A dynamic mixed integer bio-economic farm scale model
FARMDYN provides a flexible, modular template to simulate different farming systems (dairy, mother cows, beef fattening, pig fattening, piglet production, arable farming, biogas plants) at single farm scale.
Institute for Food and Resource Economics
Economic Modeling of Agricultural Systems Group
- Multiple dynamics including comparative-static, short run, or fully dynamic, with simulations covering several decades
- Integer variables capture indivisibilities in investments (machinery, buildings) and labour use
- Selected farm management decisions (e.g. feeding, manure management, labour use) depicted with a sub-annual temporal resolution, partially bi-weekly
- Farm labour, machinery and stable use are modelled in rich detail
- Highly detailed farm branch activities (e.g. intensities for arable and grass crops, differentiated feeding schemes for all animal types accounting for lactation/feeding phase, etc.)
- The machinery park is available in different mechanization levels
- Environmental accounting modules including the flow from different nitrogen compounds, CO2eq, phosphorus compounds
- A wide range of agri- and agrienvironmental policies including CAP, German implementation of Nitrate Directive, agri-environmental schemes, etc.
- Multiple biodiversity indicators
- Parameterized for multiple countries besides Germany it includes Switzerland, Norway, Netherlands
- Deterministic or stochastic programming version. The latter treats all variables as state dependent, allows for sceneario tree reduction and covers different risk measures (value at risk, MOTAD ...)
For the German version, the model is parameterized using highly detailed farm planning data provided by KTBL in combination with farm structural statistics. It offers a complementary approach to other farm scale models used in the institute such as the farm group models integrated in CAPRI or FADN based farm-scale progamming models which both are comparative-static, calibrated against observed farm programs with Positive Mathematical Programming while being far less detailed with regard to technology, and not comprising explicit investement decisions.
The model is realized in GAMS, solved with the industry MIP solver CPLEX, linked to a Graphical User Interface realized in GGIG and hosted on a Software Versioning System. Design of experiments, building on R routines directly called from GAMS, can be used in combination with farm structural statistics to systematically simulate different farm realizations (assets, farm branches) and boundary conditions such as input and output prices or emisisons ceilings using a computing server to solve several instances in parallel. That approach has e.g. been used to estimate a statistical meta model for Marginal Abatement Costs of Green House Gases from dairy farms. Code development and testing follows agreed upon guidelines.