Advances in Crop Science and Technology
Open Access

Maize cultivars; Calibration; Validation; Sensitivity; Plant density; Date of planting

Introduction

DSSAT is one best crop simulation software, which is showing best promises in yield estimation and yield prediction over long time enabling decision support to different parties (Abera et al., 2018; Hoogenboom, et al.2012). DSSAT Software package has gone through different progress since its development with only few crops and four crop models (CERES maize, CERES-wheat, SOYGROW, NUTGROW) (Jone et al., 2003; Kaleita et al., 2020) to more than 42 crops and addition of new modules (Kaleita et al., 2008; Sachin et al., 2019; Abayechaw, 2021). Its advancement for use in many operating systems (Thorp et al., 2011), data interchange system and its ability to be integrated with different software (Dzotzi et al., 2013; Thorp et al., 2008) inclusion of new modules  (Kaleita et al., 2020) used  in precision agriculture (Paz et al., 2001a, 2003; Thorp et al., 2008) and with its betterment in the challenges in formatting input and output files (Kaleita et al., 2008) it has become more user friendly, and is well validated for a number of regions and crops (Throp et al., 2008) [1].

In the same line as the progress in DSSAT developers the progress that demand in the users of DSSAT in the agricultural sector should also be in line by evaluating and validating the different new interfaces of DSSAT added from time to time and covering most areas of the agricultural ecosystem while, adjusting crop coefficients of cultivars is needed for its efficient utilization (Hoogmboom et al., 2020) [2]. The demand continues in evaluation and validation of crop parameters for various crop genes by environment interactions so that we can address the different combination of effects from changes occurring on the environment (Abera et al., 2018) and changes occurring by varying different technologies (Thorp et al., 2008) and at the end it will be possible to find ways to modify or optimize the models within DSSAT for our local condition and specific crop (Jing-yi et al., 2012) [3]. 

Environmental changes that arise due to climate change and variability across agro-ecosystems (Abera et al., 2018) as well as the differential potentials of crop cultivars across different agroecologies, technologies used in the management practices are the major causes (Jone et al., 2003) for fluctuations and yield gap problems that are widely seen in most cropping systems and have predicted future impact on the crops productivity (Abera et al., 2018; Mulune et al., 2015) [4]. For example in Ethiopia climate change perdition across time between 2010-2099 showed a decrease in maize yield by more than 24 % at the end of the century. The use of adaptation strategies, such as, the best cultivar of maize and change in the date of planting will have a salvaging effect up to 12% yield reduction during predicted years of 2012 -2040 (Eulenstein et al., 2017). Being at the start of long term climate change predictions this study is done with the objective of Evaluating and validating different cultivars using DSSAT-CSM, while testing the maize cultivar sensitivity based on some anticipated climate change scenarios [5].

Materials and Methods

Description of the Study Site

The experiment was conducted for two seasons in a warm sub-humid lowlands agro-ecology possessing one altitude feature in the region. The major agroecology covering vast area of the region are the warm moist lowlands and warm sub-humid lowlands having distribution in all the Zones and most districts. Assosa district is selected to represent warm moist lowlands agroecology. The study site is geographically located at 34o 31’E longitude and 10o04’N latitude with an altitude 1580 meters and it is approximately 660 km west of the capital, Addis Ababa [6].

Description of the study materials

Five maize varieties adapted to the agroecology, which are high yielding; resistant to disease and recently released varieties, were selected for the study. Based on the selection criteria the five varieties were, Shone (Pioneer), Melkasa6, MH138, BH545 and local variety. Blended NPS and KCl (60% K) fertilizer was used to supply the three major nutrients, nitrogen, phosphorus and potassium and one micro-nutrient which is deficient in the soil of the study site (Table 1) [7].

Table 1: Descriptions of Maize varieties selected for the study

Soil Sample and sampling methods

Soil samples were taken from the whole field at 10 points and from 4 depths (0-20, 20-40, 40-60 and 60-80 cm) before treatment application and from each plot at 3 points diagonally after crop harvest from 30 cm depth of and samples from the similar experimental unit were composited [8].

Soil physical properties like soil texture and soil dry bulk density, accompanied by chemical properties were tested following standard methods, in the Assosa University Soil Lab. Soil texture were determined using density method proposed by Bouyoucos (2003); the dry bulk density were measured by core sampling method of Black (2003); the soil pH (1:2.5) by pH meter (potentiometric analysis) (Jackson, 2003); the percent organic carbon content using wet potassium dichromate oxidation method (Walkley and Black, 2003);  while the exchangeable K was measured by flame photometer; total N by kjeldahl digestion method (Jackson, 2003); and available P by Bray No 1 method (Bray et al., 1945) [9].

Treatments and design of the experiments

The five cultivars of maize SHONE (Pioneer), MELKASA6, MH138, BH545; and one local cultivar that was planted under two nutrient condition one with (NPS and KCl) and the second one without (NPS and KCl) fertilizers. The four cultivars and the local cultivar with two nutrient situations were planted for two seasons as single factor experiment. The six (6) treatments was, planted on plot size of 4.5 m x 4.5 m with plant spacing of 75 cm and 30 cm between rows and between plants on a row, respectively, for all the cultivars. The experiment was laid in RCBD design, with three replications.

Data collected 

Crop data

Data on days to emergence, days to anthesis and silking and days to physiological maturity; and crop data on four important stages was measured to calculate the genetic parameter, like P1, P2, P5, G2, G3 and PHINT at each leaf appearance. P1 is the thermal time from seedling emergence to the end of the juvenile phase (expressed in degree days, oC day, above a base temperature of 8 oC) during which the plant is not responsive to changes in photoperiod. P2 is the extent to which development (expressed as days) is delayed for each hour increase in photoperiod above the longest photoperiod at which development proceeds at a maximum rate (which is considered to be 12.5 h). P5 is the thermal time from silking to physiological maturity (expressed in degree days above a base temperature of 8 oC). G2 is Maximum possible number of kernels per plant. G3 kernel filling rate during the linear grain filling stage and under optimum conditions (mg day−1). And PHINT is the Phyllochron interval; the interval in thermal time (degree days) between successive leaf tip appearances, to record the phyllochron the plants were observed every day starting from emergence until flowering [10].

Growths of maize such as the leaf area index and plant height were recorded at each full appearance of new leaves until the end of leaf growth and start of flowering. Plant samples were selected from the central plant rows for measuring the LAI and the plant height. Yield component data were also measured at physiological maturity, while the grain yield and final dry biomass yield were taken at time of full maturity. All plant parts (leaves, stalk and the husk) were separated dried and summed up for dry biomass yield [11].

Climate and soil data

A 40 year data between (1980-2020) on five climate variables, solar radiation, maximum temperature, minimum temperature, relative humidity and rainfall was collected from Ethiopian meteorology agency of Benishangul Gomez region, however because of high numbers of missing data between (1980-2000) only 20 year data between (2000-2020) was used for the calibration and validation purpose as completeness of climate data is more important than the numbers of years (Hognboom, et al., 2012).

During the period of the experiment in the 2019 the amount of rainfall in the growing period was 967 mm which is slightly lower than the rain fall amount in the growing period of 2020 experiment year 987 mm. The min and max temperature during 2019 was slightly higher than 2020 in the growing period the different climatic variable in the two growing period (Figure 1,2).

Figure 1: The weather condition in the growing period of 2019.

Figure 2: The weather condition in the growing period during 2020.

Soil analysis was made by taking soil samples from four depths of 0–20, 20–40, 40–60, and 60–80 and 80-100 cm and soil physical properties like: texture, dry bulk density, and some soil chemical properties, such as pH, Organic carbon, Total N, available P, available K were taken before and after the experiment [12].

Data analysis

Crop data in the first season, was used to calibrate the CERES maize model of DSSAT software version 4.7.5, while in the second season the crop data were used to validate using the statistic root means squared error (RMSE), Normalized difference RMSE (nRMSE), mean absolute error (MAE), index of agreement (d) of DSSAT (Yang et al., 2014). Then crop growth and yield of maize scenarios were estimated by changing climatic variables that approximate the El-nino periods of historic El nino years, and different adaptation strategies were tested based on changes in the length of growing period and maize varieties using the DSSAT software.

Result and Discussion

Soil Test Results

The soil analysis across five depth showed that the dry bulk density increased downward in the range between 2.00 - 2.20, while the soil pH; organic carbon, total nitrogen and available phosphorus decreased across depth within range between, 5.50-5.30, 2.84-0.62, 0.49-0.12, and 27.0-25.2 respectively, while the available potassium increased downward from 0.5-0.8 (Table 2).

Table 2: Initial soil characteristics of experimental field

Evaluation of calibrated result

Calibration of five maize varieties using the DSSAT Software: the days to anthesis, the days to physiological maturity, grain yield, harvest index, unit grain weight, and kernels number are simulated with acceptable RSME and d-stat values, higher R2 values and within the ranges of crop coefficient limits.

The minimum and maximum DSSAT-CSM crop coefficients and the new calibrated coefficients of the five cultivars of maize are shown. Comparison of the cultivars from the genetic coefficient may show that variety SHONE is the highest in grain filling rate, while it was the second in number of kernels compared to BH545 and MH138, the delay in development due to photo period sensitivity is higher for SHONE and MELKASA6 compared to the other three MH138, BH545 and L_ASOSA (Table 3).

Table 3: CSM-CERES cultivar coefficient limits and maize cultivars coefficient

The normalized RMSE (nRMSE) for each cultivar based on the five parameters found in the rules file was zero(0) for anthesis, maturity date and grain yield and it ranged between zero(0) and 20 for leaf numbers and biomass yield as calculated with in DSSAT internal statistics and taken during calibration. The values indicated that the former three parameters were estimated as excellent, while it was very good and good for next two parameters. It is suggested that if the nRMSE is less than 10% the simulation is considered excellent; good if it is between 10 and 20%, fair if it is between 20 and 30% otherwise poor if greater than 30% (Jamieson et al., 1991) (Table 4).

Table 4:  Normalized RMSE obtained from DSSAT during calibration

The statistics in the evaluation output for all measured crop parameters high values in the R2, d-stat values, while lower values in the RMSE values support the same, which is sufficient evidence that the genetic coefficients can represent the characteristics of cultivars in target (Table 5) [13].

Table 5: RMSE, R2 and d-stat values used to evaluate simulated maize in 2019

Evaluation of the measured crop parameters during the 2019 between the observed and simulated value cross the origin of the coordinate plane with the indication that there is similarity between the observed and simulated values (Figure 3).

Figure 3: Evaluation of Observed vs. Simulated values of all crop parameters (days to emergence, canopy height, leaf number, leaf area index, days to physiological maturity, dry biomass yield, grain yield, number of grains per ear, unit grain weight).

The normalized RMSE (nRMSE) difference in percent between observed and simulated was less than 10% for pheology maize: days to anthesis (0.024%), days to physiological maturity (0.00%), days to emergency (2.38%, ) while growth and yield of maize: leaf number (5.57%), LAI 7.97%, grain weight (2.22%), grain yield at maturity (5.46%) and top weights/dry biomass yield (6.16%), but harvest index showed high difference (22.45%) compared to all others during calibration year (2019). During the validation year (2020) the phenology of maize: days to emergence (2.38%), days to anthesis (2.29%), physiological maturity (0.026%); while the growth and yield of maize: the leaf number (4.41%), the LAI (16.57%), the grain weight (5.17%), grain yield at maturity (2.15%), top weight/dry biomass yield (9.37%), in similar manner the harvest index (14.33%) showed a greater than 10% variability as in the calibration year. In both calibration and validation year all parameter showed less than 10% different except for harvest index in the first year and LAI and harvest index in the second year.

The nRMSE comparison between Calibrated vs. Validated show that values for validated year are higher in six measured parameters among nine. This can be an indication that better estimation was done during the calibration than validation year hence. Improving high values seen in the harvest index during the calibration year by improving the genetic coefficient could improve the simulation during validation. The R2 and d-stat values are greater than 0.70 for at least four parameter in the calibration year (2019) compared to the validation year (2020) which showed less values for most parameters, exhibiting variability in yield estimation.

Phenology, grow and yield between Observed vs. Simulated 

The calibration results of 2019 showed that the observed and simulated values of maize cultivars are very near to each other with ratio of near 1, for at least four measured values of days to emergence, anthesis day, physiological maturity and the dry biomass yield for which they are simulated below the observed values, except days to emergence, while grain yield and number of leaves are also have ration near to 1, for which they are simulated above the observed values. The unit grain weight at maturity showed nearly equal values for two cultivars SHONE (0.335 and 0.343 g) AND MELKASA6 (0.239 and 0.229 g) for observed vs. simulated values, but there were slight variability for the other three varieties. The values presented in table are for cultivars which are treated the same, but the second cultivar L_ASOSA, was treated differently (no fertilizer) compared to L_ASOSA in the table and the simulation values are not presented here. The grain number at maturity, the dry biomass yield, LAI, HI and leaf number at maturity are not very well simulated for some cultivars than the others; for example, grain number per shoot is simulated well for MH138 (7 grains above), BH 545 (20 grains above) and L-¬_ASOSA (15 grains below) measured data. The LAI is 0.7 lower that the measured value for MELKASA6, while higher for other cultivars. The leaf number at maturity matched between observed and simulated values for the cultivar SHONE, while not good for others. The top weight is among poorly simulated character for all cultivar in both evaluation and validation years. There is much similarity in simulation between the evaluation and validation years, however in validation year the phenology and yield estimation were lower for some cultivars like MELKASA6, which also showed simulation of grain yield much below the observed value (Table 6,7).

Table 6: Measured vs. simulated values (calibration) of the six cultivars of maize during 2019

Table 7: Measured vs. simulated values (validation) of six cultivars of maize during 2020

Sensitivity analysis on the planting date and plant spacing

Date of Planting

Changing time of planting known in the area from June 1-15 to a different date of planting between May 1 to June 30 showed a change both in phenology and yield of maize. When the planting date is shifted 1 month before the known time sowing in the site June 1-15 the phenology as well as the yield of maize decreased. Increasing the panting date between May 10 to May 20 the phenology and the yield increased compared to May 1 planting, but without difference between the two days of planting (May 10 and May 20), however, planting date when shifted to June 24 the phenology are approximately similar to the known planting date of the site, but the grain yield and dry biomass decreased for both calibrated and validated simulations in 2019 and 2020, data presented here is only validation year. Therefore, the time of planting of the area could still be as appropriate using the DSSAT model, but with changes in the climate changes or fluctuation the panting date between May 10 to end of May could be beneficial during the occurrence of short rains due to El Nion (Table 8,9).

Table 8:  Sensitivity analysis in planting date simulated on maize phenology for 2020

Table 9: Sensitivity analysis in planting date simulated on yield for 2020

Plant population (plant spacing)

The plant spacing was made for the corresponding plant population without changing on row spacing, but making change only on plant spacing (Table 10).

Table 10: Plant population per meter square and its corresponding plant spacing

Changing the plant spacing (plant desity), from 4.4 plants m-2 to 5.3 and up to 12 plants m-2 consciously increased the grain yield and biomass yield without change on the phenology of maize (days to anthesis, days to physiological maturity). However, increasing the plant density to 14 plants m-2 decreased the grain, but kept increasing the dry biomass yield of maize cultivars, during both evaluation years (2019) and validation year (2020) data presented here is only validation year (Table 11,12) [14].

Table 11: Sensitivity analysis in population density simulated on phenology for 2020

Table 12: Sensitivity analysis in population density simulated on yield for 2020

Conclutions

The R2, RMSE and d-stat values are used for comparison for the maize cultivars responses. During the Calibration year the R2 values are higher for five responses of the maize cultivars with higher correlation coefficient and r values, the RMSE and Index of agreement (d) value are within the acceptable range for five responses: Anthesis, Physiological maturity, Unit grain weight, Dry biomass yield, Grain yield at maturity, for both calibrated and validated years (2019 and 2020). The calibration result genetic coefficient during 2019 showed that the observed and simulated values of maize cultivars are similar with ratio of near 1 to most of measured parameter. The validation result showed similar trend as the calibration result, but with higher RMSE, and lower d-stat values and with comparatively higher difference between observed and simulated values compared to the calibration result.

