Module 3: estimating cultivar coefficients (GLUE), soil data and water-limited production

This notebook follows the third day of a DSSAT training course:

estimate the coefficients of a soybean cultivar from field data with GLUE, ported to Python in docs/notebooks/glue.py from DSSAT’s GLUE tool;
read and estimate soil water limits;
switch the water balance on and look at water-limited production: soil water in each layer, the soil’s capacity to store water, and automatic irrigation.

Data. Soybean experiments IUAM8801, IUAM9001, OHWO8801, OHWO9001 and UFGA7902 (Williams-82; the course’s GLUE exercise) and IUAM9701, IUAM9702 for an independent check; peanut UFGA8401; maize UFGA8201 and the soil file, all from DSSAT/dssat-csm-data (BSD-3) at a pinned commit. From the course data (notebooks/course-data/): the Colorado maize experiments of GLUE self test 4 and the Umariya and Orangeburg soil sheets. GLUE follows DSSAT/glue (BSD-3) at commit e86093e.

Needs. As in modules 1 and 2. The GLUE part runs about 36,000 soybean seasons; on a many-core machine that takes a few minutes (it runs up to 99 parameter sets per experiment file and many files at once).

engine   : C:\TEMP\dsn\dssat\dscsm048.exe
csm-data : C:\Users\benny\OneDrive\Documents\Github\dssat-csm-data @ 99c95b34
GLUE     : https://github.com/DSSAT/glue/blob/e86093e400fbbdfc8e2c751c98acc2136835c3af/Calculation.r
workers  : 48

1. Estimating cultivar coefficients with GLUE### The observationsThe five treatments the course’s GLUE exercise selects for Williams-82 soybean (a maturity group 3cultivar) are used for calibration: treatment 1 of IUAM8801, IUAM9001, OHWO8801 and OHWO9001, and the fullyirrigated treatment 4 of UFGA7902. Two more(IUAM9701 and IUAM9702) are kept back to test the result: only their yield was measured.

	exp	ADAP	MDAP	HWAM	CWAM	PWAM	LAIX	HWUM
0	IUAM8801	51.0	129	2368.0	6117.0	3604.0	5.57	0.1270
1	IUAM9001	66.0	139	3168.0	6674.0	4552.0	6.29	0.1682
2	OHWO8801	71.0	145	3976.0	8090.0	5194.0	7.65	0.1510
3	OHWO9001	NaN	150	3149.0	7113.0	4915.0	7.21	0.1640
4	UFGA7902	38.0	93	2474.0	3893.0	NaN	3.15	0.1750

Priors: what may each coefficient be?GLUE starts from a range for every coefficient. DSSAT keeps them in the cultivar file itself: a `MINIMA` anda `MAXIMA` row, and a `!Calibration` row that flags each coefficient as P (phenology), G (growth) orN (not calibrated). GLUE draws parameter sets uniformly within the ranges(`Uniform.r`).

	coef	flag	min	max	default	outside_range
0	CSDL	P	11.780	14.600	13.400	False
1	PPSEN	P	0.129	0.385	0.285	False
2	EM-FL	P	9.000	28.900	19.000	False
3	FL-SH	N	5.000	10.000	8.300	False
4	FL-SD	P	11.000	22.000	14.200	False
5	SD-PM	P	22.000	37.700	32.200	False
6	FL-LF	N	18.000	26.000	28.000	True
7	LFMAX	G	1.000	1.400	0.970	True
8	SLAVR	G	300.000	400.000	385.000	False
9	SIZLF	G	137.000	230.000	180.000	False
10	XFRT	N	1.000	1.000	1.000	False
11	WTPSD	G	0.150	0.190	0.180	False
12	SFDUR	G	17.000	25.500	26.000	True
13	SDPDV	G	1.700	2.440	2.400	False
14	PODUR	N	10.000	10.000	12.000	True
15	THRSH	N	77.000	78.000	77.000	False
16	SDPRO	N	0.400	0.405	0.405	False
17	SDLIP	N	0.200	0.205	0.205	False

The likelihoodFor each measurement GLUE computes how likely the simulated value is if the measurement has a normal error,`exp(-(M - S)^2 / (2 var)) / sqrt(2 pi var)`(`Calculation.r`).The variances are fixed in`MeasurementVariance_All.csv`:a standard deviation of 3 days for anthesis and 7 days for maturity, and a coefficient of variation forweights (0.3 for yields and tops, 0.1 for seed size, 0.4 for LAI). Likelihoods are multiplied over themeasurements of an experiment, normalised over the parameter sets, multiplied over the experiments andnormalised again. Round 1 fits the phenology coefficients to the dates; round 2 keeps the best round-1values and fits the growth coefficients to the weights.

