AquaCrop-OSPy: Bridging the gap between research and practice in crop-water modelling¶

This series of notebooks provides users with an introduction to AquaCrop-OSPy, an open-source Python implementation of the U.N. Food and Agriculture Organization (FAO) AquaCrop model. AquaCrop-OSPy is accompanied by a series of Jupyter notebooks, which guide users interactively through a range of common applications of the model. Only basic Python experience is required, and the notebooks can easily be extended and adapted by users for their own applications and needs.

This notebook series consists of four parts:

1. Running an AquaCrop-OSPy model
2. Estimation of irrigation water demands
3. Optimisation of irrigation management strategies
4. Projection of climate change impacts

Install and import AquaCrop-OSPy¶

Install and import aquacrop as we did in Notebook 1.

In [ ]:

# !pip install aquacrop

# from google.colab import output
# output.clear()

# !pip install aquacrop # from google.colab import output # output.clear()

If for any reason you would rather not compile the aquacrop modules ahead-of-time, you can run the following cell to run the notebook in pure python (N.B. it will run slower).

In [ ]:

# import os
# os.environ['DEVELOPMENT'] = 'True'

# import os # os.environ['DEVELOPMENT'] = 'True'

In [1]:

from aquacrop import AquaCropModel, Soil, Crop, InitialWaterContent, IrrigationManagement
from aquacrop.utils import prepare_weather, get_filepath

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

from aquacrop import AquaCropModel, Soil, Crop, InitialWaterContent, IrrigationManagement from aquacrop.utils import prepare_weather, get_filepath import pandas as pd import numpy as np import matplotlib.pyplot as plt import seaborn as sns

Notebook 2: Estimating irrigation water demands under different irrigation strategies¶

In Notebook 1, we learned how to create an AquaCropModel by selecting a weather data file, Soil, Crop and InitialWaterContent (initial water content). In this notebook, we show how AquaCrop-OSPy can be used to explore impacts of different irrigation management strategies on water use and crop yields. The example workflow below shows how different irrigation management practices can be defined in the model, and resulting impacts on water use productivity explored to support efficient irrigation scheduling and planning decisions.

We start by creating a weather DataFrame containing daily measurements of minimum temperature, maximum temperature, precipitation and reference evapotranspiration. In this example we will use the built in file containing weather data from Champion, Nebraska, USA. (link).

In [2]:

path = get_filepath('champion_climate.txt')
wdf = prepare_weather(path)
wdf

path = get_filepath('champion_climate.txt') wdf = prepare_weather(path) wdf

Out[2]:

	MinTemp	MaxTemp	Precipitation	ReferenceET	Date
0	-21.11	3.33	0.0	1.59	1982-01-01
1	-10.00	0.56	0.0	0.86	1982-01-02
2	-11.67	-2.22	0.0	0.72	1982-01-03
3	-12.22	7.22	0.0	1.71	1982-01-04
4	-14.44	-1.11	0.0	0.92	1982-01-05
...	...	...	...	...	...
13509	-7.93	-2.01	0.0	0.52	2018-12-27
13510	-15.52	-5.35	0.0	0.62	2018-12-28
13511	-17.23	2.76	0.0	1.37	2018-12-29
13512	-11.28	10.96	0.0	2.13	2018-12-30
13513	-13.80	0.19	0.0	1.04	2018-12-31

13514 rows × 5 columns

We will run a 37 season simulation starting at 1982-05-01 and ending on 2018-10-30

In [3]:

sim_start = '1982/05/01'
sim_end = '2018/10/30'

sim_start = '1982/05/01' sim_end = '2018/10/30'

Next we must define a soil, crop and initial soil water content. This is done by creating a Soil, Crop and InitialWaterContent. In this example we select a sandy loam soil, a Maize crop, and with the soil initially at Field Capacity.

