Difference between revisions of "Weather Monitoring"

From Agri4castWiki
Jump to: navigation, search
(Aggregation to regions)
(Climatology and analysis)
 
(200 intermediate revisions by 3 users not shown)
Line 1: Line 1:
{{Popular}}
+
__NOTOC__
 +
{{Non expert}}
 
==General description==
 
==General description==
[[File:architecture_weather_monitoring.jpg|thumb|right|300px
+
[[File:Flowchart_mcyfs_modules_weather.jpg|thumb|right|200px|The role of weather monitoring within the MCYFS]]
|General layout of the weather monitoring components in MCYFS]]
+
The weather monitoring module is one of the five modules of the MCYFS and can be split in four procedures.
 +
#[[#Acquisition|Acquisition]]
 +
#[[#Interpolation|Interpolation]]
 +
#[[#Aggregation|Aggregation]]
 +
#[[#Climatology and analysis|Climatology and analysis]]
  
The weather monitoring module is one of the main elements of the MCYFS. Input data that go into the module are weather station observations and weather forecast data.
 
  
 +
The output of the weather monitoring module is used in two ways: firstly, to derive agro-meteorological indicators for a direct evaluation of alarming situations such as drought, extreme rainfall during sowing, flowering or harvest etc., and secondly, as input to the [[Crop Simulation|crop simulation]] module to simulate crops behaviour and to evaluate the effect of weather on crops.
  
The module contains 5 procedures. The first procedure handles data [[#Acquisition of observed weather|acquisition]] of observed weather. The second procedure [[#Interpolation of observed weather to 25x25km grid|interpolates]] the cleaned station weather to a 25x25km grid. A third and fourth procedure handle the [[#Acquisition of forecasted weather|acquisition]] and [[#Downscaling of forecasted weather to 25x25km grid|downscaling]] of forecasted weather from European Centre for Medium-Range Weather Forecasts (ECMWF) to the same 25x25km grid. Finally the grid weather from observations and forecasts is [[#Aggregation to regions|aggregated to regions]] in a fifth procedure.
+
==Acquisition==
 +
[[File:Weather_station.jpg|thumb|right|200px|Weather Station, Garreg Fawr, Aberdaron]]
 +
====Observed weather====
 +
Each day raw data of at least 4200 synoptic weather stations, that regularly collect and supply one or more meteorological variables, are acquired over Europe and its neighbourhood and are added as raw data to the station weather database. The variables collected include air temperature, precipitation, radiation, air humidity, and wind speed. All incoming data are checked for errors and dubious values, such as e.g. air temperatures outside a realistic range. The incoming 3- or 6-hourly data are then converted to daily values that fit in an uniform station weather database. Some variables, required for the crop simulation module, are not (or not regularly) observed. Such variables, for example solar radiation or evapotranspiration, are derived from the measured data and also added to the database.
  
 +
[[File:Supercomputers_at_ecmwf.jpg|thumb|right|200px|Supercomputer at ECMWF]]
  
The output of the weather monitoring module is used in two ways for crop yield evaluations. In the first place as input for the [[Crop Simulation|crop simulation]] module to simulate crops behaviors and evaluate the effects of weather on crops yields. Secondly as weather indicators for a direct evaluation of alarming situations such as drought, extreme rainfall during sowing, flowering or harvest etc.
+
====Forecasted weather====
 +
In addition to observed weather data also weather forecast data are loaded into the system so that crop development and biomass accumulation can be simulated into the future (see [[Crop Simulation|crop simulation module]]), reaching closer to the end of the crop season and therefore advancing crop yield forecasts along the season (see [[Yield Forecasting|yield forecasting module]]).
  
