Estimating grain yield using PRFT¶

Here we use Photosynthesis Reduction Factor function with different cardinal temperature combinations with and without VPD stress function for the period from heading to maturity (grain filling period).

Authors:

• Azam Lashkari (CIMMYT-China, azam.lashkari@jic.ac.uk)
• Urs Christoph Schulthess (CIMMYT-China, U.Schulthess@cgiar.org)
• Ernesto Giron Echeverry (Independent Researcher, e.giron.e@gmail.com)

Last updated: September 28, 2023

Load libraries¶

In [1]:

import os, sys, gc
import pandas as pd
#import datetime as dt
pd.set_option('display.max_columns', None)

# Import Temperature Response Functions package
import tfunct
from tfunct.data import load_dataset # Function to load existing dataset

print(tfunct.__version__)

import os, sys, gc import pandas as pd #import datetime as dt pd.set_option('display.max_columns', None) # Import Temperature Response Functions package import tfunct from tfunct.data import load_dataset # Function to load existing dataset print(tfunct.__version__)

tfunct version 1.0.0

Load Phenology, NDVI and Weather example datasets¶

In [2]:

# Load example dataset (Phenology, NDVI and Weather data for each site)
data = load_dataset()
# Display available datasets
print(data.keys()) # ['Pheno', 'NDVI', 'Weather']

print(data['Pheno'].shape, data['NDVI'].shape, data['Weather'].shape)

# Load example dataset (Phenology, NDVI and Weather data for each site) data = load_dataset() # Display available datasets print(data.keys()) # ['Pheno', 'NDVI', 'Weather'] print(data['Pheno'].shape, data['NDVI'].shape, data['Weather'].shape)

dict_keys(['Pheno', 'NDVI', 'Weather'])
(50, 10) (436, 11) (246702, 14)

Display phenology¶

In [3]:

data['Pheno'].head()

data['Pheno'].head()

Out[3]:

	country	location	loc_code	lat	lon	cycle	SowingDateQC	Days_To_Heading	Days_To_Maturity	ObsYield
0	Nepal	Bhairahawa	BHR	27.500	83.450	2019	2018-11-26	89	122	2.96685
1	Pakistan	Faisalabad	FAS	31.398	73.054	2014	2013-11-23	95	139	4.88205
2	Pakistan	Faisalabad	FAS	31.398	73.054	2015	2014-12-03	77	121	3.59100
3	Pakistan	Faisalabad	FAS	31.398	73.054	2016	2015-12-07	89	133	2.86425
4	Pakistan	Faisalabad	FAS	31.398	73.054	2017	2016-11-21	94	138	3.93300

Display NDVI¶

In [4]:

data['NDVI'].head()

data['NDVI'].head()

Out[4]:

	country	location	loc_code	lat	lon	cycle	phenotype_date	Year	Month	Day	NDVI
0	Nepal	Bhairahawa	BHR	27.5	83.45	2019	2019-01-28	2019	1	28	0.73
1	Nepal	Bhairahawa	BHR	27.5	83.45	2019	2019-02-03	2019	2	3	0.73
2	Nepal	Bhairahawa	BHR	27.5	83.45	2019	2019-02-11	2019	2	11	0.75
3	Nepal	Bhairahawa	BHR	27.5	83.45	2019	2019-03-01	2019	3	1	0.62
4	Nepal	Bhairahawa	BHR	27.5	83.45	2019	2019-03-06	2019	3	6	0.57

Display Weather¶

AgERA5 weather dataset

In [5]:

data['Weather'].head()

data['Weather'].head()

Out[5]:

	location	Year	Month	Day	lat	lon	Precipitation [mm/d]	Relative Humidity max [%]	Relative Humidity min [%]	Shortwave Radiation [MJ/m2/d]	TemperatureMax [C]	TemperatureMin [C]	Wind Speed 2m [m/s]	Vapor Pressure Deficit max [kPa]
0	Arizona	1979	1	1	33.06	-111.98	0.00	83.59	36.51	12.98	11.46	0.99	2.09	0.86
1	Arizona	1979	1	2	33.06	-111.98	0.00	47.63	32.73	11.17	10.74	-1.33	2.04	0.87
2	Arizona	1979	1	3	33.06	-111.98	0.00	77.74	38.90	10.13	14.33	1.55	1.16	1.00
3	Arizona	1979	1	4	33.06	-111.98	0.00	87.12	46.08	12.77	16.59	0.88	1.14	1.02
4	Arizona	1979	1	5	33.06	-111.98	3.72	86.54	64.45	9.07	17.12	4.10	1.40	0.69

Create a model¶

The model contains all of the required functions to analyse the data

In [6]:

%%time 

# Load module to create a model
from tfunct.model import Model

# ------------------------
# MODEL CONFIGURATION
# ------------------------
# Define the Path where the output data will be stored
PATH_PRJ = '/Users/ernestogiron/Desktop/TemperatureFunctions/'
RESULTS_PATH = PATH_PRJ + 'results/'

config = {
    "PROJECT_PATH": PATH_PRJ,
    "RESULTS_PATH": RESULTS_PATH, #'./', # Results will be put in the same folder where the model is running
}

# Parameters used by default
parameters = dict(
                RUE = 3,
                DRYMATTER = 0.8,
                FACTOR_TON_HA = 0.01,
                YIELD_FACTOR = 0.8 * 0.01,
                TMIN_PERC_FACTOR = 0.25,
                NDVI_lowerThreshold = 0.16,
                Toptmin = 15,
                Topt = 18,
                Toptmax = 25,
                Tmin = 9,
                Tmax = 34,
                Lvpd = 1,
                Uvpd = 4,
                SFvpd_Lthres = 0.2,
                SFvpd_Uthres = 1,
            )

# create model to estimate grain yield
#model = Model(config, parameters) # Use this if you change any parameter above
model = Model(config)
# Preprocess datasets
model.preprocess_raw_datasets(data)

%%time # Load module to create a model from tfunct.model import Model # ------------------------ # MODEL CONFIGURATION # ------------------------ # Define the Path where the output data will be stored PATH_PRJ = '/Users/ernestogiron/Desktop/TemperatureFunctions/' RESULTS_PATH = PATH_PRJ + 'results/' config = { "PROJECT_PATH": PATH_PRJ, "RESULTS_PATH": RESULTS_PATH, #'./', # Results will be put in the same folder where the model is running } # Parameters used by default parameters = dict( RUE = 3, DRYMATTER = 0.8, FACTOR_TON_HA = 0.01, YIELD_FACTOR = 0.8 * 0.01, TMIN_PERC_FACTOR = 0.25, NDVI_lowerThreshold = 0.16, Toptmin = 15, Topt = 18, Toptmax = 25, Tmin = 9, Tmax = 34, Lvpd = 1, Uvpd = 4, SFvpd_Lthres = 0.2, SFvpd_Uthres = 1, ) # create model to estimate grain yield #model = Model(config, parameters) # Use this if you change any parameter above model = Model(config) # Preprocess datasets model.preprocess_raw_datasets(data)