In [4]:

soil= Soil('SandyLoam')

crop = Crop('Maize',planting_date='05/01')

initWC = InitialWaterContent(value=['FC'])

soil= Soil('SandyLoam') crop = Crop('Maize',planting_date='05/01') initWC = InitialWaterContent(value=['FC'])

Irrigation management parameters are selected by creating an IrrigationManagement object. With this class we can specify a range of different irrigation management strategies. The 6 different strategies can be selected using the IrrMethod argument when creating the class. These strategies are as follows:

• IrrMethod=0: Rainfed (no irrigation)
• IrrMethod=1: Irrigation is triggered if soil water content drops below a specified threshold (or four thresholds representing four major crop growth stages (emergence, canopy growth, max canopy, senescence).
• IrrMethod=2: Irrigation is triggered every N days
• IrrMethod=3: Predefined irrigation schedule
• IrrMethod=4: Net irrigation (maintain a soil-water level by topping up all compartments daily)
• IrrMethod=5: Constant depth applied each day

The full list of parameters you can edit are:

Variable Name | Type | Description | Default --- | --- | --- | --- IrrMethod| int | Irrigation method: | 0 || 0 : rainfed | || 1 : soil moisture targets || 2 : set time interval | || 3: predefined schedule | || 4: net irrigation | || 5: constant depth | SMT | list[float] | Soil moisture targets (%TAW) to maintain in each growth stage | [100,100,100,100] || (only used if irrigation method is equal to 1) | IrrInterval | int | Irrigation interval in days | 3 || (only used if irrigation method is equal to 2) | Schedule | pandas.DataFrame | DataFrame containing dates and depths | None || (only used if irrigation method is equal to 3) | NetIrrSMT | float | Net irrigation threshold moisture level (% of TAW that will be maintained) | 80. || (only used if irrigation method is equal to 4) | depth | float | constant depth (mm) to apply on each day | 0. || (only used if irrigation method is equal to 5) | WetSurf | int | Soil surface wetted by irrigation (%) | 100 AppEff | int | Irrigation application efficiency (%) | 100 MaxIrr | float | Maximum depth (mm) that can be applied each day | 25 MaxIrrSeason | float | Maximum total irrigation (mm) that can be applied in one season | 10_000

For the purposes of this demonstration we will investigate the yields and irrigation applied for a range of constant soil-moisture thresholds. Meaning that all 4 soil-moisture thresholds are equal. These irrigation strategies will be compared over a 37 year period. The cell below will create and run an AquaCropModel for each irrigation strategy and save the final output.

In [5]:

# define labels to help after
labels=[]

outputs=[]
for smt in range(0,110,20):
    crop.Name = str(smt) # add helpfull label
    labels.append(str(smt))
    irr_mngt = IrrigationManagement(irrigation_method=1,SMT=[smt]*4) # specify irrigation management
    model = AquaCropModel(sim_start,
                        sim_end,
                        wdf,
                        soil,
                        crop,
                        initial_water_content=initWC,
                        irrigation_management=irr_mngt) # create model
    model.run_model(till_termination=True) # run model till the end
    outputs.append(model._outputs.final_stats) # save results

# define labels to help after labels=[] outputs=[] for smt in range(0,110,20): crop.Name = str(smt) # add helpfull label labels.append(str(smt)) irr_mngt = IrrigationManagement(irrigation_method=1,SMT=[smt]*4) # specify irrigation management model = AquaCropModel(sim_start, sim_end, wdf, soil, crop, initial_water_content=initWC, irrigation_management=irr_mngt) # create model model.run_model(till_termination=True) # run model till the end outputs.append(model._outputs.final_stats) # save results

Combine results so that they can be easily visualized.

In [6]:

import pandas as pd

dflist=outputs
labels[0]='Rainfed'
outlist=[]
for i in range(len(dflist)):
    temp = pd.DataFrame(dflist[i][['Yield (tonne/ha)',
                                   'Seasonal irrigation (mm)']])
    temp['label']=labels[i]
    outlist.append(temp)

all_outputs = pd.concat(outlist,axis=0)

import pandas as pd dflist=outputs labels[0]='Rainfed' outlist=[] for i in range(len(dflist)): temp = pd.DataFrame(dflist[i][['Yield (tonne/ha)', 'Seasonal irrigation (mm)']]) temp['label']=labels[i] outlist.append(temp) all_outputs = pd.concat(outlist,axis=0)

In [7]:

# combine all results
results=pd.concat(outlist)

# combine all results results=pd.concat(outlist)

Use matplotlib and seaborn to show the range of yields and total irrigation for each strategy over the simulation years.

