Scholarlink

<p>The coconut palm is a major tropical crop, providing food, fibers, and essential oil for both edible and industrial uses. Estimating its productivity based on agrometeorological variables is a valuable tool for management practices aimed at minimizing climate-related losses and maximizing profits. Traditional estimation methods are time-consuming and costly, whereas statistical and machine learning models can streamline decision-making. This study used both statistical and machine learning models to estimate coconut palm productivity. High and low productivity harvests (April and November, respectively) were analyzed using nine years of field experimental data combined with meteorological data during the fruiting period. Variables included temperature, humidity, precipitation, wind, solar radiation, and radiation balance, along with derived variables such as vapor pressure deficit, evapotranspiration, and soil water deficit and surplus. The models developed were multiple linear regression (MLR) and machine learning algorithms, including multilayer perceptron neural networks (MLP), support vector regression (SVR), and random forest (RF). Performance was evaluated using root mean square error (RMSE), R², and mean absolute error (MAE), with Shapley additive explanations (SHAP) applied for interpretability. Results showed RMSE values of 161–179 fruits ha⁻¹ for the high-productivity season and 79–84 fruits ha⁻¹ for the low-productivity season. Machine learning algorithms outperformed traditional methods, with MLP better suited for high productivity and RF for low productivity conditions. In the high-productivity season, solar radiation and water surplus during the final fruit maturation stage were key variables, while in the low-productivity season, relative humidity and vapor pressure deficit during fruit volume expansion were the most influential. In addition to enabling productivity forecasts months before harvest, the models provide valuable insights to support proactive management strategies in the field—optimizing irrigation, fertilization, and harvest planning—and enhance the resilience of coconut production systems in the face of increasing climatic variability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Estimating the productivity of green Dwarf coconut using Agrometeorological models in Eastern Amazonia

  • 期刊论文

  • 发表日期：

    2025年8月26日

查看全文
  • Maryelle Kleyce Machado Nery,
  • Gabriel Siqueira Tavares Fernandes,
  • João Vitor de Novoa Pinto,
  • Matheus Lima Rua,
  • Breno Rodrigues de Miranda,
  • Paulo Manoel Pontes Lins,
  • Glauco de Souza Rolim,
  • Paulo Jorge de Oliveira Ponte de Souza

摘要

The coconut palm is a major tropical crop, providing food, fibers, and essential oil for both edible and industrial uses. Estimating its productivity based on agrometeorological variables is a valuable tool for management practices aimed at minimizing climate-related losses and maximizing profits. Traditional estimation methods are time-consuming and costly, whereas statistical and machine learning models can streamline decision-making. This study used both statistical and machine learning models to estimate coconut palm productivity. High and low productivity harvests (April and November, respectively) were analyzed using nine years of field experimental data combined with meteorological data during the fruiting period. Variables included temperature, humidity, precipitation, wind, solar radiation, and radiation balance, along with derived variables such as vapor pressure deficit, evapotranspiration, and soil water deficit and surplus. The models developed were multiple linear regression (MLR) and machine learning algorithms, including multilayer perceptron neural networks (MLP), support vector regression (SVR), and random forest (RF). Performance was evaluated using root mean square error (RMSE), R², and mean absolute error (MAE), with Shapley additive explanations (SHAP) applied for interpretability. Results showed RMSE values of 161–179 fruits ha⁻¹ for the high-productivity season and 79–84 fruits ha⁻¹ for the low-productivity season. Machine learning algorithms outperformed traditional methods, with MLP better suited for high productivity and RF for low productivity conditions. In the high-productivity season, solar radiation and water surplus during the final fruit maturation stage were key variables, while in the low-productivity season, relative humidity and vapor pressure deficit during fruit volume expansion were the most influential. In addition to enabling productivity forecasts months before harvest, the models provide valuable insights to support proactive management strategies in the field—optimizing irrigation, fertilization, and harvest planning—and enhance the resilience of coconut production systems in the face of increasing climatic variability.