	measurement	kind	value	round
0	ADAP	sd	3.0	1
1	MDAP	sd	7.0	1
2	PWAM	cv	0.3	2
3	HWAM	cv	0.3	2
4	CWAM	cv	0.3	2
5	LAIX	cv	0.4	2
6	HWUM	cv	0.1	2

Before calibrationWilliams-82 as it is in the engine’s cultivar file:

	set	exp	ADAP	MDAP	HWAM	CWAM	PWAM	LAIX	HWUM
0	1	IUAM8801	58	125	2817	6341	3659	5.711085	0.157695
1	1	IUAM9001	63	133	3700	7450	4998	6.066354	0.149194
2	1	OHWO8801	76	142	3620	7491	4938	6.639524	0.155467
3	1	OHWO9001	75	145	3541	7185	4822	6.609224	0.149739
4	1	UFGA7902	39	97	2772	4422	3922	2.832199	0.145784
5	1	IUAM9701	63	131	3767	8219	5249	7.236816	0.149195
6	1	IUAM9702	65	135	3084	6676	4178	6.818458	0.152660

Round 1: phenology

round 1: 15000 seasons in 264 s; sets that matured everywhere: 2086 of 3000

	coef	best	mean	sd	prior_min	prior_max
0	CSDL	13.249898	13.255926	0.456622	11.781553	14.599494
1	PPSEN	0.267109	0.275741	0.061453	0.129002	0.384951
2	EM-FL	17.393830	17.718176	1.703780	9.000204	28.895782
3	FL-SD	11.895984	17.096515	3.242615	11.001382	21.995934
4	SD-PM	37.552135	31.146929	3.913858	22.002702	37.697495

Round 2: growth

round 2: 15000 seasons in 250 s

	coef	best	mean	sd	prior_min	prior_max
0	LFMAX	1.032143	1.104075	0.085230	1.000186	1.399889
1	SLAVR	396.348431	359.079587	26.647380	300.076718	399.998198
2	SIZLF	175.967169	178.767408	24.894066	137.018047	229.953704
3	WTPSD	0.151384	0.155078	0.004139	0.150003	0.189991
4	SFDUR	20.764567	21.640275	2.154208	17.000121	25.496115
5	SDPDV	2.363773	2.057482	0.208301	1.700413	2.439995

Best values on the edge of their rangeA best value that sits on the edge of its prior range says the data want a value the range does notallow. Here that happens in round 2, and the engine’s own Williams-82 values for `LFMAX` and `SFDUR` already lieoutside the ranges of the cultivar file. Round 2 is run again with the ranges of those coefficients widenedby half their width on the side that was hit:

round 2, wider ranges: 15000 seasons in 271 s

	coef	best	side	old_min	old_max	new_min	new_max
0	SLAVR	396.348431	max	300.00	400.00	300.00	450.00
1	WTPSD	0.151384	min	0.15	0.19	0.13	0.19

	coef	best	mean	sd	prior_min	prior_max
0	LFMAX	1.020757	1.106673	0.089082	1.000264	1.399933
1	SLAVR	436.675873	387.829469	39.393867	300.099944	449.890096
2	SIZLF	174.747192	176.937396	24.665975	137.001935	229.961366
3	WTPSD	0.146365	0.147339	0.007548	0.130023	0.189992
4	SFDUR	21.500243	20.678409	2.180812	17.000519	25.498369
5	SDPDV	1.728097	2.062645	0.210283	1.700155	2.439532

After calibration, and the independent check

version	GLUE	GLUE, wider ranges	default
variable
ADAP	3.50	3.50	4.58
CWAM	573.07	485.88	509.15
HWAM	504.77	494.17	476.82
HWUM	0.01	0.02	0.02
LAIX	0.55	0.47	0.56
MDAP	3.55	3.55	4.52
PWAM	470.52	483.59	262.74

WindowsPath('C:/Users/benny/OneDrive/Documents/Github/xdssat/docs/learn/data/module3/glue_info.csv')

GLUE self test 4: a new maize cultivar from ColoradoThe course’s last GLUE exercise calibrates a cultivar that is not yet in DSSAT: the maize hybrid DEKALB52-59 AF, grown at Greeley, Colorado, in 2008 and 2010 with full irrigation and with 40% of it(Anothai et al. 2013). The experiment files, weather and soil file come with the course(`course-data/colorado/`). The exercise asks for a new cultivar entry `CO0801` built from an existing variety;the notebook uses DEKALB 485 (`IB1072`) of the engine’s cultivar file as that starting point. All fourtreatments are used; their FileA has anthesis, maturity, yield and tops weight.