In [8]:

# import plotting libraries
import matplotlib.pyplot as plt
import seaborn as sns

# create figure consisting of 2 plots
fig,ax=plt.subplots(2,1,figsize=(10,14))

# create two box plots
sns.boxplot(data=results,x='label',y='Yield (tonne/ha)',ax=ax[0])
sns.boxplot(data=results,x='label',y='Seasonal irrigation (mm)',ax=ax[1])

# labels and font sizes
ax[0].tick_params(labelsize=15)
ax[0].set_xlabel('Soil-moisture threshold (%TAW)',fontsize=18)
ax[0].set_ylabel('Yield (t/ha)',fontsize=18)

ax[1].tick_params(labelsize=15)
ax[1].set_xlabel('Soil-moisture threshold (%TAW)',fontsize=18)
ax[1].set_ylabel('Total Irrigation (ha-mm)',fontsize=18)

plt.legend(fontsize=18)

# import plotting libraries import matplotlib.pyplot as plt import seaborn as sns # create figure consisting of 2 plots fig,ax=plt.subplots(2,1,figsize=(10,14)) # create two box plots sns.boxplot(data=results,x='label',y='Yield (tonne/ha)',ax=ax[0]) sns.boxplot(data=results,x='label',y='Seasonal irrigation (mm)',ax=ax[1]) # labels and font sizes ax[0].tick_params(labelsize=15) ax[0].set_xlabel('Soil-moisture threshold (%TAW)',fontsize=18) ax[0].set_ylabel('Yield (t/ha)',fontsize=18) ax[1].tick_params(labelsize=15) ax[1].set_xlabel('Soil-moisture threshold (%TAW)',fontsize=18) ax[1].set_ylabel('Total Irrigation (ha-mm)',fontsize=18) plt.legend(fontsize=18)

No handles with labels found to put in legend.

Out[8]:

<matplotlib.legend.Legend at 0x7f01c952d190>

Appendix A: Other types of irrigation strategy¶

Testing different irrigation strategies is as simple as creating multiple IrrMngtClass objects. The first strategy we will test is rainfed growth (no irrigation).

In [9]:

# define irrigation management
rainfed = IrrigationManagement(irrigation_method=0)

# define irrigation management rainfed = IrrigationManagement(irrigation_method=0)

The second strategy triggers irrigation if the root-zone water content drops below an irrigation threshold. There are 4 thresholds corresponding to four main crop growth stages (emergence, canopy growth, max canopy, canopy senescence). The quantity of water applied is given by min(depletion,MaxIrr) where MaxIrr can be specified when creating an IrrMngtClass.

In [10]:

# irrigate according to 4 different soil-moisture thresholds
threshold4_irrigate = IrrigationManagement(irrigation_method=1,SMT=[40,60,70,30]*4)

# irrigate according to 4 different soil-moisture thresholds threshold4_irrigate = IrrigationManagement(irrigation_method=1,SMT=[40,60,70,30]*4)

The third strategy irrigates every IrrInterval days where the quantity of water applied is given by min(depletion,MaxIrr) where MaxIrr can be specified when creating an IrrMngtClass.

In [11]:

# irrigate every 7 days
interval_7 = IrrigationManagement(irrigation_method=2,IrrInterval=7)

# irrigate every 7 days interval_7 = IrrigationManagement(irrigation_method=2,IrrInterval=7)

The fourth strategy irrigates according to a predefined calendar. This calendar is defined as a pandas DataFrame and this example, we will create a calendar that irrigates on the first Tuesday of each month.

In [12]:

import pandas as pd # import pandas library

all_days = pd.date_range(sim_start,sim_end) # list of all dates in simulation period

new_month=True
dates=[]
# iterate through all simulation days
for date in all_days:
    #check if new month
    if date.is_month_start:
        new_month=True

    if new_month:
        # check if tuesday (dayofweek=1)
        if date.dayofweek==1:
            #save date
            dates.append(date)
            new_month=False

import pandas as pd # import pandas library all_days = pd.date_range(sim_start,sim_end) # list of all dates in simulation period new_month=True dates=[] # iterate through all simulation days for date in all_days: #check if new month if date.is_month_start: new_month=True if new_month: # check if tuesday (dayofweek=1) if date.dayofweek==1: #save date dates.append(date) new_month=False

Now we have a list of all the first Tuesdays of the month, we can create the full schedule.

