comparison of six weather based risk models for...
TRANSCRIPT
COMPARISON OF SIX WEATHER BASED RISK MODELS FOR THE DEVELOPMENT OF
POTATO LATE BLIGHT
Anne-Grete Roer Hjelkrem, Mladen Cucak & Jens Grønbech Hansen
EuroBlight 13.05.2019
IPMBLIGHT2.0
14.05.2019 2
WP 1 – sampling and genotyping P. infestans populations
WP 2 – phenotyping P. infestans
WP 3 – developing improved LB DSS – IPM 2.0
WP
0 –
coo
rdin
atin
gan
d d
isse
min
atin
gIP
MB
ligh
t2
.0
Isolate sampling & Reference isolate collections
Aggregated data on individualisolatesField monitoring of
virulence
Data management and analytic tools (e.g. poppr)
New DSS simulation models/ modules
Papers and conferencepresentations
Website
Potato LateBlightToolbox
Genetic (SSR) fingerprints
Ad
viso
ryB
oar
d
Existing DSS improved and tested
Fungiciode resistancetesting
Stakeholderinteraction
EUROPEAN DECISION SUPPORT SYSTEMS
14.05.2019 3Anne-Grete Roer Hjelkrem
GOAL
– Compare European weather based sub-models for estimation of risk of potato late blight
– Four research questions:
1. Is there a seasonal effect of Blight risk?
2. Is the Blight risk different between regions?
3. Is the Blight risk different between years?
4. Is the estimation of Blight risk different between models?
14.05.2019 4User manual – NIBIOs power point template
Foto: Vinh Hong Le
OVERVIEW
14.05.2019 5Anne-Grete Roer Hjelkrem
Quality control
Blight Management
Nærstad Model
Irish Rules
WUR Blight
Hutton Critera
The reference model
Input data Model implementation
QUALITY CONTROL OF WEATHER DATA
14.05.2019 6Anne-Grete Roer Hjelkrem
Quality control
Input data
Apelsvoll
Ås
Dronninglund
Ikast
Dundee
Cambridge
Oak ParkMoore Park
EldenArcen
Viller Saint Christophe
Boigneville
QUALITY CONTROL OF WEATHER DATA
14.05.2019 7Anne-Grete Roer Hjelkrem
Missing values
Min and Max
Mean RH during hours with
1. precipitation > 0,5 mm
2. precipitation > 0,5 mm,
Solar radiation < 0,5 MJ/m²
and Wind speed < 5 m/s
Median RH among the 200
highest RH values each month,
sorted from highest to lowest
Ma
x [%
] to
p 2
00
Rh
by m
on
th
82
84
86
88
90
92
94
96
98
100
Jun Jul Aug May Jun Jul Aug May Jun Jul Aug
201720162015
May
Max
ho
url
yR
H [
%]
OVERVIEW
14.05.2019 8Anne-Grete Roer Hjelkrem
Quality control
Blight Management
Nærstad Model
Irish Rules
WUR Blight
Hutton Critera
The reference model
Input data Output dataModel implementation
OUTPUT DATA
– All models were ran simultaneously with the quality controlled weather data
14.05.2019 9Anne-Grete Roer Hjelkrem
Output data
BLIGHT RISK ESTIMATION
10
Dronninglund 2015
BLIGHT RISK ESTIMATION
11
Dronninglund 2015
BLIGHT RISK ESTIMATION
Blight Risk Estimations were summed by month for each region, year and model, and normalized.
