies and measures.

slide-6
SLIDE 6

24.06.2010 Scheifinger Umweltabteilu ng

COST ES0602, COST ES0603 and WMO Joint Workshop Chemical and Biological Weather Forecasting: State of the art and future perspectives – Conclusions –

The adverse health effects could also be reduced by

Zentralanstalt für Meteorologie und Geodynamik

The adverse health effects could also be reduced by implementing and using forecasting and information systems for harmful allergenic species, e.g., By issuing pre-warnings for susceptible population

  • subgroups. Such adaptation measures showed a high

positive impact in several countries but their implementation requires broad-scale operational arrangements and a legislative basis.

slide-7
SLIDE 7

Wichtigkeit und Motivation

24.06.2010 Scheifinger Umweltabteilu ng

State of the art solution

Numerical pollen forecast via numerical weather forecast +

Zentralanstalt für Meteorologie und Geodynamik

atmospheric transport (chemistry) model

slide-8
SLIDE 8

Wichtigkeit und Motivation

24.06.2010 Scheifinger Umweltabteilu ng

Main problem

Example chemical weather forecast. Input: Emission inventory Output from NWF models … and it works

Zentralanstalt für Meteorologie und Geodynamik

… and it works Numerical pollen forecast. Input: No emission inventory for pollen

slide-9
SLIDE 9

Plan

24.06.2010 Scheifinger Umweltabteilu ng

Main problem

No emission inventory for pollen Solution: model pollen emission

Input:

Zentralanstalt für Meteorologie und Geodynamik

Input:

  • Plan density distribution
  • Phenology
  • Atmospheric pollen concentration

Output:

  • Real time pollen emission in space
slide-10
SLIDE 10

24.06.2010 Scheifinger

Phenological model

Assumption: phenological inception of flowering = begin of potential pollen emission into the atmosphere

Zentralanstalt für Meteorologie und Geodynamik

Model inception of flowering of pollen emitting

  • r anemophilous species
slide-11
SLIDE 11

24.06.2010 Scheifinger

Phenological model

First step towards NPFC As stand alone real time phenological maps support qualitative pollen forecast

Zentralanstalt für Meteorologie und Geodynamik

support qualitative pollen forecast

slide-12
SLIDE 12

24.06.2010 Scheifinger

Overview

  • 1. Motivation
  • 2. Selection of 14 phases
  • 3. TSM problems
  • 4. Probabilistic representation

Zentralanstalt für Meteorologie und Geodynamik

  • 4. Probabilistic representation
  • 5. Spatial resolution
  • 6. Representation on the Web
slide-13
SLIDE 13

24.06.2010 Scheifinger Umweltabteilu ng

14 COST 725 Phases

Zentralanstalt für Meteorologie und Geodynamik

slide-14
SLIDE 14

:

24.06.2010 Scheifinger Umweltabteilu ng

List of COST725 phases

Scientific name Common name

Phase

Acer platanoides Norway mapple

First flowers open

Aesculus hippocastanum Horse chestnut

First flowers open

Alnus glutinosa Black alder

Beginning of flowering

Alopecurus pratensis Foxtail grass

Beginning of flowering

Artemisia vulgaris Mugwort

Beginning of flowering

Betula pendula Silver Birch

Beginning of flowering

Corylus avellana Common hazel

Beginning of flowering

Zentralanstalt für Meteorologie und Geodynamik

Corylus avellana Common hazel

Beginning of flowering

Forsythia suspensa Forsythia

First flowers open

Fraxinus excelsior Common ash

First flowers open

Salix caprea Goat willow

First flowers open

Sambucus nigra Elder

First flowers open

Syringa vulgaris Lilac

First flowers open

Tilia cordata Small leaved lime

First flowers open

Secale cereale (winter) Rye

First flowers open

slide-15
SLIDE 15

24.06.2010 Scheifinger

Overview

  • 1. Motivation
  • 2. Selection of 14 phases
  • 3. TSM problems
  • 4. Probabilistic representation

Zentralanstalt für Meteorologie und Geodynamik

  • 4. Probabilistic representation
  • 5. Spatial resolution
  • 6. Representation on the Web
slide-16
SLIDE 16