	exp	ADAP	MDAP	HWAM	CWAM	PWAM	LAIX	HWUM
0	GRCO0801-T1	86	150	13228.0	22112.0	NaN	NaN	NaN
1	GRCO0801-T2	86	150	8824.0	14331.0	NaN	NaN	NaN
2	GRCO1001-T1	81	136	11168.0	18881.0	NaN	NaN	NaN
3	GRCO1001-T2	81	136	5505.0	9294.0	NaN	NaN	NaN

	coef	flag	min	max	default
0	P1	P	5.0	450.0	215.0
1	P2	P	0.0	2.0	0.6
2	P5	P	580.0	999.0	785.0
3	G2	G	248.0	990.0	750.0
4	G3	G	5.0	16.5	8.7
5	PHINT	N	38.0	75.0	45.0

Colorado GLUE: 16000 seasons in 168 s

	coef	best	mean	sd	prior_min	prior_max	round	DEKALB 485
0	P1	247.395667	244.454392	22.123239	5.031595	449.343032	1	215.0
1	P2	1.237487	1.242804	0.503020	0.000625	1.997433	1	0.6
2	P5	608.446720	628.558692	34.777022	580.181616	998.953208	1	785.0
0	G2	573.037012	677.914586	195.344033	248.026413	989.798979	2	750.0
1	G3	11.836032	11.985038	2.768234	5.003465	16.492472	2	8.7

		DEKALB 485 as it is	GLUE: CO0801	observed
exp	variable
GRCO0801-T1	ADAP	75.0	84.0	86.0
	CWAM	26978.0	25165.0	22112.0
	HWAM	16006.0	11103.0	13228.0
	MDAP	151.0	151.0	150.0
GRCO0801-T2	ADAP	75.0	84.0	86.0
	CWAM	13282.0	15932.0	14331.0
	HWAM	7845.0	8518.0	8824.0
	MDAP	151.0	151.0	150.0
GRCO1001-T1	ADAP	75.0	82.0	81.0
	CWAM	24409.0	23865.0	18881.0
	HWAM	12022.0	9650.0	11168.0
	MDAP	138.0	132.0	136.0
GRCO1001-T2	ADAP	75.0	82.0	81.0
	CWAM	12606.0	13213.0	9294.0
	HWAM	5577.0	5281.0	5505.0
	MDAP	138.0	132.0	136.0

WindowsPath('C:/Users/benny/OneDrive/Documents/Github/xdssat/docs/learn/data/module3/co_round2.csv')

Calibration with in-season data: peanut UFGA8401The course’s second calibration exercise works by hand, one coefficient at a time, against growth analysisdata taken during the season: the irrigated treatment of the 1984 Florunner peanut experiment UFGA8401. Itasks for four statistics of four variables (tops weight, seed weight, seed number and pod harvest index): theRMSE, the intercept `a` and slope `b` of the regression of simulated on observed values, and Willmott’s `d`.The suggested coefficients: `LFMAX` (the size of tops and seed weight), `FL-SH` and `FL-SD` moved together(the timing of seed growth), `PODUR` (the first rise of seed number and pod harvest index) and `XFRT` (theshare of growth going to pods and seeds).Instead of one run at a time, the notebook runs a grid of all four, so the page can move each one freely.

                     TNAME  CU  MI
N                                 
1                IRRIGATED   1   1
2  RAINFED, EXCEPT IRRIGAT   1   2
cultivar IB0002 FLORUNNER {'LFMAX': np.float64(1.4), 'FL-SH': np.float64(9.2), 'FL-SD': np.float64(18.8), 'PODUR': np.float64(24.0), 'XFRT': np.float64(0.92)}
observed variables: ['CWAD', 'GWAD', 'G#AD', 'HIPD'] 19 dates

375 combinations; XFRT values [np.float64(0.83), np.float64(0.92), 1.0]
soil IBPN910015
375 peanut seasons with daily output in 31 s

Florunner as it is: set 188 mean d 0.979 | best mean d: set 324 0.988

	set	variable	n	rmse	a	b	d	LFMAX	shift	PODUR	XFRT	FL-SH	FL-SD
748	188	CWAD	19	1285.118	-332.731	0.893	0.976	1.40	0.0	24.0	0.92	9.2	18.8
749	188	G#AD	19	112.553	-0.454	0.837	0.972	1.40	0.0	24.0	0.92	9.2	18.8
750	188	GWAD	19	590.725	-96.562	0.849	0.974	1.40	0.0	24.0	0.92	9.2	18.8
751	188	HIPD	19	0.034	-0.012	0.987	0.993	1.40	0.0	24.0	0.92	9.2	18.8
1292	324	CWAD	19	1130.797	-254.886	0.906	0.982	1.61	-2.0	24.0	1.00	7.2	16.8
1293	324	G#AD	19	80.448	8.530	0.918	0.987	1.61	-2.0	24.0	1.00	7.2	16.8
1294	324	GWAD	19	402.287	-73.019	0.933	0.989	1.61	-2.0	24.0	1.00	7.2	16.8
1295	324	HIPD	19	0.033	-0.007	1.060	0.994	1.61	-2.0	24.0	1.00	7.2	16.8