In [13]:

depths = [25]*len(dates) # depth of irrigation applied
schedule=pd.DataFrame([dates,depths]).T # create pandas DataFrame
schedule.columns=['Date','Depth'] # name columns

schedule

depths = [25]*len(dates) # depth of irrigation applied schedule=pd.DataFrame([dates,depths]).T # create pandas DataFrame schedule.columns=['Date','Depth'] # name columns schedule

Out[13]:

	Date	Depth
0	1982-05-04	25
1	1982-06-01	25
2	1982-07-06	25
3	1982-08-03	25
4	1982-09-07	25
...	...	...
433	2018-06-05	25
434	2018-07-03	25
435	2018-08-07	25
436	2018-09-04	25
437	2018-10-02	25

438 rows × 2 columns

Then pass this schedule into our IrrMngtClass.

In [14]:

irrigate_schedule = IrrigationManagement(irrigation_method=3,schedule=schedule)

irrigate_schedule = IrrigationManagement(irrigation_method=3,schedule=schedule)

The fifth strategy is net irrigation. This keeps the soil-moisture content above a specified level. This method differs from the soil moisture thresholds (second strategy) as each compartment is filled to field capacity, instead of water starting above the first compartment and filtering down. In this example the net irrigation mode will maintain a water content of 70% total available water.

In [15]:

net_irrigation = IrrigationManagement(irrigation_method=4,NetIrrSMT=70)

net_irrigation = IrrigationManagement(irrigation_method=4,NetIrrSMT=70)

Now its time to compare the strategies over the 37 year period. The cell below will create and run an AquaCropModel for each irrigation strategy and save the final output.

In [16]:

# define labels to help after
labels=['rainfed','four thresholds','interval','schedule','net']
strategies = [rainfed,threshold4_irrigate,interval_7,irrigate_schedule,net_irrigation]

outputs=[]
for i,irr_mngt in enumerate(strategies): # for both irrigation strategies...
    crop.Name = labels[i] # add helpfull label
    model = AquaCropModel(sim_start,
                        sim_end,
                        wdf,
                        soil,
                        crop,
                        initial_water_content=initWC,
                        irrigation_management=irr_mngt) # create model
    model.run_model(till_termination=True) # run model till the end
    outputs.append(model._outputs.final_stats) # save results

# define labels to help after labels=['rainfed','four thresholds','interval','schedule','net'] strategies = [rainfed,threshold4_irrigate,interval_7,irrigate_schedule,net_irrigation] outputs=[] for i,irr_mngt in enumerate(strategies): # for both irrigation strategies... crop.Name = labels[i] # add helpfull label model = AquaCropModel(sim_start, sim_end, wdf, soil, crop, initial_water_content=initWC, irrigation_management=irr_mngt) # create model model.run_model(till_termination=True) # run model till the end outputs.append(model._outputs.final_stats) # save results

The final strategy to show is for a custom irrigation strategy. This is one of the key features of AquaCrop-OSPy as users can define an a complex irrigation strategy that incorperates any external data, code bases or machine learning models. To showcase this feature, we will define a function that will irrigate according to the follwong logic:

1. There will be no rain over the next 10 days -> Irrigate 10mm

2. There will be rain in the next 10 days but the soil is over 70% depleted -> Irrigate 10mm

3. Otherwise -> No irrigation

In [17]:

# function to return the irrigation depth to apply on next day
def get_depth(model):    
    t = model._clock_struct.time_step_counter # current timestep
    # get weather data for next 7 days
    weather10 = model._weather[t+1:min(t+10+1,len(model._weather))]
    # if it will rain in next 7 days
    if sum(weather10[:,2])>0:
        # check if soil is over 70% depleted
        if t>0 and model._init_cond.depletion/model._init_cond.taw > 0.7:
            depth=10
        else:
            depth=0
    else:
        # no rain for next 10 days
        depth=10


    return depth

# function to return the irrigation depth to apply on next day def get_depth(model): t = model._clock_struct.time_step_counter # current timestep # get weather data for next 7 days weather10 = model._weather[t+1:min(t+10+1,len(model._weather))] # if it will rain in next 7 days if sum(weather10[:,2])>0: # check if soil is over 70% depleted if t>0 and model._init_cond.depletion/model._init_cond.taw > 0.7: depth=10 else: depth=0 else: # no rain for next 10 days depth=10 return depth