Select a phenological model

24.06.2010 Scheifinger Umweltabteilung

TSM

) (

2

T R F

f t

∑

=

{

b f

T T if T T T T if R ≥ − < =

Zentralanstalt für Meteorologie und Geodynamik

) (

1

T R F

f t

∑

=

{

b b f

T T if T T R ≥ − =

slide-17
SLIDE 17

Select a phenological model

24.06.2010 Scheifinger Umweltabteilung

  • Phenological model, TSM, 2003, Vienna
  • Horse chestnut beginning of flowering

100 200 300 ure (1/10°C) 6000 8000 10000

um (1/10° C days)

Zentralanstalt für Meteorologie und Geodynamik

1 31 61 91 121

Yearday

  • 100

100 Temperatu 2000 4000

Temperature su entry date = function(mean daily temperature, begin date of summation, temperature sum at entry date, temperature threshold)

slide-18
SLIDE 18

Select a phenological model

24.06.2010 Scheifinger Umweltabteilung

Task: model fitting

  • Given for model fitting: T (time series of

mean daily temperatures) and t2 (entry date)

  • Searched for: the 3 model parameters

Zentralanstalt für Meteorologie und Geodynamik

  • Searched for: the 3 model parameters

values (t1, F, Tb) have to be chosen such that the TSM gives lowest RMSE

  • Solution: Model fitting via inverse techniques
slide-19
SLIDE 19

Select a phenological model

24.06.2010 Scheifinger Umweltabteilung

Model fitting via inverse techniques

  • Metropolis algorithm (Press et al., 1992)
  • + fast
  • - not absolutely reliable
  • - only one solution, unstable

Zentralanstalt für Meteorologie und Geodynamik

  • - only one solution, unstable
  • Look Up Table (LUT)
  • - slow, much computing power
  • + never fails
  • + all relevant solutions, stable
slide-20
SLIDE 20

Select a phenological model

24.06.2010 Scheifinger Umweltabteilung

Model fitting via inverse techniques

  • Preferred LUT
  • visualise cost function in parameter

space

Zentralanstalt für Meteorologie und Geodynamik

slide-21
SLIDE 21

Select a phenological model

24.06.2010 Scheifinger Umweltabteilu ng

Modelling of TSM parameters in space =

Zentralanstalt für Meteorologie und Geodynamik

= painful experience

slide-22
SLIDE 22

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-23
SLIDE 23

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-24
SLIDE 24

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-25
SLIDE 25

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-26
SLIDE 26

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-27
SLIDE 27

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-28
SLIDE 28

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Subjectively selected „optimum“ temp sum commencement date (yd 61) LT mean entry date: 142 – 132 No unique degday/tempthr pair!

Zentralanstalt für Meteorologie und Geodynamik

slide-29
SLIDE 29

Methods

24.06.2010 Scheifinger Umweltabteilung

  • Example of a LUT, 2D cut
  • Zentralanstalt für Meteorologie und Geodynamik
slide-30
SLIDE 30

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-31
SLIDE 31

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-32
SLIDE 32

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-33
SLIDE 33

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-34
SLIDE 34

24.06.2010 Scheifinger

Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

slide-35
SLIDE 35

24.06.2010 Scheifinger

Lilac bf, Rijeka (Croatia), 45° N Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

RMSE distribution in Birzai very different from that in Rijeka: lowest RMSE at higher temp/lower tsums in Birzai and lower temp/higher tsums in Rijeka yd 41 best temp sum commencement date for Rijeka, 20 days before Birzai

slide-36
SLIDE 36

24.06.2010 Scheifinger

Lilac bf, Rijeka (Croatia), 45° N Lilac bf, Birzai (Lith.), 56° N

Zentralanstalt für Meteorologie und Geodynamik

As first approximation a linear relationship, temperature sum as function of temperature threshold. Restricted to those elements of the LUT, which are members of the lowest 5% percentile.

slide-37
SLIDE 37

24.06.2010 Scheifinger

Relationship between temperature threshold and temperature sum at a station very strict (Syringa vulgaris bf):

Zentralanstalt für Meteorologie und Geodynamik

slide-38
SLIDE 38

24.06.2010 Scheifinger

Relationship between regression parameters and station coordinates noisy and weak: Shift = f(phi)

South

Zentralanstalt für Meteorologie und Geodynamik

South

540 dd

North

430 dd

slide-39
SLIDE 39

24.06.2010 Scheifinger

Relationship between regression parameters and station coordinates noisy and weak: Slope = f(lamda)