WindowsPath('C:/Users/benny/OneDrive/Documents/Github/xdssat/docs/learn/data/module3/pn_info.csv')

2. Soil dataA DSSAT soil profile lists layers with their lower limit of plant-extractable water (`SLLL`, the wiltingpoint), drained upper limit (`SDUL`, field capacity), saturation (`SSAT`), a root growth factor (`SRGF`),texture and organic carbon. The water a soil can hold for the crop is the sum over layers of`(SDUL - SLLL) x layer thickness`. The soil file of DSSAT has twelve generic profiles (three depths offour textures) and the measured Millhopper fine sand of the maize experiment:

	id	name	texture	depth_cm	paw_mm	source
0	IB00000001	DEFAULT - DEEP SILTY CLAY	SIC	210.0	276.60	SOIL.SOL
1	IB00000002	DEFAULT - MEDIUM SILTY CLAY	SIC	150.0	211.80	SOIL.SOL
2	IB00000003	DEFAULT - SHALLOW SILTY CLAY	SIC	60.0	94.20	SOIL.SOL
3	IB00000004	DEFAULT - DEEP SILTY LOAM	SIL	210.0	200.55	SOIL.SOL
4	IB00000005	DEFAULT - MEDIUM SILTY LOAM	SIL	150.0	139.35	SOIL.SOL
5	IB00000006	DEFAULT - SHALLOW SILTY LOAM	SIL	60.0	60.45	SOIL.SOL
6	IB00000007	DEFAULT - DEEP SANDY LOAM	SL	210.0	229.20	SOIL.SOL
7	IB00000008	DEFAULT - MEDIUM SANDY LOAM	SL	150.0	165.60	SOIL.SOL
8	IB00000009	DEFAULT - SHALLOW SANDY LOAM	SL	60.0	72.90	SOIL.SOL
9	IB00000010	DEFAULT - DEEP SAND	S	210.0	151.35	SOIL.SOL
10	IB00000011	DEFAULT - MEDIUM SAND	S	150.0	109.35	SOIL.SOL
11	IB00000012	DEFAULT - SHALLOW SAND	S	60.0	43.95	SOIL.SOL
12	IBSB910017	Orangeburg Sandy Loam	-99	203.0	215.10	SOIL.SOL
13	IBMZ910014	Millhopper Fine Sand	-99	180.0	160.95	SOIL.SOL

Estimating water limits from textureWhen only texture and organic matter are known, a pedotransfer function estimates the limits. xdssat usesSaxton and Rawls (2006) (`src/soils.py`); DSSAT’s SBuild tool uses a Saxton and Rawls based method too(Gijsman et al. 2002, as the course exercise says). Applied to the layers above (organic matter taken as 1.724 xorganic carbon):

RMSE LL, DUL, SAT: [0.037, 0.045, 0.059]

	id	name	bottom	clay	silt	OC	LL	LL_sr	DUL	DUL_sr	SAT	SAT_sr
0	IB00000001	DEFAULT - DEEP SILTY CLAY	5.0	50.0	45.0	1.75	0.228	0.292640	0.385	0.422030	0.481	0.544967
1	IB00000001	DEFAULT - DEEP SILTY CLAY	15.0	50.0	45.0	1.75	0.228	0.292640	0.385	0.422030	0.481	0.544967
2	IB00000001	DEFAULT - DEEP SILTY CLAY	30.0	50.0	45.0	1.60	0.249	0.292714	0.406	0.422954	0.482	0.542675
3	IB00000001	DEFAULT - DEEP SILTY CLAY	45.0	50.0	45.0	1.45	0.249	0.292788	0.406	0.423878	0.465	0.540385
4	IB00000001	DEFAULT - DEEP SILTY CLAY	60.0	50.0	45.0	1.45	0.249	0.292788	0.406	0.423878	0.465	0.540385
5	IB00000001	DEFAULT - DEEP SILTY CLAY	90.0	50.0	45.0	1.10	0.308	0.292960	0.456	0.426039	0.468	0.535043
6	IB00000001	DEFAULT - DEEP SILTY CLAY	120.0	50.0	45.0	0.65	0.207	0.293181	0.341	0.428824	0.452	0.528182
7	IB00000001	DEFAULT - DEEP SILTY CLAY	150.0	50.0	45.0	0.30	0.243	0.293353	0.365	0.430995	0.455	0.522851
8	IB00000001	DEFAULT - DEEP SILTY CLAY	180.0	50.0	45.0	0.10	0.259	0.293451	0.361	0.432238	0.457	0.519806
9	IB00000001	DEFAULT - DEEP SILTY CLAY	210.0	50.0	45.0	0.01	0.259	0.293495	0.361	0.432798	0.457	0.518437
10	IB00000002	DEFAULT - MEDIUM SILTY CLAY	5.0	50.0	45.0	1.74	0.228	0.292645	0.385	0.422091	0.463	0.544814
11	IB00000002	DEFAULT - MEDIUM SILTY CLAY	15.0	50.0	45.0	1.74	0.228	0.292645	0.385	0.422091	0.463	0.544814

