Trend analysis and forecasting of meteorological variables in the lower Thoubal river watershed, India using non-parametrical approach and machine learning models
Climate change, variability and their impact assessment are major concerns of the scientific community across the world. Changes and variations in meteorological variables have caused deleterious effects on water, agriculture, and forests globally. Manipur is a high rainfall deficit state in India....
Gespeichert in:
Veröffentlicht in: | Modeling earth systems and environment 2024-02, Vol.10 (1), p.551-577 |
---|---|
Hauptverfasser: | , , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Climate change, variability and their impact assessment are major concerns of the scientific community across the world. Changes and variations in meteorological variables have caused deleterious effects on water, agriculture, and forests globally. Manipur is a high rainfall deficit state in India. Therefore, the lower Thoubal River watershed is highly sensitive to minor climatic variations, which may significantly affect the socio-economic conditions of around 54% of the total population depending on agricultural activities. Hence, it has become imperative to analyze past trends of climate and ascertain future scenarios. Several researchers have investigated climatic variations; however, the existing literature has paid less attention to micro-level variations. To address this gap, the present study attempts to quantify temperature and precipitation trends in the lower Thoubal river watershed during 1981–2020 using daily gridded meteorological data. Sen's slope estimator was used to quantify the rate of change in rainfall and temperature, and the Mann–Kendall (MK) test was utilized to examine the direction of change and significance level. The study also provides a new insight to forecast climate scenarios in the watershed during 2021–2030 using two machine learning algorithms: random forest and artificial neural network-multilayer perceptron (ANN-MLP). Three statistical performance assessors and coefficient of determination (R
2
) were used to select the best forecasting model. The trend analysis results revealed a declining trend of rainfall at the rate of 10.30 mm/year with high variability. The annual maximum, minimum, and mean temperatures, as well as the diurnal temperature range (DTR), have also exhibited a statistically significant increasing trend, with rates of change at 0.035 °C, 0.01 °C, 0.025 °C, and 0.017 °C/year, respectively. The seasonal forecasting result indicate increase in temperature and decrease in rainfall were anticipated for the next 10 years. The random forest model has proved effective for forecasting of meteorological variables in micro-scale level. Such a trend will likely affect the agricultural productivity, streamflow and flooding, groundwater recharge, vegetation cover and water supply in the watershed. The findings of the study will be helpful for the local community and policy makers for management of natural resources in the watershed. The methodology adopted in the study could be expanded for other geographical region |
---|---|
ISSN: | 2363-6203 2363-6211 |
DOI: | 10.1007/s40808-023-01799-y |