Zentralanstalt für Meteorologie und Geodynamik

East

47 dd/° C

West

57 dd/° C

slide-40
SLIDE 40

24.06.2010 Scheifinger

Conclusions

Modelling of TSM parameters in space is not trivial There exist weak relationships between TSM

Zentralanstalt für Meteorologie und Geodynamik

There exist weak relationships between TSM parameters and space, which can be used for modelling of TSM parameters in space I am still searching for a suitable mathematical procedure to model TSM parametes in space satisfactorily

slide-41
SLIDE 41

24.06.2010 Scheifinger

Conclusions

Three common assumptions concerning TSMs shattered:

  • 1. There exists a unique set of optimum TSM

Zentralanstalt für Meteorologie und Geodynamik

  • 1. There exists a unique set of optimum TSM
  • parameters. This is definitely not the case.
  • 2. Numerical search algorithm for TSM

parameters finds the best solution.

  • Because of point 1 this is impossible.
  • Nevertheless numerical solutions do

work

slide-42
SLIDE 42

24.06.2010 Scheifinger

Conclusions

Three common assumptions concerning TSMs shattered:

  • 3. One can work with one set of TSM

Zentralanstalt für Meteorologie und Geodynamik

  • 3. One can work with one set of TSM

parameters anywhere in the world.

  • Still to be show conclusively to be wrong
slide-43
SLIDE 43

24.06.2010 Scheifinger

Tentative procedure

The TSM parameters calculated at the climate stations are being interpolated to the ECMWF model grid via HRIDW

Zentralanstalt für Meteorologie und Geodynamik

slide-44
SLIDE 44

24.06.2010 Scheifinger

Validation

Spatial cross validation TSM run with interpolated TSM parameters, TSM run with TSM parameters deduced with

Zentralanstalt für Meteorologie und Geodynamik

TSM run with TSM parameters deduced with

  • bserved data

results (RMSE) compared with observed entry dates at ECSN stations

slide-45
SLIDE 45

24.06.2010 Scheifinger

Validation

Zentralanstalt für Meteorologie und Geodynamik

slide-46
SLIDE 46

24.06.2010 Scheifinger

Validation

Zentralanstalt für Meteorologie und Geodynamik

slide-47
SLIDE 47

24.06.2010 Scheifinger Umweltabteilu ng

Validation of the operation model on the grid

Subjectively compared model output in Vienna with current development of vegetation, where I live and work

Zentralanstalt für Meteorologie und Geodynamik

I live and work Model 1 to 2 weeks behind Checked ECMWF 2 m temperature data and found negative bias of 1 – 2° C !!!!

slide-48
SLIDE 48

24.06.2010 Scheifinger Umweltabteilu ng

Validation of the operation model on the grid

Zentralanstalt für Meteorologie und Geodynamik

slide-49
SLIDE 49

24.06.2010 Scheifinger Umweltabteilu ng

Validation of the operation model on the grid

Need to implement MOS correction

Zentralanstalt für Meteorologie und Geodynamik

slide-50
SLIDE 50

24.06.2010 Scheifinger

Overview

  • 1. Motivation
  • 2. Selection of 14 phases
  • 3. TSM problems
  • 4. Probabilistic representation

Zentralanstalt für Meteorologie und Geodynamik

  • 4. Probabilistic representation
  • 5. Spatial resolution
  • 6. Representation on the Web
slide-51
SLIDE 51

24.06.2010 Scheifinger

Probalistic representation

  • Hight reduced entry dates of neighbouring stations

show a large scatter (Siljamo et al., 2008)

  • Causes manyfold: genetic differences, topography,
  • bserver error, microclimate, …
  • TSM produces 2 states: phase entry has not yet

Zentralanstalt für Meteorologie und Geodynamik

  • TSM produces 2 states: phase entry has not yet
  • ccurred or has occurred => sharp line divides

space => not real

  • Alternative: represent the scatter as standard

deviation from historical phenological observations

slide-52
SLIDE 52

24.06.2010 Scheifinger Umweltabteilu ng

Probabilistic representation

Two sources of scatter:

  • 1. s1 : Standard deviation of the height reduced

entry dates = genetic differences, topography,

  • bserver error, microclimate, … )

