Phenology calibration program

Submitted by pepijn.vanoort on Mon, 02/08/2016 - 14:26
Phenology calibration program
Program type
Calibration program
Available since

A program that allows you to estimate cardinal temperatures and daylength sensitivity parameters and transplanting shock parameter from field observations.

Given sufficient data (phenology and nearby weatherstation), either from multiple locations or from multiple observations at one or more locations, it is possible to estimate phenological parameters from field data. Estimation of these parameters is important to simulate effects of variety choice, sowing date and climate change on the expected days from sowing to flowering and maturity. Use of invalid cardinal temperatures can lead to overestimation of the shortening of this period (and associated yield loss) when temperatures rise due to climate change, see Zhang et al. (2008) and van Oort et al. (2011).

This calibration program is a more general version than the original rice phenology calibration program published by van Oort et al. (2011). The more general calibration program posted here was developed by Pepijn van Oort. This more general version was developed for ongoing rice phenology calibration at AfricaRice and in a small BSc project by Gerdine van Schothorst at Wageningen university, using data collected by Michiel de Vries and Joordens' Zaadhandel B.V.

Important changes relative to the original rice phenology calibration program are:

  1. The sinus diurnal temperature model has been replaced with the more accurate sinus-exponential model (SUBDD3.f90). More on this model is found in Parton and Logan (1981), Goudriaan and van Laar (1994) and Ephrath et al. (1996)
  2. Photoperiodic Daylength (DAYLP) can now be calculated by defining till how many degrees below horizon there is still light (FUNCTION DAYLP_CAL in Common_opt3.f90). Default setting is 4 degrees below horizon (BASE = -4.). If you like you can change this parameter.
  3. It is now possible to simulate both short and long day plants. Activate the appropriate code lines in SUBDD3.f90, search for "!PO20150408: delay in development for long days". (note the original rice phenology calibration program would only simulate short day plants, while we found the two brassica species White mustard (Sinapis alba) and Oilseed radish (Raphanus sativus var. Oleiformis) to be long day plants. Hence this extension to the original calibration program)
  4. The model calculates the average daylength from emergence to flowering (DLMEFL) and sends this to output. Plotting observed duration from emergence to flowering (EFLOBS) against DLMEFL (both can be found in "Output_model.txt") provides a quick indication about whether there is daylength sensitivity.
  5. The option to simulate with daylength sensitivity only has been added. This option is activated by setting TBD = 99oC.
  6. The option to simulate with development rate increasing linearly with temperature without plateau at a given optimal temperature TOD has been added. Activate the bilinear0 model for this option.

Yes I know, it's all a bit nuts and bolts. And yes, you need to do a bit of programming there yourself. No, there is no user interface. You have compile code yourself, which is actually quite easy if you read the manual. You have to open output text files and sort them by lowest RMSE yourself. And yes I know we shouldn't be specifying parameters in the source code (which forces you to recompile every time again). But hey it works! Give me more time and I do a better job. On the positive side: source code gives you full control. If you want, you can dig in the code, find out exactly what it's doing. And if you don't want to know what you're doing, then why are you here?

Scale of application
Spatial resolution
Key outputs
Cardinal temperatures Daylength sensitivity parameters Temperature sums Model accuracy for each set of parameters You have to select the best set yourself, based on your criteria. The model gives you the accuracy for each possible set of parameters so you can also analyse parameter convergence. If you get the same accuracy with very different parameter sets that probably means your sample is too small for parameter estimation, which is also an important outcome.
Time horizon
Time step of modeling
Required to run
  1. Weather data (daily Tmin and Tmax)
  2. Latitude (used for daylength calculation)
  3. Phenology observations:
    1. Emergence or sowing date
    2. transplanting date (if no transplanting trick the model by setting transplanting date to emergence date plus one and set transplantingf shock parameter to zero)
    3. Panicle initiation date (if no data then estimate, or choose to simulate by making the complete period from emergence to flowering photo period sensitive)
    4. Flowering date (if no data then estimate, and report how you did this)
    5. Maturity date (if no data then estimate, and report how you did this)
Required to develop
Fortran compiler. Check here for a instruction for how to get a freeware fortran compiler up and running. Yes I know, it's all a bit nuts and bolts. I know we shouldn't be specifying parameters in the source code (which forces you to recompile every time again). But hey it works. Give me more time and I do a better job.
Database I/O
Applications & Use
Once an accurate model and parameters have been found, this model can be used to simulate phenology and incorporated in a crop growth model. One possible application is the following, where the model is used for sowing date optimisation and variety choice. Cover crops such as White mustard (Sinapis alba) and Oilseed radish (Raphanus sativus var. Oleiformis) are used as cover crops. They are often grown in a crop rotation sown in autumn after harvesting a winter cereal. The cover crop dies during winter or is plowed into the soil in winter/spring. They protect the soil and add organic matter. Crop grows flattens off after flowering. If allowed to flower, seeds can unwantedly emerge in the subsequent crop. So farmers want the crop as long in the field as possible to maximise soil protection and biomass production, but without reaching the flowering stage. Once calibrated, a model can help simulate number of days to flowering for a range of sowing dates and from this the optimum sowing window can be selected.
P.A.J. van Oort