The course’s soil exercise: Umariya and OrangeburgThe exercise gives the laboratory sheets of two soil series and asks for a DSSAT profile of each, withSBuild filling in the water limits from texture. The sheets’ values are in `course-data/umariya_soil.csv`(Umariya series, Shahdol district, India) and `course-data/orangeburg_soil.csv` (Orangeburg series pedon 2,Peach County, Georgia, from Perkins 1987). The notebook builds both profiles with `sol.SoilProfile`,estimating the water limits with Saxton and Rawls and the root growth factor with `soils.root_growth_factor`.

particle sizes adding up to 100%?
horizon  top  bottom  sand  silt  clay   sum
     Ap    0      11  62.3  13.2  24.5 100.0
     B1   11      31  55.2  18.4  36.4 110.0
    B12   31      50  45.6  11.1  43.3 100.0
     B2   50      75  38.9  12.9  48.2 100.0
    B31   75     100  33.8  13.6  52.6 100.0
    B32  100     125  32.4  10.7  56.9 100.0
    B33  125     150  30.3  14.3  55.4 100.0

The Umariya sheet’s 11-31 cm layer adds up to 110%: the sheet itself has 55.2% sand, 18.4% silt and 36.4% clay.Saxton and Rawls use sand and clay only, so the notebook keeps those two as printed and shows below what thelayer’s water limits would be if the sand were 45.2% instead (the value that would make the sum 100%).

*SOILS: Course soil exercise: Umariya and Orangeburg

*CMUM000001  course      SCL     150 Umariya series (course sheet)
@SITE        COUNTRY          LAT     LONG SCS FAMILY
 -99         India         23.500   80.750 -99
@ SCOM  SALB  SLU1  SLDR  SLRO  SLNF  SLPF  SMHB  SMPX  SMKE
   -99  0.13   9.3  0.60  75.0  1.00  1.00 IB001 IB001 IB001
@  SLB  SLMH  SLLL  SDUL  SSAT  SRGF  SSKS  SBDM  SLOC  SLCL  SLSI  SLCF  SLNI  SLHW  SLHB  SCEC  SADC
    11   -99 0.157 0.251 0.417 1.000  1.10  1.54  0.97  24.5  13.2   -99   -99   6.7   -99  18.5   -99
    31   -99 0.223 0.327 0.421 0.657  0.20  1.53  0.66  36.4  18.4   -99   -99   6.4   -99  19.3   -99
    50   -99 0.261 0.377 0.442 0.445  0.07  1.48  0.50  43.3  11.1   -99   -99   6.4   -99  24.5   -99
    75   -99 0.288 0.409 0.461 0.287  0.04  1.43  0.40  48.2  12.9   -99   -99   6.4   -99  31.3   -99
   100   -99 0.312 0.435 0.480 0.174  0.02  1.38  0.30  52.6  13.6   -99   -99   5.7   -99  33.5   -99
   125   -99 0.337 0.459 0.494 0.105  0.01  1.34  0.25  56.9  10.7   -99   -99   5.6   -99  36.4   -99
   150   -99 0.328 0.451 0.494 0.064  0.02  1.34  0.10  55.4  14.3   -99   -99   5.7   -99  39.5   -99

*CMOG000002  course      LS      180 Orangeburg pedon 2 (Perkins 1987)
@SITE        COUNTRY          LAT     LONG SCS FAMILY
 -99         USA          -99.000  -99.000 -99
@ SCOM  SALB  SLU1  SLDR  SLRO  SLNF  SLPF  SMHB  SMPX  SMKE
   -99  0.13   9.3  0.60  75.0  1.00  1.00 IB001 IB001 IB001
@  SLB  SLMH  SLLL  SDUL  SSAT  SRGF  SSKS  SBDM  SLOC  SLCL  SLSI  SLCF  SLNI  SLHW  SLHB  SCEC  SADC
    18   -99 0.053 0.124 0.437 0.835  7.64  1.71  1.02   6.7  15.8   -99   -99   5.2   -99   7.9   -99
    28   -99 0.079 0.149 0.393 0.631  3.56  1.68  0.22  13.1  13.8   -99   -99   5.0   -99   4.7   -99
    38   -99 0.118 0.196 0.389 0.517  1.74  1.72  0.18  19.5  12.4   -99   -99   5.1   -99   5.2   -99
    74   -99 0.190 0.287 0.401 0.326  0.37  1.74  0.10  31.6  10.4   -99   -99   5.5   -99   6.7   -99
   122   -99 0.213 0.313 0.408 0.141  0.21  1.72  0.08  35.4   8.0   -99   -99   5.8   -99   6.1   -99
   150   -99 0.233 0.337 0.415 0.066  0.12  1.74  0.04  38.9   6.6   -99   -99   5.9   -99   5.0   -99
   180   -99 0.236 0.340 0.415 0.037  0.11  1.71  0.06  39.4   5.2   -99   -99   5.0   -99   5.6   -99