 +
Six data products from {{Gloshint|ECMWF|European Centre for Medium-Range Weather Forecasts. |ECMWF}} are loaded into the system:
 +
*ERA-Interim (ERA)
 +
*Analysis model (HIS)
 +
*Deterministic forecast model (OPE)
 +
*Ensemble Prediction System (ENS)
 +
*Monthly forecast model (ENSEXT)
 +
*Seasonal forecast model (SEAS)
  
The crops behaviors are mainly influenced by the atmospheric conditions near the earth surface. Considering the data availability, resources and purpose of the system a time scale of one day and a spatial scale of 25x25km are chosen as the resolutions to estimate crop yields at European scale.
+
These products have a different number of forecast days (forecast depth) and a varying number of possible weather realizations called 'members'. Different members can be thought of as model runs with a slightly different initialization and thus slightly different results, but with equal validity. Similar to observed weather, meteorological variables such as air temperature, precipitation, and solar radiation are directly retrieved from the forecast data. Additional parameters such as evapotranspiration are calculated from these variables within the MCYFS.
  
'''External links:'''
+
{{Scientific_box_2|
*[http://www.ecmwf.int/ European Centre for Medium-Range Weather Forecasts (ECMWF)]
+
*[[Meteorological data from ground stations]]
 
+
*[[Meteorological data from ECMWF models]]
==Acquisition of observed weather==
+
}}
[[file:acquisition_of_observed_weather.jpg|thumb|right|300px|Acquisition of observed weather]]
 
Every day the raw data of at least 3000 stations that regularly collect one or more parameters and supply them in near real time are acquired and added to the raw station weather database. For the year 2010 over 3400 stations have contributed the database. Over time another 2800 stations have a usefull archive. Weather parameters that are collected include:
 
* Precipitation (daily and 6-hourly)
 
* Temperature (daily maximum, daily minimum and 3-hourly)
 
* Measured radiation
 
* Sunshine
 
* Cloud cover
 
* Vapour pressure
 
* Wind speed
 
* Snow depth
 
* Humidity (3-hourly)
 
 
 
 
 
The archive data and near real time incoming parameters values are checked for errors such as temperatures that are too low or parameters values that don't change over time. Errors are corrected and the data are processed to daily parameters that fit in a uniform station weather database.
 
 
 
{{Scientific_box|[[Meteorological data from ground stations]]}}
 
 
 
Because some parameters that are needed in the crop simulation module are not sufficiently measured by weather stations some advanced parameters are calculated from basis parameters:
 
* Calculated radiation at surface
 
* Transpiration of water surface
 
* Transpiration of wet bare soil
 
* Evapotranspiration
 
  
{{Scientific_box|[[Calculation of advanced parameters]]}}
+
==Interpolation==
 +
[[File:Interpolating_observed_weather.jpg‎‎|Interpolation from weather stations to  25 x 25 km regular climate grid.|thumb|200px]]
 +
====Observed weather====
 +
Observed weather, aquired from weather stations, have an irregular distribution and density in space. Data of a single weather station are representative for the location of that station only. To construct weather data for locations in between stations a conversion is needed. Interpolation (constructing new data points within the range of a discrete set of known data points) is one of the methods to do this. In the MCYFS this procedure is used to convert irregular distributed station data to regular distributed data. The regular distribution is organized as a grid with side by side grid cells of 25 kilometre width and 25 kilometre length that covers the entire region of interest (e.g. Europe). This is called the regular climatic grid. The interpolation is managed by a sub-system called {{Hint|CGMS|Crop Growth Monitoring System}}.
  
 +
[[File:Downscalling_eps_mon.jpg‎|Interpolation from 0.5 x 0.5 degrees grid to  25 x 25 km regular climate grid.|thumb|200px]]
  
The observed and calculated data are stored as archive and used for:
+
====Forecasted weather====
* [[#Climatology (long term average weather)|climatolgy]]
+
Weather forecast data come already as spatial fields, not point data as observations, but in varying spatial resolutions and projections that are different from the regular climatic grid used in the MCYFS. Therefore, forecast data have to be interpolated from the 'source' grid to the 'target' grid: the regular climatic grid of 25 by 25 km. This specific interpolation procedure is also called "downscaling", because usually it converts a low resolution source data into higher resolution target data.
* [[Crop Simulation|crop simulation]] until the current day.
 
