Statistical downscaling and projection of climatic extremes using machine learning algorithms

• Published in: Theoretical and applied climatology Vol. 153; no. 3-4; pp. 1033 - 1047
• Main Authors: Maqsood, Junaid, Afzaal, Hassan, Farooque, Aitazaz A., Abbas, Farhat, Wang, Xander, Esau, Travis
• Format: Journal Article
• Language: English
• Published: Vienna Springer Vienna 01.08.2023; Springer; Springer Nature B.V
• Subjects: Agriculture; Algorithms; Analysis; Annual precipitation; Aquatic Pollution; Atmospheric Protection/Air Quality Control/Air Pollution; Atmospheric Sciences; Climate change; Climate science; Climate variability; Climate variations; Climatic extremes; Climatology; Daily precipitation; Data mining; Earth and Environmental Science; Earth Sciences; Environmental impact; Extreme weather; Global temperature changes; Growing season; Heavy precipitation; Heavy rainfall; Machine learning; Mathematical models; Maximum temperatures; Minimum temperatures; Multilayer perceptrons; Original Paper; Potatoes; Precipitation; Precipitation days; Rainfall intensity; Scaling; Seasons; Statistical analysis; Statistical models; Support vector machines; Temperature; Waste Water Technology; Water Management; Water Pollution Control; Weather; Weather stations; Wet days
• ISSN: 0177-798X; 1434-4483
• DOI: 10.1007/s00704-023-04532-y

Climate change impacts all fields of life including agriculture. This study aimed to determine the historical and future climatic variations for the rainfed Prince Edward Island (PEI). Statistical downscaling model (SDSM), and support vector regression (SVR), multilayer perceptron (MLP), and random forest (RF) algorithms were applied to downscale climatic extremes, i.e., daily precipitation, maximum temperature (Tmax), and minimum temperature (Tmin) at 8 meteorological stations across the island for the baseline period (1976–2003). The MLP algorithm was further applied to project the climatic extremes for the future period (2006–2100) under three representative concentration pathways (RCP 2.6, RCP 4.5, and RCP 8.5) due to its better performance. Linear scaling was used to reduce the biases from the outputs of MLP. The annual and seasonal (potato growing season of May to October) outputs revealed that Tmax and Tmin are expected to increase in the future under all the RCPs, with the maximum increment observed for RCP 8.5. The increments in Tmax and Tmin for the growing season were 0.72–5.37 °C and 0.87–5.91 °C, respectively, irrespective of the RCPs. The spatial pattern of average annual precipitation in the growing season showed high (578–966 mm), moderate (558–625 mm), and low (449–664 mm) precipitation at the eastern, central, and western parts of PEI for both baseline and future periods. The highest changes were observed under RCP 8.5 as the warmest climate associated with this scenario. The projected precipitation extreme indices trends are likely to increase in the future. The maximum changes/year were observed under RCP8.5, which are 1.20 days/year for days with heavy precipitation (R10mm), 2.44 days/year for the days with very heavy precipitation (R20mm), 7.60 mm/year for total precipitation from heavy rainy days (R95p), 3.76 mm/year for total precipitation from very heavy precipitation days (R99p), 1.10 days/year for continuous wet days (CWD), and 0.08 mm/day for precipitation intensity (SDII) for a year. The findings of this study will help the farmers and government policymakers to get a clear picture of the climatic variability and strategize to mitigate the climate change impact on the island’s agriculture in the future.