CPU times: user 356 ms, sys: 30.8 ms, total: 387 ms
Wall time: 391 ms

Prepare locations¶

In [7]:

%%time
# Prepare dataset to run all process in parallel using NDVIA GPU if available
sites = model.prepareData()

%%time # Prepare dataset to run all process in parallel using NDVIA GPU if available sites = model.prepareData()

50it [00:01, 36.35it/s]

CPU times: user 1.33 s, sys: 66 ms, total: 1.4 s
Wall time: 1.4 s

In [8]:

#sites[0].inputWPN
sites[0].attributes

#sites[0].inputWPN sites[0].attributes

Out[8]:

{'country': 'Nepal',
 'location': 'Bhairahawa',
 'loc_code': 'BHR',
 'lat': 27.5,
 'lon': 83.45,
 'cycle': 2019,
 'Days_To_Heading': 89,
 'Days_To_Maturity': 122,
 'ObsYield': 2.96685,
 'Sowing_date': '2018-11-26',
 'Heading_date': '2019-02-23',
 'Maturity_date': '2019-03-28',
 'UID': 1,
 'ndays_tmn_lt9': 1,
 'ndays_tmx_gt34': 0,
 'avg_Tdaymax': 24.578,
 'avg_NDVI': 0.447,
 'avg_iPAR': 0.369}

Photosynthetic Reduction Factor (PRFT)¶

Calculating grain yield using Ritchie's Temperature-based function affecting Photosynthetic Reduction Factor (PRFT)

No stress conditions¶

In [9]:

%%time

PRFT_noStress = model.getYield(sites=sites) #Defaults: tfun='PRFT', sites=None, is_VPDStress=False, expCSV=True
PRFT_noStress.head()

%%time PRFT_noStress = model.getYield(sites=sites) #Defaults: tfun='PRFT', sites=None, is_VPDStress=False, expCSV=True PRFT_noStress.head()

CPU times: user 746 ms, sys: 35.3 ms, total: 781 ms
Wall time: 790 ms

Out[9]:

	UID	country	location	loc_code	cycle	ObsYield	RUE	TminFactor	Topt	SimYield
0	1	Nepal	Bhairahawa	BHR	2019	2.97	3.0	0.25	18.0	2.67
1	2	Pakistan	Faisalabad	FAS	2014	4.88	3.0	0.25	18.0	4.54
2	3	Pakistan	Faisalabad	FAS	2015	3.59	3.0	0.25	18.0	3.22
3	4	Pakistan	Faisalabad	FAS	2016	2.86	3.0	0.25	18.0	3.95
4	5	Pakistan	Faisalabad	FAS	2017	3.93	3.0	0.25	18.0	4.45

Stressed VPD¶

In [10]:

%%time

PRFT_SFvpd = model.getYield(sites=sites, is_VPDStress=True)
PRFT_SFvpd.head()

%%time PRFT_SFvpd = model.getYield(sites=sites, is_VPDStress=True) PRFT_SFvpd.head()

CPU times: user 135 ms, sys: 4.17 ms, total: 139 ms
Wall time: 139 ms

Out[10]:

	UID	country	location	loc_code	cycle	ObsYield	RUE	TminFactor	Topt	Lvpd	Uvpd	SFvpd_Lthres	SFvpd_Uthres	SimYield
0	1	Nepal	Bhairahawa	BHR	2019	2.97	3.0	0.25	18.0	1.0	4.0	0.2	1.0	1.71
1	2	Pakistan	Faisalabad	FAS	2014	4.88	3.0	0.25	18.0	1.0	4.0	0.2	1.0	3.34
2	3	Pakistan	Faisalabad	FAS	2015	3.59	3.0	0.25	18.0	1.0	4.0	0.2	1.0	2.69
3	4	Pakistan	Faisalabad	FAS	2016	2.86	3.0	0.25	18.0	1.0	4.0	0.2	1.0	2.44
4	5	Pakistan	Faisalabad	FAS	2017	3.93	3.0	0.25	18.0	1.0	4.0	0.2	1.0	2.50

Display Grain Yield¶

Create a figure to compare simulated grain yield against observed.

In [11]:

from tfunct.util import figures

dirname=os.path.join(config['RESULTS_PATH'], 'PRFT', 'Figures')
figures.chart_compareResults(df_result=PRFT_noStress, fld1="ObsYield", fld2="SimYield", alpha=.75, s=45, xy_lim=2, hue='loc_code', 
                             loc_leg=2, ncol=2, ha='left', va='top',
                             title='PRFT\nNo streess condition', #xlabel='', ylabel='', 
                             dirname=dirname, fname='PRFT_noStress', 
                             dispScore=True, dispLegend=True, saveFig=True, showFig=True, fmt='jpg')

from tfunct.util import figures dirname=os.path.join(config['RESULTS_PATH'], 'PRFT', 'Figures') figures.chart_compareResults(df_result=PRFT_noStress, fld1="ObsYield", fld2="SimYield", alpha=.75, s=45, xy_lim=2, hue='loc_code', loc_leg=2, ncol=2, ha='left', va='top', title='PRFT\nNo streess condition', #xlabel='', ylabel='', dirname=dirname, fname='PRFT_noStress', dispScore=True, dispLegend=True, saveFig=True, showFig=True, fmt='jpg')

In [12]:

dirname=os.path.join(config['RESULTS_PATH'], 'PRFT', 'Figures')
figures.chart_compareResults(df_result=PRFT_SFvpd, fld1="ObsYield", fld2="SimYield", alpha=.75, s=45, xy_lim=2, hue='loc_code', 
                             loc_leg=2, ncol=2, ha='left', va='top',
                             title='PRFT\nVPD streess condition', #xlabel='', ylabel='', 
                             dirname=dirname, fname='PRFT_SFvpd', 
                             dispScore=True, dispLegend=True, saveFig=True, showFig=True, fmt='jpg')

dirname=os.path.join(config['RESULTS_PATH'], 'PRFT', 'Figures') figures.chart_compareResults(df_result=PRFT_SFvpd, fld1="ObsYield", fld2="SimYield", alpha=.75, s=45, xy_lim=2, hue='loc_code', loc_leg=2, ncol=2, ha='left', va='top', title='PRFT\nVPD streess condition', #xlabel='', ylabel='', dirname=dirname, fname='PRFT_SFvpd', dispScore=True, dispLegend=True, saveFig=True, showFig=True, fmt='jpg')

In [13]:

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT')
figures.plot_corrTempFunct(cmb_noStress=PRFT_noStress, cmb_noStress_filtered=None, 
                           cmb_SFvpd=PRFT_SFvpd, cmb_SFvpd_filtered=None,
                   functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=80, alpha=0.95, xy_lim=1, 
                   fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=True, saveFig=True, showFig=True,
                   path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_1_errorbar', fmt='pdf')

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT') figures.plot_corrTempFunct(cmb_noStress=PRFT_noStress, cmb_noStress_filtered=None, cmb_SFvpd=PRFT_SFvpd, cmb_SFvpd_filtered=None, functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=80, alpha=0.95, xy_lim=1, fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=True, saveFig=True, showFig=True, path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_1_errorbar', fmt='pdf')

In [ ]:

Combinations¶

To establish the optimum temperature response for grain-filling period, you can run several models using a wide range of cardinal temperatures.

The optimum temperature for photosynthesis depends on the choosen temperature function.

PRFT combinations¶

No stress conditions¶

In [14]:

%%time

functype='PRFT'
isVPDStress=False
df_GYield, data_input, cols = model.setup_dataInput_forCombinations(sites) # Setup input data
# Combinations
RUE = [3.0] #[2.8, 2.9, 3.0, 3.1, 3.2]
Topt = [x for x in range(15, 26)]
TminFactor = [0.25] #[0.0, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5]
# No stress conditions
array_params_to_run, array_results = model.getCombinations(functype=functype, cols=cols, RUE=RUE, Topt=Topt, 
                                                     TminFactor=TminFactor, isVPDStress=isVPDStress)

cmb_PRFT_noStress = model.getGYield_forCombinations(functype, df_GYield, data_input, array_params_to_run, 
                                           isVPDStress, array_results, saveFile=True)
cmb_PRFT_noStress

%%time functype='PRFT' isVPDStress=False df_GYield, data_input, cols = model.setup_dataInput_forCombinations(sites) # Setup input data # Combinations RUE = [3.0] #[2.8, 2.9, 3.0, 3.1, 3.2] Topt = [x for x in range(15, 26)] TminFactor = [0.25] #[0.0, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5] # No stress conditions array_params_to_run, array_results = model.getCombinations(functype=functype, cols=cols, RUE=RUE, Topt=Topt, TminFactor=TminFactor, isVPDStress=isVPDStress) cmb_PRFT_noStress = model.getGYield_forCombinations(functype, df_GYield, data_input, array_params_to_run, isVPDStress, array_results, saveFile=True) cmb_PRFT_noStress

CPU times: user 274 ms, sys: 13.9 ms, total: 288 ms
Wall time: 290 ms

Out[14]:

	UID	country	location	loc_code	cycle	ObsYield	RUE	TminFactor	Topt	SimYield
0	1	Nepal	Bhairahawa	BHR	2019	2.97	3.0	0.25	15.0	2.36
1	2	Pakistan	Faisalabad	FAS	2014	4.88	3.0	0.25	15.0	4.16
2	3	Pakistan	Faisalabad	FAS	2015	3.59	3.0	0.25	15.0	3.02
3	4	Pakistan	Faisalabad	FAS	2016	2.86	3.0	0.25	15.0	3.34
4	5	Pakistan	Faisalabad	FAS	2017	3.93	3.0	0.25	15.0	3.89
...	...	...	...	...	...	...	...	...	...	...
545	46	India	Pusa	PUS-LS	2018	1.44	3.0	0.25	25.0	2.28
546	47	India	Pusa	PUS-LS	2019	3.10	3.0	0.25	25.0	4.75
547	48	India	Ludhiana	LDH-NS	2022	6.09	3.0	0.25	25.0	5.07
548	49	India	Ludhiana	LDH-LS	2022	4.77	3.0	0.25	25.0	3.99
549	50	Spain	Spain	SPN	2018	8.35	3.0	0.25	25.0	7.08

550 rows × 10 columns

No stress Metrics¶

In [15]:

%%time

m_PRFT_noStress = model.getCombinations_Metrics(functype, isVPDStress, df_GYield, 
                                                 array_params_to_run, array_results, saveFile=True) #, fmt='parquet')
m_PRFT_noStress

%%time m_PRFT_noStress = model.getCombinations_Metrics(functype, isVPDStress, df_GYield, array_params_to_run, array_results, saveFile=True) #, fmt='parquet') m_PRFT_noStress

100%|██████████████████████████████████████████████████████████████████████████████████████████████| 11/11 [00:30<00:00,  2.79s/it]

CPU times: user 28.3 s, sys: 2.11 s, total: 30.4 s
Wall time: 30.7 s

Out[15]:

	RUE	TminFactor	Topt	MAE	MSE	RMSE	RMSRE	MAPE	pvalue	R2	EF	intercept	slope	Cb	CCC	Accuracy
0	3.0	0.25	15.0	0.752	0.923	0.961	20.672	17.186	4.414394e-15	0.726	0.641	0.4338	0.8093	0.96	0.84	82.81
1	3.0	0.25	16.0	0.692	0.759	0.871	20.055	16.308	1.964853e-15	0.735	0.705	0.5507	0.8355	0.99	0.87	83.69
2	3.0	0.25	17.0	0.685	0.728	0.853	21.322	16.539	1.660386e-15	0.737	0.717	0.6656	0.8561	1.00	0.88	83.46
3	3.0	0.25	18.0	0.736	0.787	0.887	23.654	18.217	2.118803e-15	0.734	0.694	0.7716	0.8723	0.99	0.87	81.78
4	3.0	0.25	19.0	0.788	0.905	0.951	26.438	19.954	3.803797e-15	0.727	0.648	0.8701	0.8844	0.98	0.85	80.05
5	3.0	0.25	20.0	0.853	1.056	1.028	29.297	21.957	9.429942e-15	0.717	0.590	0.9627	0.8917	0.96	0.83	78.04
6	3.0	0.25	21.0	0.912	1.213	1.101	31.889	23.732	2.729854e-14	0.704	0.529	1.0464	0.8948	0.94	0.81	76.27
7	3.0	0.25	22.0	0.958	1.358	1.165	34.180	25.195	8.858484e-14	0.689	0.472	1.1232	0.8934	0.93	0.79	74.81
8	3.0	0.25	23.0	0.995	1.481	1.217	36.113	26.381	3.282640e-13	0.672	0.424	1.1967	0.8870	0.92	0.77	73.62
9	3.0	0.25	24.0	1.018	1.575	1.255	37.669	27.214	1.359359e-12	0.652	0.388	1.2627	0.8761	0.92	0.76	72.79
10	3.0	0.25	25.0	1.032	1.632	1.277	38.734	27.753	5.840609e-12	0.631	0.366	1.3177	0.8613	0.92	0.75	72.25

stressed Vapor pressure deficit (VPD)¶

In [16]:

%%time

functype='PRFT'
isVPDStress=True
df_GYield, data_input, cols = model.setup_dataInput_forCombinations(sites) # Setup input data
# Combinations
RUE = [3.0]
Topt = [x for x in range(15, 26)]
TminFactor = [0.25] #[0.0, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5]
Lvpd = [0.5, 1, 1.5, 2, 2.5, 3, 3.5]
Uvpd = [1, 1.5, 2, 2.5, 3, 3.5, 4]
SFvpd_Lthres = [0.2, 0.4, 0.6, 0.8] 
SFvpd_Uthres = [1]
# No stress conditions
array_params_to_run, array_results = model.getCombinations(functype=functype, cols=cols, RUE=RUE, Topt=Topt, TminFactor=TminFactor,  
                                                           Lvpd=Lvpd, Uvpd=Uvpd, SFvpd_Lthres=SFvpd_Lthres, SFvpd_Uthres=SFvpd_Uthres,
                                                           isVPDStress=isVPDStress)

cmb_PRFT_SFvpd = model.getGYield_forCombinations(functype, df_GYield, data_input, array_params_to_run, 
                                           isVPDStress, array_results, saveFile=True)
cmb_PRFT_SFvpd

%%time functype='PRFT' isVPDStress=True df_GYield, data_input, cols = model.setup_dataInput_forCombinations(sites) # Setup input data # Combinations RUE = [3.0] Topt = [x for x in range(15, 26)] TminFactor = [0.25] #[0.0, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5] Lvpd = [0.5, 1, 1.5, 2, 2.5, 3, 3.5] Uvpd = [1, 1.5, 2, 2.5, 3, 3.5, 4] SFvpd_Lthres = [0.2, 0.4, 0.6, 0.8] SFvpd_Uthres = [1] # No stress conditions array_params_to_run, array_results = model.getCombinations(functype=functype, cols=cols, RUE=RUE, Topt=Topt, TminFactor=TminFactor, Lvpd=Lvpd, Uvpd=Uvpd, SFvpd_Lthres=SFvpd_Lthres, SFvpd_Uthres=SFvpd_Uthres, isVPDStress=isVPDStress) cmb_PRFT_SFvpd = model.getGYield_forCombinations(functype, df_GYield, data_input, array_params_to_run, isVPDStress, array_results, saveFile=True) cmb_PRFT_SFvpd

CPU times: user 952 ms, sys: 33.9 ms, total: 986 ms
Wall time: 995 ms

Out[16]:

	UID	country	location	loc_code	cycle	ObsYield	RUE	TminFactor	Topt	Lvpd	Uvpd	SFvpd_Lthres	SFvpd_Uthres	SimYield
0	1	Nepal	Bhairahawa	BHR	2019	2.97	3.0	0.25	15.0	0.5	1.0	0.2	1.0	0.54
1	2	Pakistan	Faisalabad	FAS	2014	4.88	3.0	0.25	15.0	0.5	1.0	0.2	1.0	0.95
2	3	Pakistan	Faisalabad	FAS	2015	3.59	3.0	0.25	15.0	0.5	1.0	0.2	1.0	1.19
3	4	Pakistan	Faisalabad	FAS	2016	2.86	3.0	0.25	15.0	0.5	1.0	0.2	1.0	0.94
4	5	Pakistan	Faisalabad	FAS	2017	3.93	3.0	0.25	15.0	0.5	1.0	0.2	1.0	0.78
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
61595	46	India	Pusa	PUS-LS	2018	1.44	3.0	0.25	25.0	3.5	4.0	0.8	1.0	1.26
61596	47	India	Pusa	PUS-LS	2019	3.10	3.0	0.25	25.0	3.5	4.0	0.8	1.0	3.44
61597	48	India	Ludhiana	LDH-NS	2022	6.09	3.0	0.25	25.0	3.5	4.0	0.8	1.0	4.76
61598	49	India	Ludhiana	LDH-LS	2022	4.77	3.0	0.25	25.0	3.5	4.0	0.8	1.0	3.31
61599	50	Spain	Spain	SPN	2018	8.35	3.0	0.25	25.0	3.5	4.0	0.8	1.0	6.90

61600 rows × 14 columns

In [17]:

%%time

m_PRFT_SFvpd = model.getCombinations_Metrics(functype, isVPDStress, df_GYield, 
                                                 array_params_to_run, array_results, saveFile=True)
m_PRFT_SFvpd

%%time m_PRFT_SFvpd = model.getCombinations_Metrics(functype, isVPDStress, df_GYield, array_params_to_run, array_results, saveFile=True) m_PRFT_SFvpd

100%|████████████████████████████████████████████████████████████████████████████████████████| 1232/1232 [00:00<00:00, 1372.88it/s]

CPU times: user 744 ms, sys: 1.02 s, total: 1.76 s
Wall time: 910 ms

Out[17]:

	RUE	TminFactor	Topt	Lvpd	Uvpd	SFvpd_Lthres	SFvpd_Uthres	MAE	MSE	RMSE	RMSRE	MAPE	pvalue	R2	EF	intercept	slope	Cb	CCC	Accuracy
0	3.0	0.25	15.0	0.5	1.0	0.2	1.0	3.522	14.317	3.784	77.338	76.465	5.004774e-05	0.293	-4.566	0.1838	0.1904	0.12	0.07	23.53
1	3.0	0.25	15.0	0.5	1.0	0.4	1.0	2.748	8.912	2.985	60.492	59.236	9.871293e-10	0.544	-2.464	0.2445	0.3462	0.22	0.17	40.76
2	3.0	0.25	15.0	0.5	1.0	0.6	1.0	1.976	4.875	2.208	44.225	42.019	1.148104e-12	0.655	-0.895	0.3118	0.5008	0.42	0.35	57.98
3	3.0	0.25	15.0	0.5	1.0	0.8	1.0	1.237	2.213	1.488	29.692	25.959	3.544116e-14	0.701	0.140	0.3757	0.6561	0.72	0.62	74.04
4	3.0	0.25	15.0	0.5	1.5	0.2	1.0	3.142	11.904	3.450	71.356	68.675	4.596118e-04	0.228	-3.627	0.0768	0.2972	0.24	0.12	31.33
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
1227	3.0	0.25	25.0	3.0	4.0	0.8	1.0	0.756	0.927	0.963	26.268	19.119	5.743903e-13	0.665	0.640	0.9409	0.7550	0.99	0.82	80.88
1228	3.0	0.25	25.0	3.5	4.0	0.2	1.0	0.822	1.185	1.088	27.784	20.216	1.095611e-10	0.584	0.539	1.0992	0.7310	1.00	0.78	79.78
1229	3.0	0.25	25.0	3.5	4.0	0.4	1.0	0.759	0.959	0.979	26.979	19.394	2.090076e-12	0.646	0.627	1.0919	0.7566	1.00	0.82	80.61
1230	3.0	0.25	25.0	3.5	4.0	0.6	1.0	0.710	0.863	0.929	27.223	18.864	1.524712e-13	0.682	0.665	1.0818	0.7830	1.00	0.84	81.14
1231	3.0	0.25	25.0	3.5	4.0	0.8	1.0	0.747	0.895	0.946	28.420	19.816	9.480468e-14	0.689	0.652	1.0718	0.8093	0.99	0.84	80.18

1232 rows × 20 columns

Display grain yield comparison with and without VPD stress¶

In [18]:

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT')
figures.plot_corrTempFunct(cmb_noStress=cmb_PRFT_noStress, cmb_noStress_filtered=cmb_PRFT_noStress, 
                   cmb_SFvpd=cmb_PRFT_SFvpd, cmb_SFvpd_filtered=cmb_PRFT_SFvpd,
                   functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=20, alpha=0.65, xy_lim=1, 
                   fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=False, saveFig=True, showFig=True,
                   path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_nofilters', fmt='jpg')

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT') figures.plot_corrTempFunct(cmb_noStress=cmb_PRFT_noStress, cmb_noStress_filtered=cmb_PRFT_noStress, cmb_SFvpd=cmb_PRFT_SFvpd, cmb_SFvpd_filtered=cmb_PRFT_SFvpd, functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=20, alpha=0.65, xy_lim=1, fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=False, saveFig=True, showFig=True, path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_nofilters', fmt='jpg')

In [19]:

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT')
figures.plot_corrTempFunct(cmb_noStress=cmb_PRFT_noStress, cmb_noStress_filtered=cmb_PRFT_noStress, 
                   cmb_SFvpd=cmb_PRFT_SFvpd, cmb_SFvpd_filtered=cmb_PRFT_SFvpd,
                   functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=80, alpha=0.95, xy_lim=1, 
                   fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=True, saveFig=True, showFig=True,
                   path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_nofilters_errorbar', fmt='jpg')

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT') figures.plot_corrTempFunct(cmb_noStress=cmb_PRFT_noStress, cmb_noStress_filtered=cmb_PRFT_noStress, cmb_SFvpd=cmb_PRFT_SFvpd, cmb_SFvpd_filtered=cmb_PRFT_SFvpd, functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=80, alpha=0.95, xy_lim=1, fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=True, saveFig=True, showFig=True, path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_nofilters_errorbar', fmt='jpg')

Select the best results using filters¶

After running several models you can filter results to see which one look better.

The following table is an example of one filter used to evaluate the best models, it only use three (3) evaluation metrics (Cb, CCC, and RMSRE) to select the top ones.

Temperature Function	Cb	CCC	RMSRE
PRFT	0.98	> 0.8
PRFT vpd stress	0.98	> 0.8	< 22

In [20]:

# Load module to filter combinations in an easy way
from tfunct.util import filterSimulations

# Load module to filter combinations in an easy way from tfunct.util import filterSimulations

filter PRFT manually¶

In [21]:

m_PRFT_noStress[(m_PRFT_noStress['Cb']>0.98) & (m_PRFT_noStress['CCC']>0.80) ]

m_PRFT_noStress[(m_PRFT_noStress['Cb']>0.98) & (m_PRFT_noStress['CCC']>0.80) ]

Out[21]:

	RUE	TminFactor	Topt	MAE	MSE	RMSE	RMSRE	MAPE	pvalue	R2	EF	intercept	slope	Cb	CCC	Accuracy
1	3.0	0.25	16.0	0.692	0.759	0.871	20.055	16.308	1.964853e-15	0.735	0.705	0.5507	0.8355	0.99	0.87	83.69
2	3.0	0.25	17.0	0.685	0.728	0.853	21.322	16.539	1.660386e-15	0.737	0.717	0.6656	0.8561	1.00	0.88	83.46
3	3.0	0.25	18.0	0.736	0.787	0.887	23.654	18.217	2.118803e-15	0.734	0.694	0.7716	0.8723	0.99	0.87	81.78

In [22]:

m_PRFT_SFvpd[(m_PRFT_SFvpd['Cb']>0.98) & (m_PRFT_SFvpd['CCC']>0.80) & (m_PRFT_SFvpd['RMSRE']<22)]

m_PRFT_SFvpd[(m_PRFT_SFvpd['Cb']>0.98) & (m_PRFT_SFvpd['CCC']>0.80) & (m_PRFT_SFvpd['RMSRE']<22)]

Out[22]:

	RUE	TminFactor	Topt	Lvpd	Uvpd	SFvpd_Lthres	SFvpd_Uthres	MAE	MSE	RMSE	RMSRE	MAPE	pvalue	R2	EF	intercept	slope	Cb	CCC	Accuracy
551	3.0	0.25	19.0	3.0	3.5	0.8	1.0	0.689	0.755	0.869	21.910	16.914	5.709861e-15	0.723	0.706	0.7573	0.8061	1.00	0.86	83.09
651	3.0	0.25	20.0	2.5	3.0	0.8	1.0	0.674	0.758	0.871	21.581	16.520	3.880880e-15	0.727	0.705	0.7142	0.8024	0.99	0.86	83.48
859	3.0	0.25	22.0	2.0	2.5	0.8	1.0	0.646	0.732	0.856	20.753	15.672	1.853433e-15	0.735	0.715	0.7631	0.7889	0.99	0.86	84.33
971	3.0	0.25	23.0	2.0	2.5	0.8	1.0	0.661	0.740	0.860	21.571	16.249	3.514474e-15	0.728	0.712	0.8163	0.7835	0.99	0.86	83.75

No stress conditions¶

In [23]:

%%time
# path of the combinations and metrics files
c_PRFT = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_noStress", "combinations_Yield_PRFT_noStress.parquet")
m_PRFT = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_noStress", "metrics_Yield_PRFT_noStress.parquet")
# Filters
cmb_filters = dict(
                RUE = ('RUE', '=', 3.0 ),
                TminFactor = ('TminFactor', '=', 0.25 )
            )
met_filters = dict(
        Cb = ('>', 0.98),
        CCC = ('>', 0.80)
    )
outputPath = os.path.join(config['RESULTS_PATH'], "PRFT")

# Use pdFormat=True to display filtered table. Use when the results are small to avoid freeze you computer session
# Use fmt='parquet' and avoid fmt='csv' when the results are quite large.
# Use dispFig=True if the filtered table is small.
filtered_c_PRFT = filterSimulations(functype='PRFT', VPDstress=False, cmb=c_PRFT, met=m_PRFT, 
                                 cmb_filters=cmb_filters, met_filters=met_filters, pdFormat=True, 
                                 dispFig=True, saveFig=True, figname='Fig_topComb_avgYield', figfmt='pdf',
                                 saveResults=True, outputPath=outputPath, fmt='parquet')

filtered_c_PRFT

%%time # path of the combinations and metrics files c_PRFT = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_noStress", "combinations_Yield_PRFT_noStress.parquet") m_PRFT = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_noStress", "metrics_Yield_PRFT_noStress.parquet") # Filters cmb_filters = dict( RUE = ('RUE', '=', 3.0 ), TminFactor = ('TminFactor', '=', 0.25 ) ) met_filters = dict( Cb = ('>', 0.98), CCC = ('>', 0.80) ) outputPath = os.path.join(config['RESULTS_PATH'], "PRFT") # Use pdFormat=True to display filtered table. Use when the results are small to avoid freeze you computer session # Use fmt='parquet' and avoid fmt='csv' when the results are quite large. # Use dispFig=True if the filtered table is small. filtered_c_PRFT = filterSimulations(functype='PRFT', VPDstress=False, cmb=c_PRFT, met=m_PRFT, cmb_filters=cmb_filters, met_filters=met_filters, pdFormat=True, dispFig=True, saveFig=True, figname='Fig_topComb_avgYield', figfmt='pdf', saveResults=True, outputPath=outputPath, fmt='parquet') filtered_c_PRFT

Loading combinations...
Loading metrics...
11 combinations found
3 combinations found after applied filter
Filtered data set saved at /Users/ernestogiron/Desktop/TemperatureFunctions/results/PRFT/PRFT_filtered/filtered_metrics_combinations_Yield_PRFT_noStress_20231001.parquet
CPU times: user 1.17 s, sys: 130 ms, total: 1.3 s
Wall time: 677 ms

Out[23]:

	UID	country	location	loc_code	cycle	ObsYield	RUE	TminFactor	Topt	SimYield	MAE	MSE	RMSE	RMSRE	MAPE	pvalue	R2	EF	intercept	slope	Cb	CCC	Accuracy
0	1	Nepal	Bhairahawa	BHR	2019	2.97	3.0	0.25	16.0	2.48	0.692	0.759	0.871	20.055	16.308	1.964853e-15	0.735	0.705	0.5507	0.8355	0.99	0.87	83.69
1	2	Pakistan	Faisalabad	FAS	2014	4.88	3.0	0.25	16.0	4.31	0.692	0.759	0.871	20.055	16.308	1.964853e-15	0.735	0.705	0.5507	0.8355	0.99	0.87	83.69
2	3	Pakistan	Faisalabad	FAS	2015	3.59	3.0	0.25	16.0	3.10	0.692	0.759	0.871	20.055	16.308	1.964853e-15	0.735	0.705	0.5507	0.8355	0.99	0.87	83.69
3	4	Pakistan	Faisalabad	FAS	2016	2.86	3.0	0.25	16.0	3.57	0.692	0.759	0.871	20.055	16.308	1.964853e-15	0.735	0.705	0.5507	0.8355	0.99	0.87	83.69
4	5	Pakistan	Faisalabad	FAS	2017	3.93	3.0	0.25	16.0	4.10	0.692	0.759	0.871	20.055	16.308	1.964853e-15	0.735	0.705	0.5507	0.8355	0.99	0.87	83.69
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
145	46	India	Pusa	PUS-LS	2018	1.44	3.0	0.25	18.0	1.64	0.736	0.787	0.887	23.654	18.217	2.118803e-15	0.734	0.694	0.7716	0.8723	0.99	0.87	81.78
146	47	India	Pusa	PUS-LS	2019	3.10	3.0	0.25	18.0	3.45	0.736	0.787	0.887	23.654	18.217	2.118803e-15	0.734	0.694	0.7716	0.8723	0.99	0.87	81.78
147	48	India	Ludhiana	LDH-NS	2022	6.09	3.0	0.25	18.0	4.48	0.736	0.787	0.887	23.654	18.217	2.118803e-15	0.734	0.694	0.7716	0.8723	0.99	0.87	81.78
148	49	India	Ludhiana	LDH-LS	2022	4.77	3.0	0.25	18.0	2.90	0.736	0.787	0.887	23.654	18.217	2.118803e-15	0.734	0.694	0.7716	0.8723	0.99	0.87	81.78
149	50	Spain	Spain	SPN	2018	8.35	3.0	0.25	18.0	7.32	0.736	0.787	0.887	23.654	18.217	2.118803e-15	0.734	0.694	0.7716	0.8723	0.99	0.87	81.78

150 rows × 23 columns

VPD stress conditions¶

In [24]:

%%time

# Path of the combinations and metrics files
c_PRFT_SFvpd = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_SFvpd", "combinations_Yield_PRFT_SFvpd.parquet")
m_PRFT_SFvpd = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_SFvpd", "metrics_Yield_PRFT_SFvpd.parquet")
cmb_filters = dict(
                RUE = ('RUE', '=', 3.0 ),
                TminFactor = ('TminFactor', '=', 0.25 )
            )
met_filters = dict(
        Cb = ('>', 0.98),
        CCC = ('>', 0.80)
    )
outputPath = os.path.join(config['RESULTS_PATH'], "PRFT") 

# Use pdFormat=True to display filtered table. Use when the results are small to avoid freeze you computer session
# Use fmt='parquet' and avoid fmt='csv' when the results are quite large.
# Use dispFig=True if the filtered table is small.
filtered_c_PRFT_SFvpd = filterSimulations(functype='PRFT', VPDstress=True, cmb=c_PRFT_SFvpd, met=m_PRFT_SFvpd, 
                                 cmb_filters=cmb_filters, met_filters=met_filters, pdFormat=True, 
                                 dispFig=True, saveFig=True, figname='Fig_topComb_avgYield', figfmt='pdf',
                                 saveResults=True, outputPath=outputPath, fmt='parquet')

filtered_c_PRFT_SFvpd

%%time # Path of the combinations and metrics files c_PRFT_SFvpd = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_SFvpd", "combinations_Yield_PRFT_SFvpd.parquet") m_PRFT_SFvpd = os.path.join(config['RESULTS_PATH'], "PRFT", "PRFT_SFvpd", "metrics_Yield_PRFT_SFvpd.parquet") cmb_filters = dict( RUE = ('RUE', '=', 3.0 ), TminFactor = ('TminFactor', '=', 0.25 ) ) met_filters = dict( Cb = ('>', 0.98), CCC = ('>', 0.80) ) outputPath = os.path.join(config['RESULTS_PATH'], "PRFT") # Use pdFormat=True to display filtered table. Use when the results are small to avoid freeze you computer session # Use fmt='parquet' and avoid fmt='csv' when the results are quite large. # Use dispFig=True if the filtered table is small. filtered_c_PRFT_SFvpd = filterSimulations(functype='PRFT', VPDstress=True, cmb=c_PRFT_SFvpd, met=m_PRFT_SFvpd, cmb_filters=cmb_filters, met_filters=met_filters, pdFormat=True, dispFig=True, saveFig=True, figname='Fig_topComb_avgYield', figfmt='pdf', saveResults=True, outputPath=outputPath, fmt='parquet') filtered_c_PRFT_SFvpd

Loading combinations...
Loading metrics...
1232 combinations found
51 combinations found after applied filter
Filtered data set saved at /Users/ernestogiron/Desktop/TemperatureFunctions/results/PRFT/PRFT_filtered/filtered_metrics_combinations_Yield_PRFT_SFvpd_20231001.parquet
CPU times: user 1.25 s, sys: 151 ms, total: 1.4 s
Wall time: 784 ms

Out[24]:

	UID	country	location	loc_code	cycle	ObsYield	RUE	TminFactor	Topt	Lvpd	Uvpd	SFvpd_Lthres	SFvpd_Uthres	SimYield	MAE	MSE	RMSE	RMSRE	MAPE	pvalue	R2	EF	intercept	slope	Cb	CCC	Accuracy
0	1	Nepal	Bhairahawa	BHR	2019	2.97	3.0	0.25	18.0	3.5	4.0	0.6	1.0	2.47	0.778	0.954	0.977	23.351	18.737	2.337009e-13	0.677	0.629	0.6547	0.8021	0.99	0.83	81.26
1	2	Pakistan	Faisalabad	FAS	2014	4.88	3.0	0.25	18.0	3.5	4.0	0.6	1.0	4.48	0.778	0.954	0.977	23.351	18.737	2.337009e-13	0.677	0.629	0.6547	0.8021	0.99	0.83	81.26
2	3	Pakistan	Faisalabad	FAS	2015	3.59	3.0	0.25	18.0	3.5	4.0	0.6	1.0	3.22	0.778	0.954	0.977	23.351	18.737	2.337009e-13	0.677	0.629	0.6547	0.8021	0.99	0.83	81.26
3	4	Pakistan	Faisalabad	FAS	2016	2.86	3.0	0.25	18.0	3.5	4.0	0.6	1.0	3.58	0.778	0.954	0.977	23.351	18.737	2.337009e-13	0.677	0.629	0.6547	0.8021	0.99	0.83	81.26
4	5	Pakistan	Faisalabad	FAS	2017	3.93	3.0	0.25	18.0	3.5	4.0	0.6	1.0	4.01	0.778	0.954	0.977	23.351	18.737	2.337009e-13	0.677	0.629	0.6547	0.8021	0.99	0.83	81.26
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2545	46	India	Pusa	PUS-LS	2018	1.44	3.0	0.25	25.0	3.5	4.0	0.8	1.0	1.26	0.747	0.895	0.946	28.420	19.816	9.480468e-14	0.689	0.652	1.0718	0.8093	0.99	0.84	80.18
2546	47	India	Pusa	PUS-LS	2019	3.10	3.0	0.25	25.0	3.5	4.0	0.8	1.0	3.44	0.747	0.895	0.946	28.420	19.816	9.480468e-14	0.689	0.652	1.0718	0.8093	0.99	0.84	80.18
2547	48	India	Ludhiana	LDH-NS	2022	6.09	3.0	0.25	25.0	3.5	4.0	0.8	1.0	4.76	0.747	0.895	0.946	28.420	19.816	9.480468e-14	0.689	0.652	1.0718	0.8093	0.99	0.84	80.18
2548	49	India	Ludhiana	LDH-LS	2022	4.77	3.0	0.25	25.0	3.5	4.0	0.8	1.0	3.31	0.747	0.895	0.946	28.420	19.816	9.480468e-14	0.689	0.652	1.0718	0.8093	0.99	0.84	80.18
2549	50	Spain	Spain	SPN	2018	8.35	3.0	0.25	25.0	3.5	4.0	0.8	1.0	6.90	0.747	0.895	0.946	28.420	19.816	9.480468e-14	0.689	0.652	1.0718	0.8093	0.99	0.84	80.18

2550 rows × 27 columns

In [25]:

path_to_save_results=os.path.join(config['RESULTS_PATH'],'PRFT', 'PRFT_filtered')
figures.plot_corrTempFunct(cmb_noStress=filtered_c_PRFT, cmb_noStress_filtered=filtered_c_PRFT, 
                           cmb_SFvpd=filtered_c_PRFT_SFvpd, cmb_SFvpd_filtered=filtered_c_PRFT_SFvpd,
                   functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=40, alpha=0.95, xy_lim=1, 
                   fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=False, saveFig=True, showFig=True,
                   path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_filtered', fmt='jpg')

path_to_save_results=os.path.join(config['RESULTS_PATH'],'PRFT', 'PRFT_filtered') figures.plot_corrTempFunct(cmb_noStress=filtered_c_PRFT, cmb_noStress_filtered=filtered_c_PRFT, cmb_SFvpd=filtered_c_PRFT_SFvpd, cmb_SFvpd_filtered=filtered_c_PRFT_SFvpd, functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=40, alpha=0.95, xy_lim=1, fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=False, saveFig=True, showFig=True, path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_filtered', fmt='jpg')

In [26]:

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT', 'PRFT_filtered')
figures.plot_corrTempFunct(cmb_noStress=filtered_c_PRFT, cmb_noStress_filtered=filtered_c_PRFT, 
                           cmb_SFvpd=filtered_c_PRFT_SFvpd, cmb_SFvpd_filtered=filtered_c_PRFT_SFvpd,
                   functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=80, alpha=0.95, xy_lim=1, 
                   fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=True, saveFig=True, showFig=True,
                   path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_filtered_errorbar', fmt='jpg')

path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT', 'PRFT_filtered') figures.plot_corrTempFunct(cmb_noStress=filtered_c_PRFT, cmb_noStress_filtered=filtered_c_PRFT, cmb_SFvpd=filtered_c_PRFT_SFvpd, cmb_SFvpd_filtered=filtered_c_PRFT_SFvpd, functype='PRFT',fld1='ObsYield',fld2='SimYield',hue='location', ncol=6, s=80, alpha=0.95, xy_lim=1, fonts_axes=10, fonts_titles=12, dispScore=True, errorbar=True, saveFig=True, showFig=True, path_to_save_results=path_to_save_results, dirname='Figures', fname='Fig_2_filtered_errorbar', fmt='jpg')

Comparison of the best combinations¶

No Stress condition¶

In [27]:

# Prepare data as a input for create comparison figures by site and/or best combinations
path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT', 'PRFT_filtered')
df_TDays_vs_TempResponse, df_TDays_vs_TempResponse_mean_allSites = \
figures.createFigure_Type_I_PRFT(sites, cmb=filtered_c_PRFT, roundVal=3, maxTDay=40, saveTable=True,
                        path_to_save_results=path_to_save_results, fmt='parquet')
df_TDays_vs_TempResponse.head()

# Prepare data as a input for create comparison figures by site and/or best combinations path_to_save_results=os.path.join(config['RESULTS_PATH'], 'PRFT', 'PRFT_filtered') df_TDays_vs_TempResponse, df_TDays_vs_TempResponse_mean_allSites = \ figures.createFigure_Type_I_PRFT(sites, cmb=filtered_c_PRFT, roundVal=3, maxTDay=40, saveTable=True, path_to_save_results=path_to_save_results, fmt='parquet') df_TDays_vs_TempResponse.head()

100%|██████████████████████████████████████████████████████████████████████████████████████████████| 50/50 [00:01<00:00, 40.26it/s]

Calculating average value for all simulations...

100%|████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:01<00:00,  2.88it/s]

Out[27]:

	TDay	TempResponse	UID	country	location	loc_code	cycle	RUE	TminFactor	Topt	ObsYield	SimYield
0	0.0	0.360	1	Nepal	Bhairahawa	BHR	2019	3.0	0.25	16.0	2.97	2.576667
1	0.1	0.368	1	Nepal	Bhairahawa	BHR	2019	3.0	0.25	16.0	2.97	2.576667
2	0.2	0.376	1	Nepal	Bhairahawa	BHR	2019	3.0	0.25	16.0	2.97	2.576667
3	0.3	0.384	1	Nepal	Bhairahawa	BHR	2019	3.0	0.25	16.0	2.97	2.576667
4	0.4	0.392	1	Nepal	Bhairahawa	BHR	2019	3.0	0.25	16.0	2.97	2.576667

In [28]:

#cmb = [cmb_PRFT_noStress['UID']==uid].reset_index(drop=True)
figures.diplay_Figure_Type_I_PRFT_InOneFig(cmb=filtered_c_PRFT, fnct='PRFT', 
                                           df_tdays=df_TDays_vs_TempResponse, 
                                           df_tdays_mean=df_TDays_vs_TempResponse_mean_allSites,
                                           saveFig=True,showFig=True, fmt='jpg', 
                                           path_to_save_results=path_to_save_results)

#cmb = [cmb_PRFT_noStress['UID']==uid].reset_index(drop=True) figures.diplay_Figure_Type_I_PRFT_InOneFig(cmb=filtered_c_PRFT, fnct='PRFT', df_tdays=df_TDays_vs_TempResponse, df_tdays_mean=df_TDays_vs_TempResponse_mean_allSites, saveFig=True,showFig=True, fmt='jpg', path_to_save_results=path_to_save_results)

100%|██████████████████████████████████████████████████████████████████████████████████████████████| 50/50 [00:24<00:00,  2.01it/s]
100%|████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00,  6.12it/s]

In [29]:

%%time
#cmb = [cmb_PRFT_noStress['UID']==uid].reset_index(drop=True)
figures.display_FigTDay_vs_TempResponseBySiteYrs(cmb=filtered_c_PRFT, df_tdays=df_TDays_vs_TempResponse, 
                                         roundVal=2, maxTDay=40, fnct='PRFT', 
                                         saveFig=True, showFig=True, fmt='jpg', cols=5, 
                                         path_to_save_results=path_to_save_results)

%%time #cmb = [cmb_PRFT_noStress['UID']==uid].reset_index(drop=True) figures.display_FigTDay_vs_TempResponseBySiteYrs(cmb=filtered_c_PRFT, df_tdays=df_TDays_vs_TempResponse, roundVal=2, maxTDay=40, fnct='PRFT', saveFig=True, showFig=True, fmt='jpg', cols=5, path_to_save_results=path_to_save_results)

100%|██████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:43<00:00,  4.38s/it]

CPU times: user 51.3 s, sys: 1.25 s, total: 52.5 s
Wall time: 52.8 s

In [30]:

uid = 1
df0 = filtered_c_PRFT[filtered_c_PRFT['UID']==uid].reset_index(drop=True)
figures.diplay_Figure_Type_I_PRFT_InOneFig(cmb=df0, fnct='PRFT', 
                                           df_tdays=df_TDays_vs_TempResponse, 
                                           df_tdays_mean=df_TDays_vs_TempResponse_mean_allSites,
                                           saveFig=True,showFig=True, fmt='jpg', 
                                           path_to_save_results=path_to_save_results) #fname = "Figure_PRFT_SiteYrs_Comparison"

uid = 1 df0 = filtered_c_PRFT[filtered_c_PRFT['UID']==uid].reset_index(drop=True) figures.diplay_Figure_Type_I_PRFT_InOneFig(cmb=df0, fnct='PRFT', df_tdays=df_TDays_vs_TempResponse, df_tdays_mean=df_TDays_vs_TempResponse_mean_allSites, saveFig=True,showFig=True, fmt='jpg', path_to_save_results=path_to_save_results) #fname = "Figure_PRFT_SiteYrs_Comparison"

100%|████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00,  4.62it/s]

Conclusion¶

Congratulations! You have run a simulation using a prebuilt dataset and the Temperature Functions API.

Type 1- one cardinal temperature¶

Model Parameters	No stress	VPD stress
RUE	3.0	3.0
TminFactor	0.25	0.25
Topt	16-18 °C	18-25 °C
VPDL		2-3.5 kPa
VPDU		2.5-4.0 kPa
SFvpdL		0.4-08
SFvpdU		1

In [31]:

print(filtered_c_PRFT['Topt'].unique())
print(filtered_c_PRFT_SFvpd['Topt'].sort_values().unique())
print(filtered_c_PRFT_SFvpd['Lvpd'].sort_values().unique())
print(filtered_c_PRFT_SFvpd['Uvpd'].sort_values().unique())
print(filtered_c_PRFT_SFvpd['SFvpd_Lthres'].sort_values().unique())
print(filtered_c_PRFT_SFvpd['SFvpd_Uthres'].sort_values().unique())

print(filtered_c_PRFT['Topt'].unique()) print(filtered_c_PRFT_SFvpd['Topt'].sort_values().unique()) print(filtered_c_PRFT_SFvpd['Lvpd'].sort_values().unique()) print(filtered_c_PRFT_SFvpd['Uvpd'].sort_values().unique()) print(filtered_c_PRFT_SFvpd['SFvpd_Lthres'].sort_values().unique()) print(filtered_c_PRFT_SFvpd['SFvpd_Uthres'].sort_values().unique())

[16. 17. 18.]
[18. 19. 20. 21. 22. 23. 24. 25.]
[2.  2.5 3.  3.5]
[2.5 3.  3.5 4. ]
[0.4 0.6 0.8]
[1.]

In [ ]: