Singh et al. (2025) Integrated trend analysis and meteorological drought forecasting using ANN in the adjacent semi-arid and arid regions

Identification

• Journal: Theoretical and Applied Climatology
• Year: 2025
• Date: 2025-11-25
• Authors: Uttam Singh, Anoop Kumar Mishra, Mahender Choudhary
• DOI: 10.1007/s00704-025-05883-4

Research Groups

• Department of Civil Engineering, Malaviya National Institute of Technology, Jaipur, Rajasthan, India

Short Summary

This study integrated trend analysis and meteorological drought forecasting using an Artificial Neural Network (ANN) model in adjacent semi-arid and arid regions of Rajasthan, India, finding increasing precipitation trends in several periods and a decrease in drought severity with longer time scales.

Objective

• To estimate the modified Mann-Kendall trend and meteorological droughts at different time scales (3, 6, 9, 12, and 18 months) in observed and ANN-forecasted precipitation data across all stations of the semi-arid and arid regions of Rajasthan, India.

Study Configuration

• Spatial Scale: Jaipur and Jodhpur divisions in the northwestern Indian state of Rajasthan. Jaipur division (semi-arid) includes Alwar, Jaipur, Jhunjhunu, Khairtal-Tijara, Kotputli-Behror, Neem-Ka-Thana, and Sikar districts. Jodhpur division (arid) includes Barmer, Jaisalmer, Jalore, Jodhpur, Pali, and Sirohi districts.
• Temporal Scale:
  • Observed precipitation data: 43 years (1980–2023).
  • ANN model training/testing: 35 years (1980–2015).
  • ANN model validation: 7 years (2016–2023).
  • ANN forecasted precipitation: 5 years (2024–2028).

Methodology and Data

• Models used:
  • Modified Mann-Kendall’s (mMK) test for trend analysis.
  • Sen’s slope estimator for trend magnitude.
  • Standardized Precipitation Index (SPI) for meteorological drought assessment (at 3, 6, 9, 12, and 18-month time scales).
  • Artificial Neural Network (ANN) model for precipitation forecasting.
  • Statistical metrics: Root Mean Square Error (RMSE), Pearson's correlation coefficient (r), Coefficient of Determination (R²), Mean Bias Error (MBE).
• Data sources: Monthly station precipitation (rainfall) data from 1980 to 2023 (excluding Dausa district due to inconsistencies) obtained from the Irrigation Department of Rajasthan (https://water.rajasthan.gov.in/wr/#/home/dptHome). Average monthly precipitation time series data for each district was utilized.

Main Results

• Precipitation Trends: Modified Mann-Kendall’s test on observed and ANN-forecasted precipitation data showed increasing trends for March, May, June, summer, and yearly precipitation across most stations, while February generally showed a decreasing trend. Some stations exhibited varying trends in other months and seasons.
• Drought Severity: Standardized Precipitation Index (SPI) analysis revealed that extreme meteorological drought events were most frequent at shorter time scales (e.g., SPI 3), with their frequency and intensity significantly decreasing as the time scale increased (e.g., SPI 18 showed the minimum number of extreme droughts). SPI 12 and SPI 18 were found reliable for estimating general and long-term drought persistence.
• ANN Model Performance: The ANN model demonstrated high reliability with correlation coefficients (R² and Pearson’s r) between 0.95 and 0.98, and Root Mean Square Error (RMSE) between 6.8 mm and 18.4 mm for training, testing, and modeled precipitation across all stations. Minimum mean bias errors between observed and validated precipitation were less than 2 mm for semi-arid and less than 1 mm for arid regions.
• Forecasted Precipitation: ANN-forecasted precipitation (2024–2028) showed a continuation of seasonal peak patterns, generally below observed peaks but within the observed range, with an incremental increase in forecasted precipitation during monsoon and post-monsoon seasons, potentially due to climate change.

Contributions

• Provided an integrated framework for trend analysis and meteorological drought forecasting using ANN models in adjacent semi-arid and arid regions, which are highly vulnerable to climate variability.
• Demonstrated the high accuracy and reliability of the ANN model for precipitation forecasting in these complex hydroclimatic environments.
• Offered crucial insights into future precipitation trends and drought patterns (2024–2028), aiding in proactive water resource planning and drought management strategies.
• Quantified the relationship between drought severity and time scales, emphasizing the higher frequency of extreme droughts at shorter accumulation periods.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Singh2025Integrated,
  author = {Singh, Uttam and Mishra, Anoop Kumar and Choudhary, Mahender},
  title = {Integrated trend analysis and meteorological drought forecasting using ANN in the adjacent semi-arid and arid regions},
  journal = {Theoretical and Applied Climatology},
  year = {2025},
  doi = {10.1007/s00704-025-05883-4},
  url = {https://doi.org/10.1007/s00704-025-05883-4}
}