Irrigation demand for fruit trees under a climate change scenario using artificial intelligence

Abstract

Fruit growing, especially in family farming, has a significant income potential in small areas, but climate change is a major challenge. This study aimed to quantify the irrigation requirements for citrus, papaya, mango and passion fruit, in the Vão do Paranã region, Goiás state, Brazil. The climate data encompassed the observed periods from 1961 to 2020 and future scenarios from 2021 to 2100. The irrigation demand was obtained from the daily water balance, while the reference evapotranspiration (ETo) was estimated using the Penman-Monteith method and then compared with an artificial intelligence tool. The future scenarios indicated a higher increase for air temperature and a lower increase for rainfall. The ETo levels raised from 1,528 mm year-1, in 1991-2020, to 1,614-1,656 mm year-1, in 2021-2050. The artificial intelligence performance was limited in the ETo estimation, with a mean absolute error of 0.71 mm day-1 and an “r” value of 0.59, when considering the air temperature as the input variable. For the 2021-2050 period, when compared with 1991-2020, there was an increase in irrigation demand, in which, under the extreme scenario, the citrus demand reached 690 mm year-1 (+11 %); papaya (+10 %) and passion fruit (+5 %) surpassed 800 mm year-1; and mango reached 491 mm year-1 (+14 %). An increase in demand for irrigation was observed, with management alternatives in association with strategies for maximum cultivation area based on water supply being recommended.

KEYWORDS: Climate resilience, water demand, machine learning, future climate scenarios.