ANOVA and post hoc Tukey’s HSD test
14.05.2019 12Anne-Grete Roer Hjelkrem
𝑌𝑖𝑗𝑘𝑙 = 𝜇 +𝑚𝑜𝑛𝑡ℎ𝑖 + 𝑟𝑒𝑔𝑖𝑜𝑛𝑗 + 𝑦𝑒𝑎𝑟𝑘 +𝑚𝑜𝑑𝑒𝑙𝑙 + 𝜀𝑖𝑗𝑘𝑙
5 months × 12 region × 4 year × 6 models
Foto: Vinh Hong Le
𝑌𝑖𝑗𝑘 = 𝜇 +𝑚𝑜𝑛𝑡ℎ𝑖 + 𝑟𝑒𝑔𝑖𝑜𝑛𝑗 + 𝑦𝑒𝑎𝑟𝑘 + 𝜀𝑖𝑗𝑘
SEASONAL EFFECT OF BLIGHT RISK
14.05.2019 13Anne-Grete Roer Hjelkrem
A significant positive effect of month number
Increased estimated Blight Risk through the season
AA
SEASONAL EFFECT OF BLIGHT RISK
14.05.2019 14Anne-Grete Roer Hjelkrem
AA
YEARLY BLIGHT RISK DIFFERENCES
14.05.2019 15Anne-Grete Roer Hjelkrem
YEARLY BLIGHT RISK DIFFERENCES
14.05.2019 16Anne-Grete Roer Hjelkrem
0
5
10
15
20
May June July August September
Mean Air Temperature (°C)
2014 2015 2016 2017
0
200
400
600
800
1000
1200
1400
May June July August September
Sum Precipitation (mm)
2014 2015 2016 2017
65
70
75
80
85
90
May June July August September
Mean Relative Humidity (%)
2014 2015 2016 2017
YEARLY BLIGHT RISK DIFFERENCES
14.05.2019 17Anne-Grete Roer Hjelkrem
YEARLY BLIGHT RISK DIFFERENCES
14.05.2019 18User manual – NIBIOs power point template
Date
01-04 01-05 01-06 01-07 01-08 01-09
Norway
Russia
Scotland
Estonia
Lithuania
Finland
Ireland
Sweden
Poland
Serbia
England&Wales
Denmark
Belgium
Netherlands
Switzerland
Germany
France
= 2016, Attack in > 5 conventional fields = 2015, Attack in > 5 conventional fields
REGIONAL BLIGHT RISK DIFFERENCES
14.05.2019 19Anne-Grete Roer Hjelkrem
Vill
er S
ain
t C
hri
sto
ph
e
Bo
ign
evill
e
Ikas
t
Dro
nn
ingl
un
d
Oak
Par
l
Mo
ore
Par
k
Ap
elsv
oll Ås
Cam
bri
dge
Du
nd
ee
Eld
en
Arc
en
Esti
mat
edN
orm
aliz
edB
ligh
tR
isk
REGIONAL DIFFERENCES IN BLIGHT RISK
14.05.2019 20Anne-Grete Roer Hjelkrem
Low
Estimated risk levels, averaged over the models:
Medium
High
June 2015
REGIONAL/YEARLY DIFFERENCES IN BLIGHT RISK
14.05.2019 21Anne-Grete Roer Hjelkrem
June 2016June 2015
REGIONAL/SEASONAL DIFFERENCES IN BLIGHT RISK
14.05.2019 22Anne-Grete Roer Hjelkrem
August 2015June 2015
CONCLUSIONS
1. Is there a seasonal effect of Blight risk?
2. Is the Blight risk different between regions?
3. Is the Blight risk different between years?
4. Is the estimation of Blight risk different between models?
14.05.2019 23User manual – NIBIOs power point template
Foto: Vinh Hong Le
Increased estimated Blight Risk through the season
Generally lower and later estimated Blight Risk north
Estimated Blight Risk differ between years
Estimated Blight Risk differ between models, but the models generally agrees in estimated high and low risk periods
CONCLUSIONS
– Blight risk differ between years and regions, and decision support systems provides valuable information about the risk levels through the season
– A more robust estimate of Blight Risk can be achieved by using several models simultaneously
– Model validation including biological observations must be performed in order to compare the accuracy of the different models
14.05.2019 24User manual – NIBIOs power point template
Foto: Vinh Hong Le
THANK YOU FOR THE ATTENTION!