  
==Interpolation of observed weather to 25x25km grid==
+
{{Scientific_box_2|
The interpolation is managed by a sub-system called Crop Growth Monitoring System (CGMS). All input data and output data of CGMS is stored in a relational database.
+
*[[Interpolation of observed weather]]
 
+
*[[Interpolation of forecasted weather]]
{{analyst_box|
 
* Database structure: [[Appendix 4: CGMS DB description#Appendix 4|Appendix 4 (CGMS DB description)]]
 
* Table descriptions: [[Appendix 5: CGMS tables#Appendix 5|Appendix 5 (CGMS tables)]].
 
* Description of software procedures: [[Appendix 3: Overview of the software#Appendix 3|Appendix 3 (Overview of the software)]].
 
 
}}
 
}}
  
 +
==Aggregation==
 +
[[File:Aggregation_to_regions.jpg|Example of four different administrative levels in combination with landcover type 'arable land' on the 25 km grid.|thumb|200px]]
 +
The interpolation procedure generates gridded datasets of observed and forecast weather. In order to describe weather at a spatial domain larger than the grid size and to answer questions like e.g.:
  
[[file:interpolation_of_observed_weather.jpg|thumb|right|300px|Interpolation of observed weather]]
+
''''What was the average temperature in northern France during the last week for locations where winter wheat is grown?''''
For the current year (near real time interpolation) for each grid cell every day the most suitable stations are selected. This is based on the  stations location and altitude comparison to the grid cell location and altitude. The data from the suitable stations are used to interpolate to a specific grid value for a specific day. Due to this procedure the suitable stations can differ between days and even between parameters. In case no suitable stations are found, a long term average value is substituted to ensure spatial and temporal continuity.
 
 
 
For interpolation of archive station weather each grid cell uses the same suitable stations for a complete year. For most recent years around 2700 stations have a sufficient temporal coverage (enough observations within one year). Only these are used in the interpolation procedure. In this way a grid weather archive is build up with daily weather for each grid cell going back to 1975.
 
 
 
 
 
Having the grid weather database other weather indicators can be derived:
 
* average day temperature
 
* climatic water balance
 
 
 
 
 
Some of the weather indicators are defined over an arbitrary period:
 
* sum, max, min and average of a parameter
 
* number of heat waves
 
* longest heat wave period
 
* number of hot day's
 
* number of cold day's
 
* number of day's with significant rainfall
 
 
 
{{Scientific_box|[[Interpolation method onto regular climatic grid]]}}
 
 
 
==Acquisition of forecasted weather==
 
 
 
5 products from ECMWF are loaded into the system:
 
* Analysis model
 
* Deterministic forecast
 
* Ensemble Prediction System
 
* Monthly forecast
 
* Seasonal forecast
 
 
 
The products have a different number of forecast days and a varying number of possible results called 'members'. Different members can thought of as model runs with a slightly different initialization and thus slightly different forecast results.
 
 
 
{|class="wikitable"
 
!Model !! Abbreviation !!Forecast days !! Members !! Spatial resolution !! Delivery
 
|-
 
|Analysis model || HIS || 1 || 1 || 0.25° x 0.25° || Daily (10.30 hr)
 
|-
 
|Deterministic forecast || OPE ||10 || 1 || 0.25° x 0.25° || Daily (12.00 hr)
 
|-
 
|Ensemble Prediction System || EPS ||15 || 51 || 0.5° x 0.5° || Daily (14.00 hr)
 
|-
 
|Monthly forecast || MON ||32 || 50 || 0.5° x 0.5° || Every Friday (03.00 hr)
 
|-
 
|Seasonal forecast || SEA ||170 || 40 || 0.25° x 0.25° || Every 15th of the month (14.00 hr)
 
|}
 
 
 
The acquired weather parameters are:
 
* Precipitation
 
* Temperature (daily maximum, minimum and average)
 
* Dewpoint temperature
 
* Global radiation
 
* Snow depth
 
* Wind speed
 
 
 
Some advanced parameters are calculated from basis parameters:
 
* Vapour pressure
 
* Transpiration of water surface
 
* Transpiration of wet bare soil
 
* Evapotranspiration
 
 
 
 
 
The HIS model data are stored as archive and used for:
 
* (model based) [[#Climatology (long term average weather)|climatolgy]]
 
* (model based) [[Crop Simulation|crop simulation]] until the current day.
 
 
 
The most recent available forecast data of the other models (OPE, EPS, MON, SEA) are used to finish [[Crop Simulation|crop simulation]] as far into the future as the forecast model allows.
 
 
 
{{Scientific_box|[[Meteorological data from ECMWF models]]}}
 
 
 
'''External links:'''
 
*[http://www.ecmwf.int/ European Centre for Medium-Range Weather Forecasts (ECMWF)]
 
 
 
==Downscaling of forecasted weather to 25x25km grid==
 
[[File:Downscalling_ope_his.jpg‎|Example of source grid (black dots) and target grid (gray lines)|thumb|200px]]
 
The data of different forecast models ('source') are acquired in different spatial resolutions (sea '[[#Acquisition of observed weather|Acquisition of observed weather]]'). The downscalling procedure converts these data into a 25x25 km ('target') resolution which can be used in the [[Crop Simulation|crop simulation]] module.
 
 
 
 
 
In the first part of the downscalling method parameter values for a target gridcell are derived from the 4 closest source gridcells where closer source gridcells have a higher weight (inverse distance weighting).
 
 
 
 
 
In the second part of the downscalling a correction is made for systematic bias (systematic deviations between estimated downscalled source data and observed target data). For instance: temperature is correlated with height. The heigt values in the source grid are different from the height values in the target grid (they sample different locations). Therefor downscalled temperature will be biased depending upon the height differences.
 
 
 
{{Scientific_box|[[Processing ECMWF model data]]}}
 
 
 
 
 
'''External links:'''
 
[http://en.wikipedia.org/wiki/Inverse_distance_weighting Inverse Distance Weighting (IDW)]
 
 
 
==Aggregation to regions==
 
[[File:Aggregation_to_regions.jpg|Example of 4 different administrative levels in combination with landcover type 'arable land' on 25 x 25 km grid.|thumb|200px]]
 
The primary results of the interpolating observed stations weather and downscalled forecast weather is grid weather. To answer questions like:
 
 
 
 
 
''''What was the average temperature in France during the last week for locations where winter wheat is grown?''''
 
 
 
 
 
the grid weather data are aggregated to different levels of administrative regions and different levels of agri-environmental regions while weighting each gridcells for the area covered by a certain crop or landcover type. Observed grid weather and forecasted grid weather are both aggregated to 90 combinations:
 
* 4 levels of administrative regions for 12 crops and 7 landcover types
 
* 2 levels of agri-environmental regions for 7 landcover types.
 
 
 
The aggregation of forecasted grid weather is restricted to the HIS model, OPE model and the median of all EPS member values per forecast day.
 
  
 +
gridded weather data are aggregated to different types of regions. Observed and forecast gridded weather data are aggregated to different levels of administrative regions for a number of landcover types. This aggregation is based on a weight of each grid cell for the area covered by the selected landcover type.
  
A second set of 90 combinations is aggregated where only gridcells are taken into account that have at least 5 % coverage of a certain crop or landcover type. For example:
+
Altogether, the aggregation procedure results in many aggregated weather data sets based on the data type (e.g. observed, ERA-Interim, forecast), regions and landcover types.
*administrative region: 'country' level
 
*landcover type: 'arable land'
 
*threshold: 5%
 
  
In this combination all grid cells that have at least 5% coverage of landcover type 'arable land' contribute to the aggregated values for 'countries' they are contained by. The contribution of each grid cell that passes the threshold of 5% is weighted by the area covered with landcover type 'arable land'. Grid cells with less than 5% coverage are not taken into account.
+
{{Scientific_box_2|[[Aggregation of weather indicators]]}}
  
{{Scientific_box|[[Aggregation of weather indicators]]}}
+
==Climatology and analysis==
 +
[[File:Climatology.jpg|Long-term average air temperature over period January-June on a 25 km resolution.|thumb|200px]]
 +
Within the MCYFS climatology is considered as the long-term average of weather indicators. The long-term averages - or 'normal' conditions - are essential to understand how current weather conditions relate to the situation that was 'normal' in the past.
  
==Climatology (long term average weather)==
+
Long-term average daily values are calculated over the years in the archive for all spatial resolutions defined in each regional window. Two periods are considered: 1975-last year and 1995-last year.
Climatology is essential to understand how current weather conditions relate to the normal situation. Long term average values are available for observed weather and forecasted weather (Analysis model only, see '[[#Acquisition of forecasted weather|acquisition of forecasted weather]]') for different resolutions (25 x 25 km grid, administrative regions and agri-environmental regions).
 
  
The following parameters are available:
+
In addition to averages of the basic indicators (such as precipitation or air temperature) additional statistics are calculated, for instance the probability of having a rainy day, defined as a day receiving more than a certain amount of rainfall (5, 10, 15 mm). It allows to compare extreme weather events of the current year to extremes of the past. For example, the number of rainy days (with more than 5 mm/d) of last month can be compared to the average number of rainy days of this month that occurred in the past.
* Precipitation
 
* Temperature (daily maximum, minimum and average)
 
* Dewpoint temperature (forecasted weather only)
 
* Vapour pressure
 
* Wind speed
 
* Snow depth
 
* Calculated radiation at surface (observed), global radiation (forecasted)
 
* Transpiration of water surface
 
* Transpiration of wet bare soiles
 
* Evapotranspiration
 
* fraction of days wit daily maximum temperature greater than 25 degrees celsius
 
* fraction of days wit daily maximum temperature greater than 30 degrees celsius
 
* fraction of days wit daily maximum temperature greater than 35 degrees celsius
 
* fraction of days wit daily maximum temperature less than 0 degrees celsius
 
* fraction of days wit daily maximum temperature less than minus 8 degrees celsius
 
* fraction of days wit daily maximum temperature less than minus 10 degrees celsius
 
* fraction of days wit daily maximum temperature less than minus 18 degrees celsius
 
* fraction of days wit daily maximum temperature less than minus 20 degrees celsius
 
* fraction of days wit rainfall greater than 5 mm
 
* fraction of days wit rainfall greater than 10 mm
 
* fraction of days wit rainfall greater than 15 mm
 
  
{{Scientific_box|
+
{{Scientific_box_2|
 
* [[Calculation of Climatology]]
 
* [[Calculation of Climatology]]
 
* [[Analysis of weather indicators]]
 
* [[Analysis of weather indicators]]
* [[Time series analysis at station, grid and regional level]]
 
 
}}
 
}}
  
[[Category:MCYFS modules]]
+
[[Category:MCYFS introduction]]
 +
[[Category:Weather Monitoring]]

Latest revision as of 13:51, 13 August 2018



General description

The role of weather monitoring within the MCYFS

The weather monitoring module is one of the five modules of the MCYFS and can be split in four procedures.

  1. Acquisition
  2. Interpolation
  3. Aggregation
  4. Climatology and analysis


The output of the weather monitoring module is used in two ways: firstly, to derive agro-meteorological indicators for a direct evaluation of alarming situations such as drought, extreme rainfall during sowing, flowering or harvest etc., and secondly, as input to the crop simulation module to simulate crops behaviour and to evaluate the effect of weather on crops.

Acquisition

Weather Station, Garreg Fawr, Aberdaron

Observed weather

Each day raw data of at least 4200 synoptic weather stations, that regularly collect and supply one or more meteorological variables, are acquired over Europe and its neighbourhood and are added as raw data to the station weather database. The variables collected include air temperature, precipitation, radiation, air humidity, and wind speed. All incoming data are checked for errors and dubious values, such as e.g. air temperatures outside a realistic range. The incoming 3- or 6-hourly data are then converted to daily values that fit in an uniform station weather database. Some variables, required for the crop simulation module, are not (or not regularly) observed. Such variables, for example solar radiation or evapotranspiration, are derived from the measured data and also added to the database.

