(CRA-W) Meteorological data and lateblight modeling China-Europe Workshop of Potato Sustainable and Efficient Productions Harbin, 24th of July 2015 Damien Rosillon Version 25/04/2014
CRA-W : A public agricultural research centre based in Belgium Multidisciplinary research institution CRA-W it is 500 people 300 ha of facilities 3 sites 16 research units 4 research departments
CRA-W : Breeding and biodiversity Unit Farming systems, Territory and Information Technology Unit
Architecture of a warning system
CRA-W s Automatic weather stations network 38 automatic weather stations on an area of 16.844 km² 1 weather station for 440 km² Average distance between stations : 30 km
Automatic weather stations network
A weather station Pyranometer Wind vane Raingauge Air temperature and relative humidity sensor Anemometer Ground surface temperature probe Soil temperature probe (- 20 cm) 7
Automatic weather stations network «An automatic weather station is automatic. After its installation nothing more to do.» Wrong! Clogged raingauge
Automatic weather stations network «An automatic weather station is automatic. After its installation nothing more to do.» Wrong! Blokked raingauge
Automatic weather stations network «An automatic weather station is automatic. After its installation nothing more to do.» Wrong! Dust on relative humidity sensor
Automatic weather stations network «An automatic weather station is automatic. After its installation nothing more to do.» Wrong! «Hostile» environment
Late blight modeling Developpement of a model based on Guntz and Divoux principles Based on past weather data Helps the expert to interpret meteorological data. Are the meteorological conditions favourable to late blight development? It helps to decide if spraying fungicide is useful or not Outputs : a potential infection or fructification periods Detection of periods favourable to the infection Estimate of the infection gravity Measurment of the length of the incubation phase Detection of the day where the fructification phase can happen
Late blight modeling RH curve T curve Incubation curve Infection and level of gravity Cumulated daily rainfall rainfall
Impact of meteorological data on model outputs Impact of a slight malfunction of a relative humidity sensor on the amount of simulated infections Scenario 1 : regular situation
Impact of meteorological data on model outputs Underestimation of 5% of RH measurements
Impact of meteorological data on model outputs Overestimation of 5% on RH measurements
Impact of meteorological data on model outputs Model outputs is higlhy dependant on meteorological data Wrong meteorological data can lead to wrong decision Reliable meteorological data are crucial for a reliable late blight warning system maintenance of weather stations network is crucial Maintenance of the site Frequent recalibration of the sensors meteorological data must be validated
Meteorological data management 40 stations 6 sensors per station 1 record / hour 5760 data a day 2.102.400 data a year
Validation tools : step 1 data check Step 1 : Data check : identification of missing or wrong data Automatic procedures : Check if data every hour Check by thresholds Impossible to entirely automate validation. Automatic procedures are completed with «human» procedures «Human» procedures : Visual analysis Monthly indicators
Validation tools : step 1 data check Human procedures daily maps
Validation tools : step 1 data check Human procedures graphic analysis 40 35 30 25 20 15 10 5 0 1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191 201 211 221 231 241
Validation tools : step 1 data check Human procedures monthly indicators Indicateurs region station indic tsa indic_hra indic_ens_inc_plu indic_missing haute_belgique Libramont -0,2 0% 30% 2% 0% haute_belgique Michamps -0,1-3% 40% -22% 0% haute_belgique Amberloup 0 0% -99% 1% 14% haute_belgique Bergeval 0,7 0% 19% 9% 0% haute_belgique Willerzie -0,2 3% 10% 9% 0% haute_belgique MOYENNE 0 0% 0% 0% 3% gaume Chassepierre 0,1 2% -5% 22% 0% gaume Schockville -0,4 0% 6% -16% 0% gaume Ruette 0,4-2% 0% -6% 0% gaume MOYENNE 0 0% 0% 0% 0% intermediaire Chimay 0 0% 2% 1% 0% intermediaire Ferrieres -0,5-4% -14% -4% 0% intermediaire Haut-le-Wasti 0 3% 16% -7% 0% intermediaire Jemelle -0,1 2% -1% 2% 0% intermediaire Seny 1-5% -14% -19% 0% intermediaire Jevoumont -0,4 4% 12% 26% 0% intermediaire MOYENNE 0 0% 0% 0% 0% moyenne_belgiqulouvain-la-neu 0,3-4% 13% -34% 0% moyenne_belgiqufloriffoux -0,7 2% 1% -16% 0% moyenne_belgiquleuze -0,1-5% -2% -18% 0%
Validation tools : step 1 data check Human procedures monthly indicators June 2013
Validation tools : step 1 data check Human procedures monthly indicators August 2013
Validation tools : step 2 data correction Step 2 : Correction Integrated tools to correct the data Linear interpolation Spatial interpolation Data translation Validated meteorological data are sent to the model
Next steps «Field scale» decision support system Virtual weather stations network Use of grid data GIS technics Source : Racca Paolo et al., «Decision Support Systems in Agriculture : Administration of Meteorological Data, Use of GIS and Validation methods in crop protection warning service». ZEPP Germany Weather forecast integration
Conclusions What ever the source of the data : Observation from weather station network Grid data Weather forecast Quality weather data are crucial for a quality decision support system
Contact Websites : www.pameseb.be (weather station network website) Damien Rosillon : d.rosillon@cra.wallonie.be +32 61231010
Thank you for your attention