This site has been archived on 18/03/2015
Modelling concept for nutrients

Nutrients are essential for plants, animals and humans, but in excessive amounts they can lead to a significant decrease in soil and water quality. Strict regulations have been adopted to limit their impact on the environment, however detrimental effects are still observed nowadays. Modelling offers a unique opportunity to identify main sources of nutrients in European ecosystems, and help develop sound and sustainable scenarios and policies in view of reducing the impact of excessive nitrgen and phosphorus in the environment. A tiered approach for addressing the nutrient fate at various scale, making best use of readily pan-European available data while taking into account policy requirements has been setup within FATE. A statistical model (GREEN) was developped as a screening tool to identify hot spots and perform nitrogen and phosphorus source apportionment. A number of physically based models (e.g. EPIC, SWAT) are used to elaborate sound best farming and management practices preserving the environment. To support these modelling activities, a harmonized pan-European environmental deatabase was developped.

Statistical modelling


GREEN modeling conceptThe GREEN (Geospatial Regression Equation for European Nutrient losses) model consists of a regression equation based on spatially referenced input data. The GREEN model was developed to estimate the annual nitrogen and phosphorus loads and perform source apportionment at pan-European level (Grizzetti, 2006) . Availability of data is a major constraint in modelling based assessments. The powerful feature of GREEN is its validated capacity for accurate predictions even in cases of limited data availability.

Physically-based modelling


flow chart of the EPIC modelWe designed a specific data framework using ESRI ArcGis and a physically based model to estimate nutrient and water requirements as affected by farming practices in Europe and Africa. More specifically, we applied EPIC (Williams, 1995), a continuous physically-based farm scale plant growth model used widely to determine the effects of management strategies on agricultural production and soil and water resources.

Pan-European database


study area and monitoring pointsTo support the nutrient modelling activities, the JRC developed a EU-wide environmental database including features related to river basin morphology, nutrient pressure originating from anthropogenic activitites (e.g. agriculture, urban agglomerations) and climate. The database includes both spatial and non-spatial data and covers a time period ranging from 1985 to 2005.

load_to_seaGREEN considers two different pathways of nutrient transfer from sources to the catchment outlet. Diffuse sources (DS) that include fertilizers (artificial and manure), atmospheric deposition, and scattered dwellings, are first reduced in the soil and then partially retained in the streams. Point sources (PS), which include discharges from sewers, wastewater treatment plants, industries, and paved areas, are retained only in the streams. Retention is calculated as a exponential function of river length for in-stream retention, and as a exponential function of annual precipitation for the basin retention.


The model requires the following information to run:

  • A digital river network and river basins
  • Input of mineral and manure fertilizers (nitrogen and phosphorus)
  • Nutrient discharge from urban waste water treatment plants
  • Industrial nutrient discharge
  • Nutrient emissions from scattered dwelling
  • Nitrogen atmospheric deposition
  • Annual precipitation


The model will provide the following outputs:

  • Nutrient loads per river basin
  • Nutrient diffuse emission at river basin scale
  • Nutrient source apportionment at river basin scale

Irrigation requirementsEPIC main components include hydrology, weather, erosion, nutrient cycling and transport, plant growth, farm economic account. EPIC takes into account crop type and varieties, crop rotation, fertilizer and irrigation strategies, tillage practices. EU 27 plus Switzerland were divided into a 10X10 km grid (49,000 cells) where the top five dominent crops are reprensented. Each site can be seen as a fictitious field with an area equal to the crop area within the cell with a homogeneous soil, meteorological and topographic attributes (Bouraoui and Aloe, 2007).


The model requires the following information to run:

  • Topography
  • Soil properties
  • Land use and land management
  • Climate information


The model will provide the following outputs:

  • Nutrients uptake by crops
  • Nutrient losses via particulate or dissolved forms
  • Water balance including irrigation requirements.


The model was then extended to Africa where actual simulations are performed to estimate impacts of agricultural practices on water and nutrient management.

Point Source Emisssions

point_sourceThe estimations of N and P release from waste water treatment plants from agglomerations is based on the population density, the percentage of population connected to the sewerage system (taken from EUROSTAT), the level of treatment (EUROSTAT), the N and P abatment for each waste water treatment type, and N and P emission factor per person. The population density was obtained from the HYDE database, for the years 1980, 1990, 2000 and 2005 on a 5mn resolution (Klein Goldewijk and Van Drecht, 2006). references missing



Fertilizers Application

land_use_2000The fertilizer application rate per crop for N and P for EU27, Norway and the Balkans were obtained from the agro-economical model CAPRI ( Britz, 2004). For the remaining countries, fertilization rates were taken from FAO [FAOSTAT, 2009]. The application of fertilizers taking place only on arable land was derived from the Global Land Cover map for year 2000 (GLC2000), while the crop distribtion was estimated from the SAGE database (Ramankutty et al., 2008). Fertilizer applications were derived for the years 1980, 1990, 2000 and 2005. 



Climate Database

atmospheric depositionThe climate database (Princeton Climate Database) was provided by Sheffield et al. (2006). The dataset are available on a global grid of 1 degree resolution for a time period extending from 1948 to 2006. The basic variables include daily precipitation, short and long waves radiations, surface air pressure, specific humidity and wind speed.  The CRU long term average climate information was used to downscale the Princeton data to a 10 mn resolution. This data was combined with information coming from EMEP to wet and dry N atmospheric deposition.



Faycal Bouraoui