{'LL_printed': 0.2231236088128,
 'LL_sand45': 0.2223893120128,
 'DUL_printed': 0.3274041546780846,
 'DUL_sand45': 0.33916132251787046,
 'SAT_printed': 0.4209602011587246,
 'SAT_sand45': 0.43506471383051043}

Self test 3 compares the drained upper limit and lower limit of Orangeburg with the sheet’s measured watercontents at 1/3 bar (33 kPa) and 15 bar (1500 kPa). The sheet gives them in % by weight; times bulk densitythey are volumetric. Its own “available water” column, in cm³/cm³, is the check: (8.2 - 2.8) x 1.71 / 100 =0.092 for the top layer, printed as 0.09. Self test 2 compares with soils of the same series already inSOIL.SOL: IBSB910017, an Orangeburg sandy loam.

	horizon	top	bottom	clay	bulk_density	water_33kpa_pct_wt	water_1500kpa_pct_wt	available_water_fc	DUL_measured	LL_measured	available_check	DUL_sr	LL_sr
0	Ap	0	18	6.7	1.71	8.2	2.8	0.09	0.14022	0.04788	0.09234	0.123586	0.053408
1	AB	18	28	13.1	1.68	7.5	4.2	0.06	0.12600	0.07056	0.05544	0.148969	0.078930
2	BA	28	38	19.5	1.72	10.2	6.0	0.07	0.17544	0.10320	0.07224	0.195836	0.117693
3	Bt1	38	74	31.6	1.74	14.0	9.9	0.07	0.24360	0.17226	0.07134	0.286747	0.190058
4	Bt2	74	122	35.4	1.72	16.3	11.3	0.09	0.28036	0.19436	0.08600	0.312651	0.212585
5	Bt3	122	150	38.9	1.74	17.7	12.7	0.09	0.30798	0.22098	0.08700	0.337427	0.233117
6	Bt4	150	180	39.4	1.71	16.9	12.8	0.07	0.28899	0.21888	0.07011	0.339911	0.236274

WindowsPath('C:/Users/benny/OneDrive/Documents/Github/xdssat/docs/learn/data/module3/soil_sheet_orangeburg.csv')

3. Water-limited productionThe course’s water-limited exercise builds `UFGA8212.MZX` from the potential production experiment ofmodule 1: McCurdy 84aa maize planted on 26 February 1982 (7.2 plants/m², 61 cm rows, 7 cm deep), theMillhopper fine sand, simulation from 25 February with the initial soil water the exercise lists, the waterbalance on and nitrogen off. Treatment 1 is rainfed (no irrigation at all); treatment 2 irrigatesautomatically when the available water in the top 30 cm falls below 50%. Rain and irrigation enter thesoil; water runs off, drains, evaporates from the soil and is taken up by the roots from each layer(`Soil/SoilWater/WATBAL.for`).`filex.build_filex` copies a template treatment; the rainfed treatment needs irrigation level 0, which thetemplate’s treatment 1 does not have, and the automatic treatment needs its own simulation-control level.`learnlib.set_treatment_levels` and a second simulation-control block do what XBuild does for these two.

*TREATMENTS                        -------------FACTOR LEVELS------------
@N R O C TNAME.................... CU FL SA IC MP MI MF MR MC MT ME MH SM
 1 1 0 0 rainfed                    1  1  0  1  1  0  1  0  0  0  0  0  1

