Impact analysis of moisture stress on growth and yield of cotton using DSSAT-CROPGRO-cotton model under semi-arid climate

Introduction

Although India is a major cotton producer with 13 M ha area and 29.5 M bales production its productivity of 494 kg ha⁻¹ is very low when compared to world average cotton productivity of 764 kg ha⁻¹ (Prakash, 2019; Singh, Brar & Mishra, 2021). The appearance of various biotic and abiotic constraints during crop growth period is the primary reason for the poor productivity (Mahalingam, 2015; Pandey, Ramegowda & Senthil-Kumar, 2015; Singh, Rathore & Mishra, 2022b; Habib-ur-Rahman et al., 2022). High temperatures at early growth stages increase the field evaporation and crop transpiration excessively which exacerbates the water stress in the cotton fields. Poor distribution of untimely and scanty rainfall often causes huge reductions in the cotton production. Paltasingh, Goyari & Mishra (2012) also reported that cotton yield is largely affected by both rainfall and temperature through crop-weather interactions.

In India, about 65% of the cotton cultivation is rainfed. However, in north-western India comprising the states like Punjab, Haryana and Rajasthan, cotton is grown under irrigated conditions. In Punjab, out of 4.2 M ha total cultivated area, 4.01 M ha is irrigated. Canal irrigation covers only 28% area, while 72% area is irrigated by tube wells (Brar, Buttar & Sharma, 2018). Consequently, groundwater table is depleting at a scale of 0.51 m year⁻¹ (Singh et al., 2022a). The annual water balance of the state shows a deficit of 1.6 M ha^−m and, by the year 2050, per capita availability of surface water is projected to fall drastically (Arora & Kukal, 2017). A comprehensive study elucidated that against annual availability of 21.6 billion m³ (BCM) in Punjab, the groundwater extraction is 35.8 BCM (Anonymous, 2018). The continuous Rice-Wheat cropping system (RWCS) in Punjab cumulatively needs about 2,000 mm of water in which rice crop alone requires approximately 1,600 mm (Bhatt et al., 2021). The high water requirement of RWCS necessitates replacing the rice with less water consuming crop like cotton, which offers a great potential for water resource conservation without compromising with the economic returns (Singh, Singh & Mishra, 2020).

Large fluctuations in weather conditions largely affect the cotton growth and productivity (Kaur et al., 2019). Cotton needs about 550 to 950 mm of water during its life cycle to be met either through rainfall or irrigation. In southwestern region of Punjab, the average rainfall of the cotton growing season (~365 mm) usually fails to meet the cotton crop water requirement. Furthermore, diurnal maximum temperature usually exceeds 45 °C and occasionally reaches 48 °C during sowing period or during early crop growth phase which coincides with the high evaporation period in semi-arid environments (Singh, Brar & Mishra, 2021). It is well established that rainfall and temperature jointly govern about 50% of yield variation of dry land cotton while other crop management factors affect the remaining variability.

Owing to the aforementioned climatic aberrations, the cotton crop in near future is expected to face multiple abiotic stresses, including high temperature and reduced water availability (Timothy & Michael, 2014; Rahman et al., 2020). Large variability is expected with extremely high temperatures and high precipitation (Napoli et al., 2019; Praveen et al., 2020). Consequently, with projected warming and extreme temperature, the increased crop water requirement would exert more pressure for producing more yield per drop of water (Liu et al., 2021). Thus, agriculture is expected to be badly affected from drought rather than the extreme rainfall events (Auffhammer, Ramanathan & Vincent, 2012; Neal et al., 2020). Under such circumstances, an increase of 10 mm of rainfall could enhance the cotton yield meagerly by 7 kg ha⁻¹ (Raju et al., 2014). Hence, it is confirmed that the moisture stress would result into hampered growth and physiological processes (dos Santos et al., 2022), decreased net photosynthesis (Bryant et al., 2021), transpiration (Goufo et al., 2017), stomatal conductance (Singh, Rathore & Mishra, 2022b), leaf water potential (Goche et al., 2020), lower plant height (Singh, Brar & Singh, 2018a), poor leaf area index (Singh, Singh & Mishra, 2020), and reduced root development (Daryanto, Wang & Jacinthe, 2020). Under such circumstances, Hejnák et al. (2015) reported a 50% reduction in the crop biomass besides 50–73% reduction in the potential cotton yield (Langridge & Reynolds, 2015). Although, fiber quality traits are majorly governed by inherent genetic characters (but weather conditions and moisture status during fiber cell development phase directly influence the lint quality (Kaur, Mishra & Singh, 2023). Early fiber elongation (0–15 days after anthesis) is vital for fiber quality but water stress impedes it by reducing the fiber length and uniformity. Soil water deficit negatively affects fiber strength and elongation, while adequate water boosts fiber maturity.

Plants under moisture stress exert over accumulation of reactive oxygen species such as hydrogen peroxide (Hasanuzzaman et al., 2019), superoxide (Berwal & Ram, 2018), hydroxyl and alkoxy radicals (Mittova, Volokita & Guy, 2015) in cellular organs such as chloroplasts (Dietz, Turkan & Krieger-Liszkay, 2016), mitochondria (Huang et al., 2016), and peroxisomes (Sandalio & Romero-Puertas, 2015) which cause irreversible DNA damage and cell death (Huang et al., 2019). Contrarily, externally applied Glycine betaine could encounter the harmful effect of abiotic stresses, stimulate the growth parameters and yield through various physiological and biochemical processes (Khalid et al., 2015).

Crop growth simulation models are quantitative tools extensively being used worldwide to evaluate the individual or combined effect of edapho-climatic and management on growth, and crop yield (Mishra et al., 2015; Rahman et al., 2018). The CROPGRO-cotton model is set of the crop growth algorithms available within the decision support system for agrotechnology transfer (DSSAT) cropping system model (Li et al., 2020; Mishra, Kaur & Singh, 2021). This model has been already applied to assess the impacts of various crop management options such as irrigation water requirement (Garibay et al., 2019; Rahman et al., 2019), effects of climatic variability on yield (Paz et al., 2012), effect of temperature and solar radiation on seed cotton yield (SCY) (Pal, Kataria & Singh, 2016), and other input options, Mishra, Kaur & Singh (2021) have evaluated this model for cotton growth stages, leaf area index (LAI), crop height, above ground dry matter and SCY under different sowing environments. However, impact of deficit water stress due to rainfall shortage and reduced irrigation on the cotton growth and SCY has been little reported in previous studies. Nevertheless, model assisted appropriation of irrigation frequency based on actual crop water needs offers ample opportunity to optimize the existing irrigation facilities for ensuring high SCY and environmental sustainability (Singh et al., 2022a). The field experiments and model-based results may precisely be used to improve the crop productivity under different hydro-climatic and management scenarios. Therefore, objectives of the present study were (i) to analyze the impact of optimum and deficit irrigation levels on phenology, growth and SCY, and (ii) to simulate the growth and yield attributes of cotton under optimal and sub-optimal moisture conditions using DSSAT-CROPGRO-cotton model and their comparison with two simulated treatments of automatic irrigation and rainfed scenarios.

Materials and Methods

Results

Soil and weather characteristics

Results of the soil sample analysis indicated that soil texture of the experimental field was predominantly sandy loam (Table 1). The upper 15 cm soil profile contained 0.61% OC, and 263.4, 38.1 and 230.7 kg ha⁻¹ of N, P and K, respectively. The CEC was 8.31 Cmol kg⁻¹, while BD value was 1.53 gcm⁻³. Different soil layers had a pH range of 7.16–8.50, EC value of 0.71–0.73 dSm⁻¹, 0.31–0.61% of OC, 125.4–263.4 kg ha⁻¹ of N, 20.8–38.1 kg ha⁻¹ of P, 73.92–230.72 kg ha⁻¹ of K, 7.99–8.44 C mol kg⁻¹ of CEC, 1.53–1.68 g cm⁻³ of BD and 20–23% of FC values. The fertility level gradually decreased from top to lower soil depths. The analysis of weather data during crop growth period depicted that the mean maximum and minimum temperatures were 26.6–43.0 °C and 8.8–29.8 °C during 2018, and 22.9–44.4 °C and 9.8–28.3 °C during 2019, respectively (Fig. 1). Morning and afternoon relative humidity was 30.3–92.1% and 12.4–68.6% in 2018 and 47.7–93.3% and 20.4–69.1% in year 2019, respectively. During 2018 and 2019, the cumulative bright sunshine duration was 908 and 980 h, respectively. During the first crop season, minimum and maximum temperatures were slightly higher over later years. Consequently, 132 mm more evaporation and 61.1 mm higher rainfall was recorded during 2018 as compared to the latter year (2019). The crop season rainfall of 345 mm in 21 rainy days (days having ≥2.5 mm rain) during 2018 (Fig. 2A) and 283.9 mm in 19 rainy days during 2019 were recorded. Rain showers (days with <2.5 mm rainfall amount) were observed for 5 and 14 days during 2018 and 2019, respectively. During 2018 and 2019, 125 and 118 days experienced no rainfall (i.e., dry days), respectively (Figs. 2A and 2B). The 2 years mean GDD (growing degree days), HTU (helio-thermal unit) and PTU (photo thermal unit) values were 2,622 °C day, 14,208 °C day and 34,988 °C day hour, respectively (Table 2).

Observed and simulated leaf area index and plant height

The maximum observed LAI was 3.26 ± 0.73 during 2018 and 4.55 ± 0.32 during 2019, while the corresponding simulated values of 3.27 ± 0.89 and 4.52 ± 0.37, respectively exhibited very close fit with the observed values (Fig. 3A). The values of d-stat ranged from 0.95 to 0.99 for LAI during 2019 and 2018; respectively (Table 7). The plant height was significantly affected by different irrigation regimes (Fig. 3B). Pooled results revealed that the crop supplied with adequate irrigation (Control) exhibited taller plants (1.75 m) followed by mild water stress i.e., 2-POSI (1.48 m), while significantly dwarf plants (1.28 m) were evident under extreme water stress regime (1-POSI). The observed mean plant height ranged between 1.62 m during 2018 and 1.38 m during 2019 with corresponding standard deviation (SD) of ±0.24 and ±0.22 m, but range of the model simulated plant height was 1.63 ± 0.33 m and 1.43 ± 0.26 m, respectively. The DSSAT-CROPGRO-cotton model recorded lower RMSE (0.07–0.09) and higher d-stat (0.95–0.97) for simulation of cotton plant height (Table 7). The observed LAI was higher than the simulated LAI until 80 DAS and afterward, there was a good agreement between these two (Figs. 4C and 4D). The simulated LAI for the automatic irrigation remained maximum however, minimum LAI values were found for the rainfed condition, during both years. The LAI values of the control, 2-POSI, and 1-POSI not only followed descending order but also remained between those of the automatic and rainfed environments.

Table 7:

Observed and simulated soil water content (mm³ mm⁻³) during 2018 and 2019.

Irrigation levels	Year	Mean_O	SD_O	Mean_S	SD_S	R2	MAE	MBE	RMSE	d-Stat
		Layer 1: 0–30 cm
Recommended	2018	0.122	0.031	0.132	0.031	0.909	0.012	0.010	0.013	0.997
	2019	0.117	0.036	0.107	0.043	0.950	0.013	−0.011	0.015	0.995
1-POSI	2018	0.087	0.042	0.080	0.052	0.985	0.010	−0.007	0.012	0.994
	2019	0.086	0.041	0.081	0.047	0.981	0.008	−0.005	0.009	0.996
2-POSI	2018	0.097	0.031	0.100	0.036	0.849	0.011	0.004	0.013	0.994
	2019	0.091	0.034	0.087	0.033	0.792	0.012	−0.004	0.015	0.991
		Layer 2: 30–60 cm
Recommended	2018	0.126	0.027	0.130	0.034	0.925	0.008	0.004	0.010	0.998
	2019	0.121	0.034	0.122	0.045	0.884	0.013	0.001	0.016	0.995
1-POSI	2018	0.092	0.047	0.088	0.056	0.988	0.010	−0.005	0.011	0.995
	2019	0.093	0.046	0.146	0.096	0.026	0.057	0.053	0.114	0.727
2-POSI	2018	0.094	0.046	0.101	0.049	0.970	0.010	0.008	0.011	0.996
	2019	0.093	0.045	0.093	0.037	0.880	0.011	0.000	0.015	0.992

DOI: 10.7717/peerj.16329/table-7

Observed and simulated soil water content

Different irrigation regimes considerably affected the soil water content of soil profile (Fig. 5). Among different irrigation regimes, relatively higher moisture content has been recorded in lower layer (30–60 cm) than upper layer (0–30 cm) at 30 DAS and 40 DAS. Later, considerably higher soil water in both the soil layers was observed for the subsequent period. For three (control, 1-POSI and 2-POSI irrigation regimes) plus two treatments (automatic and rainfed), the CROPGRO-cotton model predicted the soil water content of each soil layer both years (Figs. 5A and 5J) with RMSE of 0.016 mm³ mm⁻³, R² > 0.84, and d-stat > 0.72 (Table 8). The model performance for simulating soil water content was superior in 2018 to that of the following year. The observed soil water content of the recommended irrigation in 2018 (Fig. 5A) and 2019 (Fig. 5B) was 0.122 ± 0.31 and 0.117 ± 0.036 mm³ mm⁻³ for the soil layer above 30 cm and 0.126 ± 0.27 and 0.121 ± 0.034 mm³ mm⁻³ for the layer below 30 cm. However, the corresponding simulated soil water content was 0.132 ± 0.31 and 0.107 ± 0.043 mm³ mm⁻³ at 0–30 cm soil depth and 0.130 ± 0.34 and 0.122 ± 0.045 mm³ mm⁻³ at 30–60 cm soil depth, respectively. During both years, the simulated soil water content of soil layers above 30 cm well-matched with the observed range (Figs. 5A and 5B), but below 30 cm soil profiles, it remained slightly above to the observed line, especially during early and later developmental stages (Fig. 5B). The soil moisture content of 1-POSI for both years (Figs. 5C and 5D) closely matched with the observed until 45 DAS, but over forecasted for the subsequent period. Similarly in 2-POSI, the soil water content during early development stages showed considerable variance during 2018 (Fig. 5E) compared to 2019. (Fig. 5F). As on the 144^th day during 2018, (Fig. 5G) and at 148^th day in 2019, the CROPGRO-cotton model terminated the simulation process of soil water content under a rainfed environment (Fig. 5H). However, after activating the automatic irrigation option, it was extended to more than 180 days for both years (Figs. 5I and 5J). Regardless of the soil profile, the rainfed (Figs. 5G and 5H) and automatic irrigation (Figs. 5I and 5J) patterns for the soil water content were nearly similar in both years. Unadjusted plant transpiration coefficient may be a blame factor for the overestimations of soil water content. However, such discrepancies between cotton growth and yield parameters predicted by models and actually observed are well documented (Hussein, Janat & Yakoub, 2011).

Table 8:

Observed and simulated cotton phenology, growth and yield attributes during 2018 and 2019.

Error analysis	Emergence (DAS)		Anthesis (DAS)		First boll (DAS)		Maturity (DAS)
	2018	2019	2018	2019	2018	2019	2018	2019
Phenological stages
Mean_O	5.0	5.0	72.3	69.3	92.3	91.7	178.3	173.3
SD_O	–	–	2.5	2.5	6.8	8.3	9.7	10.7
Mean_S	5.5	5.0	73.0	69.7	93.0	93.0	179.0	175.7
SD_S	–	–	4.0	3.5	5.6	6.6	5.6	7.6
MAE	–	–	1.3	1.0	1.3	2.7	3.3	3.0
MAPE	–	–	1.8	1.4	1.4	2.9	1.9	1.7
MBE	–	–	0.7	0.3	0.7	1.3	0.7	2.3
RMSE	–	–	1.4	1.0	1.4	3.2	3.5	3.4
d-Stat	–	–	0.93	0.96	0.98	0.93	0.92	0.95

	Leaf area index		Plant height (m)		Seed cotton yield (kg ha−1)		Above ground biomass (kg ha−1)
Growth and yield attributes
Mean_O	3.26	4.55	1.62	1.38	3,413	2,707	9,812	8,974
SD_O	0.73	0.32	0.24	0.22	564.0	637.0	2,779	2,489
Mean_S	3.27	4.52	1.63	1.43	3,424	2,814	10,141	8,804
SD_S	0.89	0.37	0.33	0.26	656.9	476.2	3,084	2,008
MAE	0.13	0.10	0.06	0.08	81.8	175.8	403.2	416.5
MAPE	4.0	2.2	3.7	5.8	2.4	6.5	4.1	4.6
MBE	0.01	−0.03	0.02	0.05	10.7	107.2	328.8	−170.4
RMSE	0.14	0.12	0.07	0.09	91.5	183.2	452.86	533.73
d-Stat	0.99	0.95	0.97	0.95	0.99	0.96	0.99	0.98

DOI: 10.7717/peerj.16329/table-8

Note:

Where, DAS, days after sowing; O, observed; S, simulated; SD, standard deviation (±); R², coefficient of determination; MAE, mean absolute error; MBE, mean biased error; RMSE, root mean square error; d-stat, d-statistics.

Discussion

The cotton growth and yield were considerably influenced by the moisture regimes. On an average, the control (i.e., the non-stressed) crop required up to 185 days duration to achieve the physiological maturity, whereas due to enforced maturity, 1-POSI required the minimum duration of 164 days (Table 6). The impacts of water stress on crop performance vary depending upon the growth stages and health of crop. The water stress during seeding stage restricts the emergence and flowering of cotton (Li et al., 2019). Besides, it also curtailed growth and yield potential of cotton over well-watered crop (Figs. 3 and 4). The crop which received sufficient irrigation, resulted in maximum LAI (4.0 in 2018 and 4.8 in 2019), and it was decreased to (3.3–4.6) under 2-POSI and (3.3–4.2) under 1-POSI (Figs. 4B and 4C). When water supply was constrained, as it was in the case of 1-POSI, the expansion in the leaf area was significantly reduced. Like this, taller plants were observed under control (1.75 m), followed by those under 2-POSI (1.48 m) and 1-POSI (1.28 m) (Fig. 3B). These findings are well supported by Zhang et al. (2021), who reported that the moisture stress shortens plant height. During 2018 and 2019, mean observed plant height was 1.62 ± 0.24 m and 1.38 ± 0.22 m, respectively (Fig. 3). The lower rainfall may be responsible for the reduced plant height during the later year (2019), because under the limiting water conditions, the plant height as well as the leaf size and the stem girth are significantly reduced (Khan et al., 2015).

Water stress negatively affects the cellular and metabolic structure of cotton crop. Results of the pooled data analysis (Fig. 3C) further revealed that SCY was substantially greater under control (3,418 kg ha⁻¹), followed by 2-POSI (2,821 kg ha⁻¹), and least under 1-POSI (2,291 kg ha⁻¹). As levels of water stress increased, the accumulation of dry matter tended to decline, and vice versa (Fig. 3D). Dry matter accumulation under stressed environments has been reduced to 7.72 and 6.25 t ha⁻¹ under 1-POSI and 9.12 and 9.54 t ha⁻¹ under 2-POSI for the year 2018 and 2019, respectively, as compared to control (10.95 t ha⁻¹ during 2018 and 11.13 t ha⁻¹ during 2019, respectively). The drastic decrease in the crop biomass simulated by CROPGRO-cotton model for rainfed circumstances (Figs. 4A and 4B) could serve as a guide for formulating plans for growing cotton in arid/semi-arid environments. However, the results of the self irrigation simulation can be used to achieve the potential yield target. Our findings agree with Hejnák et al. (2015) who reported 50% lowered dry matter accumulation in cotton exposed to water stress. Accumulation of reduced biomass under water deficit conditions might be due to inhibited cell enlargement, which could have restricted development of plant structure due to disrupted flow of water from xylem to the nearby cells. Saifullah et al. (2015) also reported retarded shoot growth and lesser flowering in the moisture stressed cotton crop. Consequently, reduced SCY by 32.9% under 1-POSI and 17.0% under 2-POSI over control (3,418 kg ha⁻¹) has been recorded in conformity with Shahzad et al. (2019). Furthermore, between the water stressed treatments, reduced SCY by 18.8% was observed in the crop raised under extreme water stress (1-POSI) over mildly stressed crop (2-POSI).

The DSSAT-CROPGRO-cotton model was able to evaluate the impact of irrigation strategies on the phenological stages, LAI development, plant height progress and SCY (Table 7). Though, LAI and biomass were over predicted during 2019 (negative MBE) but the error lied within ±10% range, hence it was acceptable. Loison et al. (2017) reported RMSE of 5.3% for anthesis, 4.3% for boll opening, 28% for maximum LAI and 25.7% for SCY. However, Arshad et al. (2017) reported that the simulated total dry matter and SCY exhibited a RMSE of 278–573 kg ha⁻¹ and 237–422 kg ha⁻¹, respectively. Whereas in the present study, the highest RMSE for studied phenological stages was 3.5 and 3.4 days during 2018 and 2019, respectively (Table 7). Likewise, RMSE between observed and simulated SCY was 91.5 kg ha⁻¹ during 2018 and 183.2 kg ha⁻¹ during 2019. For simulation of above ground biomass, RMSE has been 453 and 534 kg ha⁻¹ during 2018 and 2019, respectively. Moreover, for the simulation of different growth and yield attributes of the cotton crop, the d-stat always remained higher than 0.92 which confirmed a close association between observed and simulated parameters.

The data on soil water content at 0–30 and 30–60 cm under different irrigation regimes exhibited remarkable differences over time, at both layers (Fig. 5). Due to adequate supply of irrigation, the control plot retained maximum soil water followed by 2-POSI and 1-POSI plots. Since high stressed plots received only single irrigation, it could retain least soil water. During early stages i.e., 10 and 45 DAS, water regimes did not influence the soil water availability as pre-sowing irrigation and 1-POSI was uniformly applied to all the experimental plots. Nevertheless, it was higher in upper layer (0–30 cm) until vegetative growth period (before 80 DAS) but towards reproductive stages, it was more in the lower layer (30–60 cm) due to more canopy coverage and lesser open field exposure. Since the last irrigation to control plots has been applied during the end of September (Table 3) and till that time soil water of the upper layer (0–30 cm) might have been depleted at higher rate than the lower layer (30–60 cm).

Conclusion

In comparison to well-watered conditions, cotton crop exposed to either level of moisture stress recorded poor growth and reduced SCY. The cotton crop supplied with only 1-POSI and 2-POSI correspondingly also lead to mild or high level of moisture stress. The crop under 1-POSI recorded a SCY reduction by 32.9% whereas under 2-POSI, reduction was only 17.0% over well-watered crop (control). For simulation of various phenological, growth and production attributes of the cotton crop, higher d-stat (0.92 to 0.99) confirmed that the DSSAT-CROPGRO-cotton model can be used as a research tool to reduce expensive and laborious field experiments. The model aided information may be immensely useful for developing strategies to tackle the water deficit situations and improving the cotton crop performance under semi-arid conditions across the globe. Nevertheless, the study needs to be validated using time series experiments over large scale.

Supplemental Information