Sensitivity test with different time of planting scenario different from June 1-15 to May 1 through to June 30 showed a change both in phenology and yield of maize. When the planting date is shifted 1 month before the known time of sowing in the site June 1-15 the phenology as well as the yield of maize decreased, but increasing the panting date between May 10 to May 20 the phenology and the yield increased compared to May 1 planting, but without difference between the two days of planting (May 10 and May 20). Changing the planting density between 4.4 plant per m2 to 12 plant per m2 could also be advantageous by predicting and mutching with the future climate condition of the site to increase grain yield, but increasing plant density above 14 plants per m2 could increase the dry biomass yield but not the grain yield.

References

1. Abayechaw D (2021) Review of the Decision Support System for Agrotechnology Transfer (DSSAT) Model. J Agrotechnology 10: 236.

Google Scholar                    

2. Abera K, Crespo O, Seid J, and Mequanent F (2018) Simulating the impact of climate change Central Rift Valley of Ethiopia. Journal of Climatology & Weather Forecasting 9: 291.

Indexed at, Google Scholar, Crossref

3. Dzotsi KA, Basso B., Jones W (2013) Development, uncertainty and sensitivity analysis of the simple SALUS crop model in DSSAT. J Ecological Modeling 260: 62-76.

Google Scholar                    

4. Eulenstein F, Lana MA, Schlindwein SL, Sheudzhen AK, Tauscke M,  et al. (2017) Regionalization of Maize Responses to Climate Change Scenarios, N Use Efficiency and Adaptation Strategies. J Horticulturae 3: 9.

Indexed at, Google Scholar, Crossref

5. Hoogenboom G,  Porter CH,  Boote KJ,  Shelia V,   Wilkens PW, et al. (2020) In Boot K. (ed.). Advances in crop modelling for a sustainable agriculture. Burleigh Dodds Science Publishing, Cambridge, UK,  978 1 78676: 240.

Google Scholar                    

6. Jing-yi Y, Hai-long L, Ping H, You-lu B, Ji-yun J,  et al. (2012) Optimizing Parameters of CSM-CERES-Maize Model to Improve Simulation Performance of Maize Growth and Nitrogen Uptake in Northeast China. Journal of Integrative Agriculture 11: 1898-1913.

Google Scholar                    

7. Jones JW,  Hoogenboom G,  Porter CH,  Boote KJ, Batchelor WD, et al. (2003) The DSSAT cropping system model. European Journal of Agronomy 18: 235- /265.

Indexed at, Google Scholar, Crossref

8. Muluneh A, Birhanu B, Stroosnijder L, Bewket W, Keesstra S, et al.  (2015) Impact of predicted changes in rainfall and atmospheric carbon dioxide on maize and wheat yields in the Central Rift Valley of Ethiopia. J Reg Environ Change. on maize production in Ethiopia, East Africa. J Environmental System Research 7:4.

Indexed at, Google Scholar, Crossref

9. Paz JO, Batchelor WD, Jones JW (2003) Estimating potential economic return for variable rate soybean variety management. Trans. ASAE 46: 1225-1234.

Indexed at, Google Scholar, Crossref

10. Paz JO, Batchelor WD, Tylka GL (2001b) Estimating potential economic return for variable rate management in soybeans. Trans. ASAE 44: 1335-1341.

Google Scholar                    

11. Sachin G, Mohammed AJ, Nagajothi K, Mamta R, Murugan BS, et al.  (2019) Automation of   the DSSAT Crop Growth Simulation Model. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W6, 2019 ISPRS GEOGLAM-ISRS Joint Int. Workshop on Earth Observations for Agricultural Monitoring 18-20.

Indexed at, Google Scholar, Crossref

12. Thorp KR, White JW, Porter CH.,  Hoogenboom G,  Nearing GS, et al. (2012)  Methodology to evaluate the performance of simulation models for alternative compiler and operating system configurations. J Computers and Electronics in Agriculture 81: 62-71.

Google Scholar                    

13. Thorpa KR, DeJonge KC, Kaleita AL, Batchelo WD, Paze JO, et al. (2008) Methodology for the use of DSSAT models for precision agriculture decision support. Computers and electronics in agriculture 64: 276-285.

Google Scholar                    

14. Yang JM, Yang JY, Liu S, Hoogenboom G (2014) An evaluation of the statistical methods for testing the performance of crop models with observed data. J Agricultural System 127: 81-89.

Indexed at, Google Scholar, Crossref