Evaluation of rainfall–runoff model performance under non-stationary hydroclimatic conditions

Figures & data

Figure 1. Location of the selected study catchments.

Figure 2. Land-use/land cover map (2012) of catchments (a) 405251 and (b) 410044.

Figure 3. Periods during which the hydrological models were calibrated and validated.

Table 1. Sensitivity analysis of SWATgrid model parameters at catchment 410044.

Parameter	Percent change in NSE (+25% change in parameter)	Percent change in NSE (−25% change in parameter)	Sensitivity level
ALPHA_BF	19.32	18.17	High
Ch_K2	3.26	2.19	Moderate
GWQMN	23.36	27.73	High
SOL_AWC	29.17	21.18	High
SOL_K	17.76	19.17	High
CN2	25.55	26.23	High
ESCO	20.34	19.90	High
CH_N2	1.16	1.92	Low
GW_DELAY	2.33	2.04	Moderate
GW_REVAP	1.77	1.86	Low
RECHARG_DP	1.69	1.96	Low
REVAPMN	7.33	6.94	Moderate
SOL_Z	0.43	0.51	Low
OV_N	0.66	0.72	Low

Table 2. IHACRES, HEC-HMS and SWATgrid model parameter definitions, units and their range used in this study.

Model	Parameter	Definition	Units	Range
IHACRES	τq	Time constant controlling the recession of quickflow	Day	0.5–10
	τs	Time constant controlling the recession of slow flow	Day	10 – 350
	vs	Fraction of slow flow to total flow	(-)	0 – 1
	d	Flow threshold	mm	5 – 550
	e	Temperature to PET conversion factor	(-)	0.01–1.5
	f	Plant stress threshold factor	(-)	0.01–3
HEC-HMS	CN	Curve Number	(-)	1 – 100
	Ip	Percent imperviousness	%	0.5 – 3
	Ia	Initial abstraction	mm	0 – 500
	Lt	Lag time	min	0.1 – 30,000
	K	Muskingum K	hr	0.1–150
	x	Muskingum x	(-)	0 – 0.5
SWATgrid	GWQMN	Threshold depth of water in the shallow aquifer required for return flow to occur	mm	0 – 5000
	SOL_AWC	Available water holding capacity of the soil	mm mm^−1	0 – 1
	ALPHA_BF	Baseflow alpha factor	Day^−1	0 – 1
	SOL_K	Saturated hydraulic conductivity	mm h^−1	0 – 0.15
	CN2	SCS-CN for moisture condition II	(-)	35 – 98
	ESCO	Soil evaporation compensation factor	(-)	0.01 – 1

Table 3. Average, Dry and Wet periods chosen for catchments 405251 and 410044.

Catchment	MAR for whole record (mm)	Periods	Years selected	Percent change in MAR relative to the MAR for the whole record
405251	967	Average1	1976-1980	−3.2
		Dry1	2000-2004	−25.5
		Wet1	1987-1991	15.8
		Average2	1981-1985	0.5
		Dry2	2005-2009	−26.5
		Wet2	1992-1996	15.0
410044	571	Average1	1971-1979	−1.6
		Dry1	1943-1951	−49.9
		Wet1	1962-1970	28.2
		Average2	1980-1988	−2.8
		Dry2	2001-2009	−29.8
		Wet2	1992-2000	22.1

Figure 4. Average1, Average2, Dry1, Dry2, Wet1 and Wet2 periods chosen for model calibration and validation in this study for catchments (a) 405251 and (b) 410044 (MAR represents mean annual rainfall for the whole data record).

Figure 5. Flow–duration curves for model comparison during (a) calibration (Average1), (b) validation (Average2), (c) validation (Dry2) and (d) validation (Wet2) in case of catchment 405251 (vertical grey lines indicate the division of 10% and 90% flows).

Table 4. Model comparison statistics during calibration in Average1 and validation in Average2, Dry2 and Wet2 periods at catchment 405251. Bold figures represent values within an acceptable range as detailed in Section 3.5.

Period	Model	NSE	RMSE (m^3 s^−1)*	PBIAS (%)
Calibration (Average1)	IHACRES	0.83	0.29	10.40
	HEC-HMS	0.80	0.31	18.02
	SWATgrid	0.94	0.17	−2.10
Validation (Average2)	IHACRES	0.76	0.61	−6.40
	HEC-HMS	0.75	0.66	−5.30
	SWATgrid	0.95	0.30	−3.81
Validation (Dry2)	IHACRES	0.47	0.22	−33.45
	HEC-HMS	0.39	0.29	−23.62
	SWATgrid	0.81	0.16	−5.40
Validation (Wet2)	IHACRES	−1.62	1.43	37.44
	HEC-HMS	−1.78	1.45	39.32
	SWATgrid	0.87	0.65	12.74

*acceptable values for RMSE are 0.34, 0.71, 0.17 and 1.02 m³ s⁻¹ for Average1, Average2, Dry2 and Wet2, respectively, and are calculated from the observed runoff for the corresponding periods (as described in Section 3.5).

Table 5. Model comparison statistics during calibration in Dry1 and validation in Average2, Dry2 and Wet2 periods at catchment 405251. Bold figures represent values within an acceptable range as detailed in Section 3.5.

Period	Model	NSE	RMSE (m^3 s^−1)*	PBIAS (%)
Calibration (Dry1)	IHACRES	0.73	0.32	7.50
	HEC-HMS	0.71	0.46	13.72
	SWATgrid	0.92	0.15	3.61
Validation (Average2)	IHACRES	0.68	1.02	−14.62
	HEC-HMS	−6.12	1.44	97.52
	SWATgrid	0.84	0.43	1.48
Validation (Dry2)	IHACRES	0.49	0.21	−3.29
	HEC-HMS	0.51	0.29	−6.86
	SWATgrid	0.87	0.16	3.59
Validation (Wet2)	IHACRES	0.70	1.04	5.10
	HEC-HMS	0.67	1.11	5.33
	SWATgrid	0.78	0.44	3.66

*Acceptable values for RMSE are 0.32, 0.71, 0.17 and 1.02 m³ s⁻¹ for Dry1, Average2, Dry2 and Wet2, respectively, and are calculated from the observed runoff for the corresponding periods (as described in Section 3.5).

Figure 6. Flow–duration curves for model comparison during (a) calibration (Dry1), (b) validation (Average2), (c) validation (Dry2) and (d) validation (Wet2) in case of catchment 405251 (vertical grey lines indicate the division of 10% and 90% flows).

Table 6. Model comparison statistics during calibration in Wet1 and validation in Average2, Dry2 and Wet2 periods at catchment 405251. Bold figures represent values within an acceptable range as detailed in Section 3.5.

Period	Model	NSE	RMSE (m^3 s^−1)*	PBIAS (%)
Calibration (Wet1)	IHACRES	0.78	0.46	−0.28
	HEC-HMS	0.77	0.47	2.41
	SWATgrid	0.95	0.21	−1.42
Validation (Average2)	IHACRES	0.75	0.63	−17.82
	HEC-HMS	0.71	0.62	−20.90
	SWATgrid	0.86	0.34	−14.43
VALIDATION (Dry2)	IHACRES	0.41	0.49	−51.15
	HEC-HMS	0.33	0.58	−54.20
	SWATgrid	0.57	0.27	−28.85
Validation (Wet2)	IHACRES	0.66	0.85	15.26
	HEC-HMS	0.71	0.96	16.21
	SWATgrid	0.87	0.59	3.39

*acceptable values for RMSE are 0.47, 0.71, 0.17 and 1.02 m³ s⁻¹ for Wet1, Average2, Dry2 and Wet2, respectively, and are calculated from the observed runoff for the corresponding periods (as described in Section 3.5).

Figure 7. Flow–duration curves for model comparison during (a) calibration (Wet1), (b) validation (Average2), (c) validation (Dry2) and (d) validation (Wet2) in case of catchment 405251 (vertical grey lines indicate the division of 10% and 90% flows).

Table 7. Model comparison statistics during calibration in Average1 and validation in Average2, Dry2 and Wet2 periods at catchment 410044. Bold figures represent values within an acceptable range as detailed in Section 3.5.

Period	Model	NSE	RMSE (m^3 s^−1)*	PBIAS (%)
Calibration (Average1)	IHACRES	−0.28	4.74	−24.72
	HEC-HMS	−0.53	4.38	−31.15
	SWATgrid	0.85	1.09	−11.82
Validation (Average2)	IHACRES	−0.38	4.39	−29.82
	HEC-HMS	−0.44	5.20	−26.82
	SWATgrid	0.90	0.89	−9.20
Validation (Dry2)	IHACRES	−0.16	2.14	−82.09
	HEC-HMS	−0.19	2.06	−56.21
	SWATgrid	0.43	0.97	−69.81
Validation (Wet2)	IHACRES	−0.06	3.78	−34.95
	HEC-HMS	−0.28	2.85	−26.69
	SWATgrid	0.67	1.89	−21.89

*acceptable values for RMSE are 3.12, 2.84, 0.52 and 2.34 m³ s⁻¹ for Average1, Average2, Dry2 and Wet2, respectively, and are calculated from the observed runoff for the corresponding periods (as described in Section 3.5).

Figure 8. Flow–duration curves for model comparison during (a) calibration (Average1), (b) validation (Average2), (c) validation (Dry2) and (d) validation (Wet2) in case of catchment 410044 (vertical grey lines indicate the division of 10% and 90% flows).

Table 8. Model comparison statistics during calibration in Dry1 and validation in Average2, Dry2 and Wet2 periods at catchment 410044. Bold figures represent values within an acceptable range as detailed in Section 3.5.

Period	Model	NSE	RMSE (m^3 s^−1)*	PBIAS (%)
Calibration (Dry1)	IHACRES	−17.42	2.22	210.41
	HEC-HMS	−14.42	2.36	200.41
	SWATgrid	0.81	0.71	−30.31
Validation (Average2)	IHACRES	−8.84	5.06	22.71
	HEC-HMS	−8.51	4.12	−31.10
	SWATgrid	0.84	1.66	18.82
Validation (Dry2)	IHACRES	−0.05	4.15	62.88
	HEC-HMS	−1.13	1.58	−66.25
	SWATgrid	0.87	0.29	−16.72
Validation (Wet2)	IHACRES	−6.90	3.86	113.46
	HEC-HMS	−4.74	2.89	−24.51
	SWATgrid	0.39	2.45	99.12

*acceptable values for RMSE are 1.08, 2.84, 0.52 and 2.34 m³ s⁻¹ for Dry1, Average2, Dry2 and Wet2, respectively, and are calculated from the observed runoff for the corresponding periods (as described in Section 3.5).

Figure 9. Flow–duration curves for model comparison during (a) calibration (Dry1), (b) validation (Average2), (c) validation (Dry2) and (d) validation (Wet2) in case of catchment 410044 (vertical grey lines indicate the division of 10% and 90% flows).

Table 9. Model comparison statistics during calibration in Wet1 and validation in Average2, Dry2 and Wet2 periods at catchment 410044. Bold figures represent values within an acceptable range as detailed in Section 3.5.

Period	Model	NSE	RMSE (m^3 s^−1)*	PBIAS (%)
Calibration (Wet1)	IHACRES	−0.11	3.55	21.60
	HEC-HMS	−0.08	3.54	30.30
	SWATgrid	0.83	1.69	6.30
Validation (Average2)	IHACRES	−0.76	3.95	37.91
	HEC-HMS	−0.95	4.18	40.21
	SWATgrid	0.91	1.18	−3.53
Validation (Dry2)	IHACRES	−0.81	1.06	23.62
	HEC-HMS	−0.67	1.12	20.31
	SWATgrid	0.49	0.81	−32.41
Validation (Wet2)	IHACRES	−0.09	3.09	38.31
	HEC-HMS	−0.06	3.34	7.01
	SWATgrid	0.88	1.36	16.12

*acceptable values for RMSE are 2.62, 2.84, 0.52 and 2.34 m³ s⁻¹ for Wet1, Average2, Dry2 and Wet2, respectively, and are calculated from the observed runoff for the corresponding periods (as described in Section 3.5).

Figure 10. Flow–duration curves for model comparison during (a) calibration (Wet1), (b) validation (Average2), (c) validation (Dry2) and (d) validation (Wet2) in case of catchment 410044 (vertical grey lines indicate the division of 10% and 90% flows).