Zentralanstalt für Meteorologie und Geodynamik

  • bserver error, microclimate, … )
slide-53
SLIDE 53

24.06.2010 Scheifinger Umweltabteilu ng

Lilac

Standard deviation on 0.25°grid, height reduced (mindist 20 km,

maxdist 100 km, minstat 10 Stationen)

Mean of 5 seasons 2000 – 2004 In case of no data total mean value = 5.16 days

Zentralanstalt für Meteorologie und Geodynamik

slide-54
SLIDE 54

24.06.2010 Scheifinger Umweltabteilu ng

Birch

Standard deviation on 0.25°grid, height reduced (mindist 20 km,

maxdist 100 km, minstat 10 Stationen)

Mean of 5 seasons 2000 – 2004 In case of no data total mean value = 7.75 days

Zentralanstalt für Meteorologie und Geodynamik

slide-55
SLIDE 55

24.06.2010 Scheifinger Umweltabteilu ng

Hazel

Standard deviation on 0.25°grid, height reduced (mindist 20 km,

maxdist 100 km, minstat 10 Stationen)

Mean of 5 seasons 2000 – 2004 In case of no data total mean value = 9.89 days

Zentralanstalt für Meteorologie und Geodynamik

slide-56
SLIDE 56

24.06.2010 Scheifinger Umweltabteilu ng

Probabilistic representation

Two sources of scatter:

  • 2. Standard deviation of the model grid
  • topography. The smaller (larger) the model grid

area, the smaller (greater) this standard

Zentralanstalt für Meteorologie und Geodynamik

area, the smaller (greater) this standard deviation: s2 = st * k

s2 : Standard deviation through topography (days) st : Standard deviation of topography (m) k: Slope of entry dates (days/m)

slide-57
SLIDE 57

24.06.2010 Scheifinger Umweltabteilu ng

Standard deviation of topography (ECMWF) ECMWF grid 0.25°

Zentralanstalt für Meteorologie und Geodynamik

slide-58
SLIDE 58

24.06.2010 Scheifinger Umweltabteilu ng

Standard deviation of topography * slope (0.03 days/m e.g.) ECMWF grid 0.25°

Zentralanstalt für Meteorologie und Geodynamik

slide-59
SLIDE 59

24.06.2010 Scheifinger Umweltabteilu ng

Probabilistic representation

Suggestion: max(height reduced s, s from topography)

Zentralanstalt für Meteorologie und Geodynamik

slide-60
SLIDE 60

24.06.2010 Scheifinger Umweltabteilu ng

Birch

max(height reduced s, s caused by top) ECMWF grid 0.25°

Zentralanstalt für Meteorologie und Geodynamik

slide-61
SLIDE 61

24.06.2010 Scheifinger

Overview

  • 1. Motivation
  • 2. Selection of 14 phases
  • 3. TSM problems
  • 4. Probabilistic representation

Zentralanstalt für Meteorologie und Geodynamik

  • 4. Probabilistic representation
  • 5. Spatial resolution
  • 6. Representation on the Web
slide-62
SLIDE 62

24.06.2010 Scheifinger

Spatial resolution

0.25°ECMWF grid too coarse. Wrong elevations in and around the Alps Interpolation to GTOPO30 1 km grid

Zentralanstalt für Meteorologie und Geodynamik

Interpolation to GTOPO30 1 km grid 25 106 grid points, special procedure, takes a few minutes to calculate

slide-63
SLIDE 63

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-64
SLIDE 64

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-65
SLIDE 65

24.06.2010 Scheifinger

Overview

  • 1. Motivation
  • 2. Selection of 14 phases
  • 3. TSM problems
  • 4. Probabilistic representation

Zentralanstalt für Meteorologie und Geodynamik

  • 4. Probabilistic representation
  • 5. Spatial resolution
  • 6. Representation on the Web
slide-66
SLIDE 66

Results

24.06.2010 Scheifinger KS Ost Klima

Zentralanstalt für Meteorologie und Geodynamik

slide-67
SLIDE 67

Results

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-68
SLIDE 68

Results

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-69
SLIDE 69

Results

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-70
SLIDE 70

Results

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-71
SLIDE 71

Results

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-72
SLIDE 72

Results

24.06.2010 Scheifinger Umweltabteilu ng

Zentralanstalt für Meteorologie und Geodynamik

slide-73
SLIDE 73

Wichtigkeit und Motivation

24.06.2010 Scheifinger Umweltabteilung

Zentralanstalt für Meteorologie und Geodynamik