Experiences from Data retrieval to crop disease modelling 31.10.2017 Experiences from Data retrieval to crop disease modelling Hanna Huitu MMEA result seminar 26.9.2012 31.10.2017
Model development and data input Biology and epidemiology check-up. Model compatibility with different input weather data sources. Framework for agro models – model was implemented in CropInfra Framework, containing interfaces for meteorological data retrieval (MMEA), field crop information, and in future model result output (MMEA) Data 1: Interface provided by FMI (MMEA 25.5.result seminar), LAPS (Limited Area Prediction System) data Data 2: SoilWeather – a-Lab in-situ measurements (XML data; MMEA platform) Experiences so far © Maa- ja elintarviketalouden tutkimuskeskus 31.10.201731.10.2017
Implementation of a disease forecast model to the 31.10.2017 Implementation of a disease forecast model to the pilot environment BASE RISK Weather input data (MMEA) Cultivation history + DAILY RISK Cultivation Sporulation Infection THRESHOLD VALUES FOR WARNING + Itiönmuodostus; ilmankosteuden gradientti Variety
Feeding the model with epidemiological data 31.10.2017 Feeding the model with epidemiological data INFECTION SPEED AT DIFFERENT TEMPERATURES NET BLOTCH TAN SPOT Infection development - spot size mm in 7 days Weather data used in model development vs weather data available in operational use (instrumented fields, or random farm in location X)
31.10.2017 CropInfra Platform Internet-based networked production infrastructure, ”Sandbox” Service Oriented Architecture Project co-operation at national and EU levels: design and carrying out case studies (AgroXML, AgriXchange, Smart Agri-Food...) Co-operation with agricultural machinery manufacturers; studying present and emerging ecosystems Assisted decision making in agriculture: Science -> knowledge models -> practise Implemented onto an operational plant production farm (150 ha) with instrumentation to monitor both environment and machinery Research themes Internet of things & Future internet Esim. Infektion leviämisnopeustutkimus tulokset käytäntöön © Maa- ja elintarviketalouden tutkimuskeskus 31.10.201731.10.2017
System of services framework DPS Service 31.10.2017 System of services framework DPS Service SOA Interfaces (ESB) Optional SOA Interfaces Independent Services FMIS Service Framework (Cloud) Discovery Services (USDL based) Embedded DPS Trusted Log In Services Trusted Profile Services Data Storage Services AgroXML Weather Services PMML Model Services FMIS Service Framework (Cloud Proxy aka Local) Disease Pressure Service Framework Embedded DPS Limited Functionality Communication Enabler User Interface for stand-alone usage SOA Interfaces for service integration Säätietopalvelu – tähän pakataan MMEA tietolähteet Model services- tilaa mallikirjastolle Service Registration Enabler Framework Specific Services – User Specific Services Framework Specific Application Logic Enabler Framework Specific Data Model Enabler Service Registry Enabler Public – Private - System
31.10.2017 FMI data retrieval Interface presented in MMEA result seminar 25.5.2012 Real-time LAPS data Partial implementation of WCS 2.0 XML data retrieval; air temperature, relative humidity, wind speed, wind direction, precipitation 1 hour Textual format, suitable for small grid datasets Data delivery was working out very well! Problems with model time synchronization (NTP) in MTT influenced data retrieval in 2012. Aikojen kohdistaminen, mallin säätietopalvelu kellotettu hakemaan dataa eri lähteistä © Maa- ja elintarviketalouden tutkimuskeskus 31.10.201731.10.2017
SoilWeather (in situ) weather station data retrieval SoilWeather observation network in Karjaanjoki area, part of CropInfra in Vihti Network monitors air, soil and water (air temperature, air humidity, wind, precipitation, soil moisture, water level, water temperature, turbidity, nitrate content..) We used the data from automatic weather station on Hovi field (temperature and air humidity) Data measured in 15 min time interval; station-specific xml request from a-Lab server every 1 hour Data delivery was established and worked out well Built procedure to replace data gaps with value from database to enable gradient calculations in the model © Maa- ja elintarviketalouden tutkimuskeskus 31.10.201731.10.2017
Temperature Modelled (LAPS) temperature and relative air humidity were well correlated with in situ measurements on the field
Comparison of two weather input datasets 31.10.2017 Model results (cumulated risk value) for individual field plots in Finland Comparison of two weather input datasets Cumulated risk value Based on in situ measurement Cumulated risk value based on FMI (1 km2) data © Maa- ja elintarviketalouden tutkimuskeskus 31.10.2017
31.10.2017 Agricultural (biological) models and data interoperability – first experiences Having choice over weather input data to agro models is excellent, and FMI weather data was seen as a fine source of information to replace or complement data from compact in-situ weather stations Easy access to data from FMI / MMEA interfaces. Biological expertise must be included in implementation process together with technical expertise. Continuous, good communication is needed because ways of thinking can differ and terminologies are often domain-specific. Information connected to measured value: parameter, unit, time and location of measurement, estimate of uncertainty of measurement...Here Input data values behaved nicely, but work with other information (especially time, location) needed and will need more attention than we expected Weather data used in model development vs weather data available in operational use (instrumented fields, or random farm in location X) © Maa- ja elintarviketalouden tutkimuskeskus 31.10.201731.10.2017
Work continues... Model testing 2012-2013, and model output for growth season 2013 Data quality assessment Weather data comparison in agricultural applications (Scientific article, 2012 manuscript) CropInfra environment will be developed towards supporting flexible service use, service relaying and service integration Support for model simulations (Monte Carlo approach) Our work will be useful for integration of other crop disease and pest models © Maa- ja elintarviketalouden tutkimuskeskus 31.10.201731.10.2017