In [18]:

model._clock_struct.time_step_counter

model._clock_struct.time_step_counter

Out[18]:

13280

In [19]:

# create model with IrrMethod= Constant depth
crop.Name = 'weather' # add helpfull label

model = AquaCropModel(sim_start,sim_end,wdf,soil,crop,initial_water_content=initWC,
                      irrigation_management=IrrigationManagement(irrigation_method=5,)) 

model._initialize()

while model._clock_struct.model_is_finished is False:    
    # get depth to apply
    depth=get_depth(model)
    
    model._param_struct.IrrMngt.depth=depth

    model.run_model(initialize_model=False)

# create model with IrrMethod= Constant depth crop.Name = 'weather' # add helpfull label model = AquaCropModel(sim_start,sim_end,wdf,soil,crop,initial_water_content=initWC, irrigation_management=IrrigationManagement(irrigation_method=5,)) model._initialize() while model._clock_struct.model_is_finished is False: # get depth to apply depth=get_depth(model) model._param_struct.IrrMngt.depth=depth model.run_model(initialize_model=False)

In [20]:

outputs.append(model._outputs.final_stats) # save results
labels.append('weather')

outputs.append(model._outputs.final_stats) # save results labels.append('weather')

Combine results so that they can be easily visualized.

In [21]:

dflist=outputs
outlist=[]
for i in range(len(dflist)):
    temp = pd.DataFrame(dflist[i][['Yield (tonne/ha)','Seasonal irrigation (mm)']])
    temp['label']=labels[i]
    outlist.append(temp)

all_outputs = pd.concat(outlist,axis=0)

dflist=outputs outlist=[] for i in range(len(dflist)): temp = pd.DataFrame(dflist[i][['Yield (tonne/ha)','Seasonal irrigation (mm)']]) temp['label']=labels[i] outlist.append(temp) all_outputs = pd.concat(outlist,axis=0)

In [22]:

# combine all results
results=pd.concat(outlist)

# combine all results results=pd.concat(outlist)

Use matplotlib and seaborn to show the range of yields and total irrigation for each strategy over the simulation years.

In [23]:

# import plotting libraries
import matplotlib.pyplot as plt
import seaborn as sns

# create figure consisting of 2 plots
fig,ax=plt.subplots(2,1,figsize=(10,14))

# create two box plots
sns.boxplot(data=results,x='label',y='Yield (tonne/ha)',ax=ax[0])
sns.boxplot(data=results,x='label',y='Seasonal irrigation (mm)',ax=ax[1])

# labels and font sizes
ax[0].tick_params(labelsize=15)
ax[0].set_xlabel(' ')
ax[0].set_ylabel('Yield (t/ha)',fontsize=18)

ax[1].tick_params(labelsize=15)
ax[1].set_xlabel(' ')
ax[1].set_ylabel('Total Irrigation (ha-mm)',fontsize=18)

plt.legend(fontsize=18)

# import plotting libraries import matplotlib.pyplot as plt import seaborn as sns # create figure consisting of 2 plots fig,ax=plt.subplots(2,1,figsize=(10,14)) # create two box plots sns.boxplot(data=results,x='label',y='Yield (tonne/ha)',ax=ax[0]) sns.boxplot(data=results,x='label',y='Seasonal irrigation (mm)',ax=ax[1]) # labels and font sizes ax[0].tick_params(labelsize=15) ax[0].set_xlabel(' ') ax[0].set_ylabel('Yield (t/ha)',fontsize=18) ax[1].tick_params(labelsize=15) ax[1].set_xlabel(' ') ax[1].set_ylabel('Total Irrigation (ha-mm)',fontsize=18) plt.legend(fontsize=18)

No handles with labels found to put in legend.

Out[23]:

<matplotlib.legend.Legend at 0x7f0133040b50>

In [ ]:

In [ ]: