Interactive effects of n fertilization rate, cultivars and planting date under climate change on maize (zea mays l.) yield using crop simulation and statistical downscaling of climate models
Loading...
Date
2019
Authors
Chisanga, Charles Bwalya
Journal Title
Journal ISSN
Volume Title
Publisher
University of Zambia
Abstract
Generating new information using traditional agricultural trials which are expensive is not sufficient to meeting novel agro-technologies. Planting date (PD), low soil fertility and climate change influences maize (Zea mays L.) growth and yield. Local-scale impacts of future climate change (CC) and variability (CV) on PD, N fertilizer rate (N), cultivar, maize growth, and yield is not well documented in Zambia. The impact of climate change and extreme climate indices on maize yield in AERII are carried out at large spatial scales, missing out on local-scale impacts, mitigation and adaptation potentials under which farmers operate.
Statistical downscaling models such as stochastic weather generator (Long Ashton Research Station Weather Generator [LARS-WG]) and delta-based methods (Agricultural Model Intercomparison and Improvement Project (AgMIP) protocols) have not been applied locally to assess the impact of climate change. Additionally, the AgMIP protocols have not been applied to downscale climate scenarios using Representative Concentration Paths (RCP4.5 and RCP8.5). Changes in temperature and precipitation would have a significant impact on maize phenology and yield. Two field experiments were conducted at Zambia Agriculture Research Institute at Mount Makulu (Lat: 15.550o S, Lon: 28.250o E, altitude: 1213 m) in Zambia to assess the effect of PD, N and cultivars on yield and yield parameters and to predict the impact of climate change on maize productivity in the 2050s. The irrigated experiment was arranged in a Split-plot design with maize cultivars (ZMS606, PHB30G19and PHB30B50) and N fertilizer rates (67.20, 134.40 and 201.60 kg N ha-1) as main-plots and sub-plots, respectively. These cultivars were selected as major cultivars planted by small scale farmers and their long commercial life. The rainfed experiment was a split-split plot design with PDs, cultivars, and N as the main-plots, sub-plot, and sub-subplots, respectively. Each field experiment had three replicates.
Daily weather data were obtained from the Zambia Meteorological Department and AgMERRA. Plant growth, grain, and biomass yield were observed at phenological stages. The ANOVA for grain yield and yield parameters were computed using the sp.plot and ssp.plot functions in Agricolae R package. Significant differences between means were tested using Fisher-LSD Test (p<0.05). Site weather data, soil data, cultivar characteristic, and management required by the crop models were also collected. The rainfed (2016/2017) and irrigated (2016) field experimental data were used for calibrating and validating the crop simulation models (CSMs), respectively.
Expert Team on Sector Specific Indices (ET-SCI) of extreme temperature and precipitation were computed after checking weather data (1963-2012) for quality, homogeneity, and trends using ClimPACT2. The APSIM-Maize v7.9 and CERES-Maize v4.7 models were calibrated using rainfed experimental site data. Days after planting (DAP) to anthesis, and maturity, grain and biomass yield, LAI and soil water content measurements were used to calibrate the CSMs. Cultivar-specific parameters (CSPs) in DSSAT CERES-Maize model were computed using the generalized likelihood uncertainty estimation (GLUE). The CSPs in APSIM-Maize were computed using a stepwise approach starting with phenology, soil water, soil N, biomass and grain yield. The models were validated using the irrigated field experimental site data. Root mean squared error (RMSE), normalized RMSE, R2, and d-stat were used to evaluate the agreement between simulated and observed values.
v
Climate scenarios were generated using LARS-WG and AgMIP protocols. The suitability of LARS-WG in generating current and future climate scenarios was evaluated using CMIP3 HadCM3 and BCM2 global climate models (GCMs) for B1 and A1B scenarios. The current (1980-2010) and future (2040-2070) climate scenarios used as inputs into the CSMs were generated using 5 GCMs (E, I, K, O, R) and RCPs (RCP4.5 and RCP8.5) scenarios using the AgMIP protocols. APSIM-Maize and CERES-Maize models were used to evaluate maize yield to response to PD, N, cultivar and climate scenarios assuming constant management. Future changes in phenology and yield were estimated as the difference between future and baseline period.
Extreme precipitation (PRCPTOT, R30 mm, RX5day, R95pTOT) indices were statistically non-significant. Mount Makulu warming are due to increase in mean and maximum temperature. LARS-WG projected an increase in temperature (Observed station data [1.50oC (B1: 2050), 1.84oC (A1B: 2050)], AgMERRA data [1.48oC (B1: 2050), 1.84oC (A1B: 2050)]) and variability in precipitation. The projected ensemble mean annual temperature using the AgMIP protocols is expected to increase by 1.82°C (RCP4.5) and 2.48oC (RCP8.5). However, rainfall is projected to decrease by 1.46% (RCP4.5) and 1.91% (RCP8.5). APSIM-Maize and CERES-Maize models simulated fewer DAP to anthesis and maturity relative to the baseline. Maize grain yield (-6.90 - +4.06 (RCP4.5), -10.80 - +5.00% (RCP8.5) [APSIM-Maize]; -0.59 to +25.77% (RCP4.5) and -6.52 to +20.21% (RCP8.5) [CERES-Maize]) would decrease or increase relative to the baseline. PD, N, and rainfall would affect future grain yield.
The simulated versus observed values of DAP to anthesis, and maturity, grain yield, grain size, and grain number m-2 and soil water content had normalized root mean square error < 20% and d-stat > 0.71. The models can be used to predict phenology and yield. Using an ensemble mean of the CSMs, DAP to anthesis (-11.28 to -9.39% [RCP4.5]; -14.28 to -12.65% [RCP8.5]) and maturity (-10.52 to -9.43% [RCP4.5]; -14.01 to -12.75% [RCP8.5]) would reduce in 2050. The % change in grain would range from 2.78 to 9.94%, -3.81 to -8.88% and -2.33 to 10.63% under N1, N, and N3, respectively. Grain yield would increase/or decrease with delay in PD (RCP4.5 [PD1 = 2.57%; PD2=3.31%; PD3=4.37%]; RCP8.5 [PD1 = -1.11%; PD2=-0.29%; PD3=1.08%]). The current PDs and cultivars with lower N (N1) would increase grain yield in future. However, grain yield would increase with higher N (N3) at PD3.
To establish credibility in CSMs for use at a local scale, they have to be adequately calibrated and validated. The calibration and validation of APSIM-Maize and CERES-Maize models were necessary for their application to new cultivars to minimize uncertainty. CSMs can be used to simulate probable outcomes in crop management strategies, N fertilizer rate, PD and impact of climate change on crop growth and yield. Earlier PDs with lower N would lead to an increase in grain yield than at higher N. Proposed improvements in the CSMs are phenological stage duration and LAI prediction under subtropical environments. The mitigation and adaptation strategies for CC includes changing PDs, N management, cultivar selection, water management, and tillage practices. Future model evaluations may be needed for new cultivars.
Key words: AgMERRA, AgMIP, APSIM-Maize, CERES-Maize, climate change, climate indices, climate variability, CSM, delta-based method, GCM, GDD, LARS-WG, nitrogen, planting date, RCPs, statistical downscaling
Description
Keywords
Climate change , Climate variability