Supercomputer at ECMWF

Forecasted weather

In addition to observed weather data also weather forecast data are loaded into the system so that crop development and biomass accumulation can be simulated into the future (see crop simulation module), reaching closer to the end of the crop season and therefore advancing crop yield forecasts along the season (see yield forecasting module).

Six data products from ECMWF are loaded into the system:

  • ERA-Interim (ERA)
  • Analysis model (HIS)
  • Deterministic forecast model (OPE)
  • Ensemble Prediction System (ENS)
  • Monthly forecast model (ENSEXT)
  • Seasonal forecast model (SEAS)

These products have a different number of forecast days (forecast depth) and a varying number of possible weather realizations called 'members'. Different members can be thought of as model runs with a slightly different initialization and thus slightly different results, but with equal validity. Similar to observed weather, meteorological variables such as air temperature, precipitation, and solar radiation are directly retrieved from the forecast data. Additional parameters such as evapotranspiration are calculated from these variables within the MCYFS.


Interpolation

Interpolation from weather stations to 25 x 25 km regular climate grid.

Observed weather

Observed weather, aquired from weather stations, have an irregular distribution and density in space. Data of a single weather station are representative for the location of that station only. To construct weather data for locations in between stations a conversion is needed. Interpolation (constructing new data points within the range of a discrete set of known data points) is one of the methods to do this. In the MCYFS this procedure is used to convert irregular distributed station data to regular distributed data. The regular distribution is organized as a grid with side by side grid cells of 25 kilometre width and 25 kilometre length that covers the entire region of interest (e.g. Europe). This is called the regular climatic grid. The interpolation is managed by a sub-system called CGMS.

Interpolation from 0.5 x 0.5 degrees grid to 25 x 25 km regular climate grid.

Forecasted weather

Weather forecast data come already as spatial fields, not point data as observations, but in varying spatial resolutions and projections that are different from the regular climatic grid used in the MCYFS. Therefore, forecast data have to be interpolated from the 'source' grid to the 'target' grid: the regular climatic grid of 25 by 25 km. This specific interpolation procedure is also called "downscaling", because usually it converts a low resolution source data into higher resolution target data.


Aggregation

Example of four different administrative levels in combination with landcover type 'arable land' on the 25 km grid.

The interpolation procedure generates gridded datasets of observed and forecast weather. In order to describe weather at a spatial domain larger than the grid size and to answer questions like e.g.:

'What was the average temperature in northern France during the last week for locations where winter wheat is grown?'

gridded weather data are aggregated to different types of regions. Observed and forecast gridded weather data are aggregated to different levels of administrative regions for a number of landcover types. This aggregation is based on a weight of each grid cell for the area covered by the selected landcover type.

Altogether, the aggregation procedure results in many aggregated weather data sets based on the data type (e.g. observed, ERA-Interim, forecast), regions and landcover types.


Climatology and analysis

Long-term average air temperature over period January-June on a 25 km resolution.

Within the MCYFS climatology is considered as the long-term average of weather indicators. The long-term averages - or 'normal' conditions - are essential to understand how current weather conditions relate to the situation that was 'normal' in the past.

Long-term average daily values are calculated over the years in the archive for all spatial resolutions defined in each regional window. Two periods are considered: 1975-last year and 1995-last year.

In addition to averages of the basic indicators (such as precipitation or air temperature) additional statistics are calculated, for instance the probability of having a rainy day, defined as a day receiving more than a certain amount of rainfall (5, 10, 15 mm). It allows to compare extreme weather events of the current year to extremes of the past. For example, the number of rainy days (with more than 5 mm/d) of last month can be compared to the average number of rainy days of this month that occurred in the past.