maize-Toff

Repository for ecohydrological modeling analysis of maize yield variability and tradeoffs between yield and crop failure. See article by Krell et al. (2021) "Consequences of dryland maize planting decisions under increased seasonal rainfall variability" in Water Resources Research.

Set-up

Create fork of maize-Toff and git clone to local machine.

1. cd maize-Toff
2. conda env create -f environment.yml -n maize-Toff
3. conda activate maize-Toff
4. jupyter notebook

Note: To update dependecies in an existing environment, use conda env update --file environment.yml after step four.

Directory structure

farm/

• where models are stored

data/

• contains CETRAD rainfall data, maize variety info

output/

• exported figures, results

notebooks/

• contains notebook that generates figures from the manuscript

How to use this model

Generate a Soil object

The soil object contains all the necessary parameters to specify soil hydraulic properties and texture. The easiest way to generate a soil instance is to use one of the standard USDA soil texture types: Sand, Loamy Sand, Sandy Loam, Silt Loam, Loam, Sandy Clay Loam, Clay Loam, Sandy Clay, Silty Clay, Clay, and Sandy Silty Loamy Clay. Just kidding. That last one is totally not a soil type.

# Creates a sand soil object:
soil = Soil('sand') 

You can also specify your own parameters:

param_dict = {
    'b': 11.4,
    'Psi_S_cm': 40.5,   # saturated water tension, cm
    'Psi_l_cm': 24.3,   # leakage water tension, cm
    'n': 0.482,         # porosity, cm^3/cm^3 (is Psi_S) in C&H,
    'Ks': 0.0077,   # saturated hydraulic conductivity, cm/min
    'S': 0.268      # sorptivity, cm/min^1/2  
}

custom_soil = Soil(params=param_dict)

Of the parameters, only b, Psi_S_cm, Psi_l_cm, n, and Ks are required. The model doesn't use S right now.

Generate a Climate object

The climate object has information on maximum evapotranspiration, as well as a time series of rainfall. The rainfall timeseries is generated stochastically using the parameters of storm depth, frequency, and the length of the season. The storm depth (alpha_r) specifies the average daily storm depth, assuming that daily storm depths are drawn from an exponential distribution. The frequency (lambda_r) is best described as the daily probability of rainfall. The length of the season ends up setting the timescale of the simulation in days.

There are several keyword arguments available:

• alpha_r Average storm depth [mm]
• lambda_r Frequency of storms [day^-1]
• t_seas Length of rainy season [days]
• ET_max Maximum evapotranspiration [mm/day]

Keyword arguments can be specified when instantiating the object, or the following are the resonable defaults values:

climate = Climate(
    alpha_r=10.0,
    lambda_r=0.3,
    t_seas=180,
    ET_max=6.5)

Generate a Plant

The plant object requires the most parameters to generate, and some of these parameters depend on the soil in which the plant is growing. In addition, the plant object can be subclassed into specific plant types to allow for varying structures and function. In this simulation, we are using the Crop subclass, which is initialized with the minimum following parameters:

• kc_max The maximum crop coefficient [dimensionless], which is a scale factor applied to ET_max to determine the maximum rate of plant transpiration, T_max.
• LAI_max The maximum crop leaf area [m^2/m^2].
• T_max The maximum rate of crop water use in [mm/day]
• soil A soil object that specifies the soil that this crop is growing in.

crop = Crop(kc_max=1.2, LAI_max=2.0, T_max=4.0, soil=soil)

Additional parameters that should be specified are:

• Zr Plant rooting depth [mm]. Default is 500mm.
• sw_MPa Plant wilting point [MPa]. Default is -1.5MPa
• s_star_MPa Water potential of maximum water use [MPa]. Default is -0.2 MPa.

If these parameters are not provided, default values are inherited from the Plant class.

Initialize and run the model

Combining the prior three steps, we can create a model instance and run the model:

# Import the necessary objects:
from farm.climate import Climate
from farm.soil import Soil
from farm.plant import Crop
from farm.model import CropModel

# Make the things
climate = Climate() # uses default climate values
soil = Soil('sand')
crop = Crop(kc_max=1.2, LAI_max=2.0, T_max=4.0, soil=soil)

# Create the model
model = CropModel(crop=crop,soil=soil,climate=climate)

# RUN IT.
model.run() # TADA!

model.output()

Getting model output

The model.output() function returns all the simulation output structured as a single pandas DataFrame. The frame has the following columns:

• kc Time series of daily crop coefficients.
• LAI Time series of crop LAI
• R Time series of daily rainfall [mm]
• s Time series of daily relative soil moisture [0-1]
• I Time series of daily interception [mm]
• E Time series of daily soil evaporation [mm]
• T Time series of daily plant transpiration [mm]
• L Time series of soil leakage loss [mm]
• Q Time series of surface runoff [mm]
• dsdt Time series of changing relative soil moisture

To plot any of this data, simply use the .plot() command:

output = model.output()

# Plots a time series of simulated evapotranspiration:
output['ET'].plot()

# Plots a time series of simulated relative soil moisture
output['s'].plot()

Test the model

Basic testing framework

From the root of the directory, run the following in the command line to put farm into your python path in editable mode.

pip install -e .

After that, run tests from the subdirectory farm/tests, for example:

nosetests -vv test_sand

Checking test coverage

Update the coverage for the model before pushing new commits, or in advance of a pull request. You can update the coverage using the following command:

nosetests -vv ./farm/tests --with-coverage --cover-package=farm --cover-html --cover-html-dir=coverage_html_report/