Addressing water stress in banana-based cropping systems: evaluating the applicability of the AquaCrop computer simulation model for banana plantations
Bananas (Musa spp.) is one of the most produced crop worldwide and the world’s most popular fruit. Water is often considered the largest abiotic threat to production. The impact of water stress on production can be simulated through well calibrated computer crop models to aid in plantation management. Currently, a widely applicable, accurate and functional computer crop model for simulating banana production and underlying soil water balances does not exist. The general aim of this PhD research was to evaluate the promising AquaCrop simulation model for a banana field over the course of two crop cycles.
A field experiment was set up in Arusha, Tanzania using two different cultivars: Mchare - Huti Green Bell (HG) and Cavendish Grande Naine (GN). Growth under optimally (‘full’) irrigated (FI) and water-limited rainfed (RF) conditions was studied to act as a database for AquaCrop modelling. Soil water contents were monitored through time domain reflectometry, and water was applied whenever more then 25% of total available soil moisture (TAW) was about to be depleted. TDR measurements revealed SWC in the FI plots were close to field capacity over the course of the entire field experiment whilst SWC followed the seasonal rainfall in RF plots. Canopy cover (CC), leaf area index (LAI) biomass (B), yield (Y) and time to phenological events (e.g., flowering, harvest) were studied under these differing SWC.
Generally irrigation sped up phenology slightly and improved growth components of a plantation (CC, LAI, B and Y). CC showed to varied diurnally under conditions of high evaporative demand, possibly increased under drought. To be able to calculate CC values based on LAI, cultivar-specific CC-LAI curves were created following the Lambert-Beer equation. Curves were different between cultivars.
Weighted least square regression models were built for (1) estimating aboveground vegetative dry biomass (ABGVD) and corm dry biomass (cormD) and (2) forecasting bunch fresh weight (bunchF), based on non-destructive parameters for our two cultivars, whereby it was checked whether relationships differed under different irrigation regimes. Pseudostem volume (Vpseudo) proved a good predictor for estimating ABGVD, but was suboptimal for cormD estimation due to its subterranean position. ABGVD and cormD models were not different for FI and RF plots. Vpseudo at flowering (Vpseudo,flower) proved to be a good predictor for bunchF in FI plots. Significant differences were observed between cultivars and between treatments indicating allometric plasticity under drought during bunch filling.
The AquaCrop model scheme was adjusted to simulate (1) successive banana crop cycles and (2) drip irrigation, using collected data from GN. Goodness-of-fit indicators used were R², relative root mean square error, Nash and Sutcliffe model performance and Willmott’s index of Agreement. Based on these indicators, FI and RF plots were simulated with good to moderate good accuracy for all parameters: CC, SWC, B and Y. Our research provided first estimates of crop parameter sets for banana in AquaCrop. Successive modelling of banana cycles can be used in initial plantations characterized by a narrow flowering and harvest timeframe. Further validation and evaluation (using different cultivars and irrigation schemes) in multi-cycle and multi-location experiments will further develop AquaCrop for banana, especially under drought and nutrient stress. Further insights into modelling mature plantations (beyond plantation development) were obtained in this modelling exercise.