*CULTIVARS
@C CR INGENO CNAME
 1 MZ IB0035 McCurdy 84aa

*FIELDS
@L ID_FIELD

level	HWAM	CWAM	LAIX	PRCM	IRCM	IR#M	ETCM	EPCM	ESCM	ROCM	DRCM	SWXM
potential	11859	23174	4.562119	NaN	0	0	NaN	NaN	NaN	NaN	NaN	NaN
rainfed	2010	7144	3.157887	661.0	0	0	330.0	240.0	90.0	35.0	309.0	139.0
automatic irrigation	11691	22990	4.562119	661.0	194	12	469.0	389.0	80.0	38.0	354.0	146.0

soil file layers: [5.0, 15.0, 30.0, 60.0, 90.0, 120.0, 150.0, 180.0]

	layer	top	bottom	LL	DUL	SAT
0	1	0.0	5.0	0.026	0.096	0.23
1	2	5.0	15.0	0.025	0.086	0.23
2	3	15.0	30.0	0.025	0.086	0.23
3	4	30.0	45.0	0.025	0.086	0.23
4	5	45.0	60.0	0.025	0.086	0.23
5	6	60.0	90.0	0.028	0.090	0.23
6	7	90.0	120.0	0.028	0.090	0.23
7	8	120.0	150.0	0.029	0.130	0.23
8	9	150.0	180.0	0.070	0.258	0.36

The daily SoilWat output reports soil water (SW1D, SW2D …) for the layers the engine simulates,which are not the layers of the soil file: with the soil layer option MESOL = 2 the engine keeps 5 and15 cm at the top and splits thick layers (LYRSET2, LMATCH.for:263).learnlib.engine_layers, used above, is a port of that rule.### Self test 2: the twelve generic soils, and the course’s twoThe rainfed treatment on each generic profile, on the measured Millhopper fine sand, and on the Umariya andOrangeburg profiles built above. Initial soil water is rebuilt for each soil at 100% of its plant-availablerange (ic_water=1.0), the default the exercise mentions before it sets the Millhopper values by hand.

	id	name	texture	depth_cm	paw_mm	source	HWAM	CWAM	ETCM	DRCM	ROCM
11	IB00000012	DEFAULT - SHALLOW SAND	S	60.0	43.950000	SOIL.SOL	1117	6664	358	221	100
5	IB00000006	DEFAULT - SHALLOW SILTY LOAM	SIL	60.0	60.450000	SOIL.SOL	1086	7766	414	99	169
8	IB00000009	DEFAULT - SHALLOW SANDY LOAM	SL	60.0	72.900000	SOIL.SOL	1098	7806	408	172	105
2	IB00000003	DEFAULT - SHALLOW SILTY CLAY	SIC	60.0	94.200000	SOIL.SOL	2537	9777	415	44	292
10	IB00000011	DEFAULT - MEDIUM SAND	S	150.0	109.350000	SOIL.SOL	4020	11515	403	211	67
4	IB00000005	DEFAULT - MEDIUM SILTY LOAM	SIL	150.0	139.350000	SOIL.SOL	5643	14510	455	121	148
9	IB00000010	DEFAULT - DEEP SAND	S	210.0	151.350000	SOIL.SOL	3856	11161	397	242	44
12	IBMZ910014	Millhopper Fine Sand	-99	180.0	160.950000	SOIL.SOL	2010	7144	330	318	35
7	IB00000008	DEFAULT - MEDIUM SANDY LOAM	SL	150.0	165.600000	SOIL.SOL	7079	16840	460	177	79
14	CMOG000002	Orangeburg pedon 2 (Perkins 1987)	LS	180.0	170.588072	course exercise	3692	10845	406	188	91
13	CMUM000001	Umariya series (course sheet)	SCL	150.0	175.347325	course exercise	5220	12750	437	132	138
3	IB00000004	DEFAULT - DEEP SILTY LOAM	SIL	210.0	200.550000	SOIL.SOL	5203	13716	448	139	127
1	IB00000002	DEFAULT - MEDIUM SILTY CLAY	SIC	150.0	211.800000	SOIL.SOL	10193	21025	483	64	264
6	IB00000007	DEFAULT - DEEP SANDY LOAM	SL	210.0	229.200000	SOIL.SOL	6995	16723	460	181	70
0	IB00000001	DEFAULT - DEEP SILTY CLAY	SIC	210.0	276.600000	SOIL.SOL	10263	21209	487	80	234

Self test 3: when to irrigateWith `IRRIG` = `A` the engine irrigates when the available water in the top `IMDEP` cm of soil falls to`ITHRL` percent of its capacity, and refills it to `ITHRU`(`Management/IRRIG.for:827`, with the available fraction from`SWDEFICIT`, lines 1025-1064). The exercise’s grid: depths of 10, 30, 50 and 70 cm and thresholds of 10, 30, 50,70 and 90%.

depth_cm	10	30	50	70
threshold
10	7403.0	9183.0	9381.0	7927.0
30	9732.0	11016.0	11505.0	10753.0
50	10194.0	11691.0	11859.0	11818.0
70	11842.0	11859.0	11859.0	11859.0
90	11824.0	11859.0	11859.0	11859.0

Self test 4: other years at GainesvilleThe same rainfed and automatically irrigated maize planted on 26 February of every year for whichdssat-csm-data has a full Gainesville weather file. All the experiment’s dates move with planting.

28 years: [2006, 2007, 2021, 2022, 1958, 1959, 1960, 1961, 1962, 1964, 1965, 1966, 1968, 1969, 1970, 1971, 1978, 1979, 1980, 1981, 1982, 1983, 1984, 1985, 1986, 1987, 1989, 1990] | skipped, unreadable: ['UFGA6701.WTH', 'UFGA7601.WTH', 'UFGA9601.WTH']

level	automatic irrigation	rainfed
year
1958	10654	7867
1959	11733	9545
1960	11418	1687
1961	12821	1673
1962	11743	2484
1964	11559	4321
1965	13407	1754
1966	12213	7788
1968	11251	3248
1969	10103	2943
1970	12149	4648
1971	13106	4277
1978	11496	2652
1979	13017	6511
1980	10870	6213
1981	10387	1324
1982	11691	2010
1983	11862	4840
1984	10493	3145
1985	9794	3356
1986	14648	4081
1987	10847	4401
1989	10183	2051
1990	11194	1754
2006	10554	1322
2007	11378	3946
2021	11564	2082
2022	9837	5559

WindowsPath('C:/Users/benny/OneDrive/Documents/Github/xdssat/docs/learn/data/module3/wl_files.csv')

Paddy rice: the flood water balanceFor rice grown in bunded, puddled fields the water balance has one more store: the flood above the soil. Bundshold rain and irrigation as standing water up to their height; the flood loses water by evaporation, byinfiltration into the puddled soil at a percolation rate, and by spilling over the bund. The engine keeps thisbalance daily (`Management/Paddy_Mgmt.for`, written to `FloodW.OUT` by `Management/OpFlood.for`) when theirrigation section of the experiment file sets the paddy operations: `IR010` puddling, `IR008` percolation rate(mm/day), `IR009` bund height (mm), `IR011` constant flood depth (mm) and `IR006` flood depth (mm) (codes andmeanings from the engine’s `DETAIL.CDE`).The runs use xdssat’s own rice template, `templates/XDRI8601.RIX` (IRRI’s IRMZ8601 experiment at Munoz,Philippines, with an added rainfed bunded water level), with its three water regimes, in the dry season as in theexperiment (transplanted early February 1986) and in the wet season (1 July 1986), nitrogen off.

	TRNO	season	regime	HWAM	PRCM	IRCM	ROCM	DRCM	ETCM	days_to_maturity	flood_days	flood_mean	flood_evap	infiltration	over_bund	WSPD	WSGD
0	1	dry season	flooded, 50 mm kept	8939	13	784	0	178	577	89	91	46.233333	151.2	596.8	0.0	0.000000	0.000000
1	2	dry season	flooded, drained twice	8939	13	859	75	178	577	89	91	44.169892	151.2	596.8	75.0	0.000000	0.000000
2	3	dry season	rainfed, bunded	8760	834	0	0	458	463	161	1	0.013939	0.0	834.0	0.0	0.328917	0.359965
3	4	wet season	flooded, 50 mm kept	8876	1109	77	317	198	541	97	100	99.623762	153.0	582.4	317.3	0.000000	0.000000
4	5	wet season	flooded, drained twice	8876	1109	241	481	198	541	97	100	78.639604	153.0	582.4	481.0	0.000000	0.000000
5	6	wet season	rainfed, bunded	8876	1109	0	258	196	538	97	99	100.369307	149.4	581.5	258.2	0.000000	0.000000

WindowsPath('C:/Users/benny/OneDrive/Documents/Github/xdssat/docs/learn/data/module3/flood_files.csv')

4. Things to try- Run GLUE with `N = 500` and again with `N = 3000`. How much do the best values move? That spread is part of the answer, not noise to hide.- Remove one calibration experiment and recalibrate. Which coefficients change most?- Build your own soil profile with `SoilProfile` and `saxton_rawls` (sand, clay and organic matter of your field) and run the rainfed maize on it.- In the automatic irrigation runs, which threshold gives most grain per mm of water applied, and is it the one that gives the highest yield?- Run GLUE for DEKALB 52-59 AF a second time with another seed, as the exercise suggests. How far apart are the two answers?- In the peanut grid, is the combination with the best mean d also best for each variable on its own?

	0
engine	C:\TEMP\dsn\dssat\dscsm048.exe
csm_data_commit	99c95b340acb9515c2d0a8ab582f5fbbf076f345
csm_os_commit	25c93b9a2f73f8ec45bf441b198dac5fb7b22849
glue_commit	e86093e400fbbdfc8e2c751c98acc2136835c3af
run_on	2026-11-07 11:10