Mthembu et al. (2025) Enhancing AquaCrop model precision for accurate simulation of sweet potato and taro landraces

Identification

• Journal: Frontiers in Sustainable Food Systems
• Year: 2025
• Date: 2025-11-03
• Authors: Thando Lwandile Mthembu, Richard Kunz, Tafadzwanashe Mabhaudhi, Shaeden Gokool
• DOI: 10.3389/fsufs.2025.1698211

Research Groups

• Centre for Water Resources Research, School of Agricultural, Earth and Environmental Science, University of KwaZulu-Natal, Pietermaritzburg, South Africa
• Centre for Transformative Agricultural and Food Systems, School of Agricultural, Earth & Environmental Sciences, University of KwaZulu-Natal, Pietermaritzburg, South Africa
• Centre on Climate Change and Planetary Health, London School of Hygiene & Tropical Medicine, London, United Kingdom
• United Nations University, Institute for Water, Environment and Health (UNU-INWEH), Richmond Hill, ON, Canada

Short Summary

This study recalibrated and validated the AquaCrop model for orange-fleshed sweet potato (OFSP) and taro using multi-location datasets in South Africa. Recalibration significantly improved simulations of canopy cover, biomass, and yield, demonstrating AquaCrop's reliability for these neglected and underutilised crops under non-stressed conditions, though performance declined under severe water limitation.

Objective

• To recalibrate and validate the AquaCrop model for orange-fleshed sweet potato (OFSP) and taro using secondary datasets from multiple locations to improve its predictive accuracy for growth and yield.

Study Configuration

• Spatial Scale: Field and greenhouse experiments across five locations in South Africa: Pietermaritzburg (University of KwaZulu-Natal, UKZN, KwaZulu-Natal province), Roodeplaat (Agricultural Research Council, ARC, Gauteng province), Fountainhill (KwaZulu-Natal province), Swayimane (KwaZulu-Natal province), and Ukulinga (UKZN research farm, KwaZulu-Natal province).
• Temporal Scale: Data collected from growing seasons spanning 2010/11 to 2023/24. Specific experiments include 2022/23 (OFSP, UKZN), 2010/11 (Taro, ARC), 2021/22 (OFSP & Taro, Fountainhill), 2023/24 (OFSP & Taro, Swayimane), and 2010/11 (Taro, Ukulinga).

Methodology and Data

• Models used: AquaCrop v7.1 (FAO crop model).
• Data sources:
  • Field and greenhouse experimental data: Measured canopy cover (CC), above-ground biomass, harvestable yield, and harvest index (HI) for OFSP and taro under various water treatments (unstressed, moderately stressed, stressed, rainfed, optimally irrigated).
  • Climate data: Daily rainfall, maximum and minimum air temperature, and reference evapotranspiration (ETO) from automatic weather stations (AWS) at or near experimental sites. Atmospheric carbon dioxide (CO2) levels from AquaCrop's default MaunaLoa.CO2 file.
  • Soil data: Soil water content at permanent wilting point (PWP), field capacity (FC), and saturation (SAT), total available water (TAW), saturated hydraulic conductivity (KSAT), and soil textural class. Determined using hydrometer method, controlled outflow pressure apparatus, and constant-head permeameter method.
  • Management data: Irrigation schedules (seasonal totals up to 385 mm), planting dates, planting densities (55,556 plants per hectare for OFSP, 20,000 plants per hectare for taro), fertilisation, weed control, and pest/disease management.
  • Crop parameters: Initial parameters sourced from existing literature (Rankine et al., 2015 for sweet potato; Mabhaudhi et al., 2014b for taro).
  • Recalibration and Validation: Recalibration performed using data from UKZN (OFSP) and ARC (Taro). Validation performed using independent datasets from Swayimane, Fountainhill (both crops), and Ukulinga (taro). AquaCrop was run in Growing Degree-Day (GDD) mode for validation.

Main Results

• Improved Recalibration: Recalibration involved adjusting 21 parameters for OFSP and 19 for taro, including reducing taro's maximum rooting depth (to 0.40 m), modifying soil water depletion thresholds, and parameterising phenology based on tuber mass stabilisation.
• Canopy Cover (CC) Simulation:
  • OFSP: R² up to 0.954, Nash-Sutcliffe efficiency (NSE) up to 0.880, and absolute deviations for final yield ≤ 6% under optimal irrigation.
  • Taro: R² up to 0.632, NSE up to 0.485 under unstressed/moderately stressed conditions, but performance declined under severe water stress (negative NSE).
• Above-ground Biomass and Yield Simulation:
  • High accuracy for both crops under unstressed conditions (R² > 0.98, NSE > 0.95 for taro).
  • Absolute deviations for final biomass and yield were ≤ 6% under optimal irrigation for both crops.
  • Simulated high HI values (76–88%) for both crops.
• Validation Performance:
  • AquaCrop reliably simulated growth and yield under non-stressed conditions (e.g., Swayimane with high rainfall, R² = 0.999 for taro CC, NSE = 0.989).
  • Performance declined under water-limited environments, particularly for taro CC and OFSP biomass at Fountainhill, indicating limitations in simulating severe water stress.
  • GDD mode improved model transferability across different agro-ecological zones.

Contributions

• Provides the first comprehensive recalibration and validation of the AquaCrop model for orange-fleshed sweet potato and taro using multi-location datasets, significantly enhancing predictive accuracy compared to previous parameterizations.
• Bridges a critical modeling gap for neglected and underutilised crop species (NUS), supporting their integration into climate adaptation strategies and promoting resilient agricultural diversification.
• Enables more accurate water management, operational yield predictions, and climate risk assessments for both smallholder and commercial farmers cultivating these nutrient-dense, climate-resilient crops.
• Demonstrates that targeted recalibration can achieve predictive performance for root and tuber crops (RTCs) on par with or better than conventional staple crops, strengthening the case for mainstreaming NUS in crop modeling.
• Emphasizes the importance of using Growing Degree-Day (GDD) mode for improved phenological representation and model transferability across variable climates, crucial for future climate change impact assessments.

Funding

• Water Research Commission (WRC), Pretoria, South Africa (WRC Project C2023/24-0254: "Developing a database and utility tool for underutilised indigenous crops for increased agricultural diversification in South Africa.")
• National Research Foundation (NRF), Pretoria, South Africa (NRF Grant PMDS23041994670)
• London School of Hygiene & Tropical Medicine, University of London

Citation

@article{Mthembu2025Enhancing,
  author = {Mthembu, Thando Lwandile and Kunz, Richard and Mabhaudhi, Tafadzwanashe and Gokool, Shaeden},
  title = {Enhancing AquaCrop model precision for accurate simulation of sweet potato and taro landraces},
  journal = {Frontiers in Sustainable Food Systems},
  year = {2025},
  doi = {10.3389/fsufs.2025.1698211},
  url = {https://doi.org/10.3389/fsufs.2025.1698211}
}