EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-SA 4.0

Integrated Drought Diagnostics in Telangana (1981–2023): Trend Analysis, Multi-Index Assessment, Quadrant Framework, and Interpretable Machine Learning

VELUSAMY GUHAN, DHARMA RAJU AKASAPU, NAGARATNA KOPPARTHI

Journal of Agrometeorology · 04.06.2026 · 28 (2) · 167-176

DOI: 10.54386/jam.v28i2.3302 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Agrometeorology, online erschienen am 04.06.2026. Verfasst von VELUSAMY GUHAN und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study aims to provide an integrated diagnostic of drought risk by investigating historical rainfall and temperature variability from 1981–2023 across six drought‑prone districts of Telangana using statistical indices, trend diagnostics, and machine learning approaches. Monthly and annual datasets from IMD gridded archives were processed to compute mean, standard deviation, and coefficient of variation (CV) for rainfall, minimum, and maximum temperatures. Multiple drought indicators, including the Z Score Index (ZSI), Deciles Index (DI), Percent of Normal Index (PNI), Standardized Precipitation Index (SPI), China Z Index (CZI), and Rainfall Anomaly Index (RAI), were applied to capture severity, duration, and spatial extent of droughts. Trend analysis using the Mann–Kendall test and Sen’s slope revealed statistically significant increases in maximum temperatures (+0.03 to +0.06 °C per year), while rainfall showed high variability (CV ranging from 22% in Khammam to 38% in Rangareddy) but no consistent long‑term trend. A quadrant‑based climate stress framework was developed by integrating rainfall magnitude, variability, extremes, and peak maximum temperature, classifying districts into Climate Stable, Rainfall Unpredictable, Dry Stable, and High Risk Climate Stress Zones. To enhance predictive capacity, machine learning models (Random Forest, Gradient Boosting, SVM, and Neural Networks) were trained on rainfall and temperature predictors, with SHAP analysis providing interpretability by identifying key drivers such as rainfall CV, Tmax slope, SPI, and ZSI. Model performance was robust, with Gradient Boosting achieving 89.1% accuracy and Random Forest 87.2%, confirming ensemble methods as the most reliable classifiers. Results confirm that all districts experienced mild to extreme drought years, with SPI identifying 6–9 severe drought years per district and Rangareddy and Mahbubnagar showing the highest risk. The integrated framework, combining statistical indices, visualization, and interpretable machine learning, provides a replicable methodology for semi‑arid regions and offers actionable insights for policymakers to strengthen agricultural resilience, water resource management, and climate adaptation strategies.

Zitieren

APA

GUHAN, V., AKASAPU, D. R., & KOPPARTHI, N. (2026). Integrated Drought Diagnostics in Telangana (1981–2023): Trend Analysis, Multi-Index Assessment, Quadrant Framework, and Interpretable Machine Learning. Journal of Agrometeorology, 28(2), 167-176. https://doi.org/10.54386/jam.v28i2.3302

Vancouver

GUHAN V, AKASAPU DR, KOPPARTHI N. Integrated Drought Diagnostics in Telangana (1981–2023): Trend Analysis, Multi-Index Assessment, Quadrant Framework, and Interpretable Machine Learning. Journal of Agrometeorology. 2026;28(2):167-176. doi:10.54386/jam.v28i2.3302

BibTeX
@article{GUHAN2026,
  author = {GUHAN, VELUSAMY and AKASAPU, DHARMA RAJU and KOPPARTHI, NAGARATNA},
  title = {Integrated Drought Diagnostics in Telangana (1981–2023): Trend Analysis, Multi-Index Assessment, Quadrant Framework, and Interpretable Machine Learning},
  journal = {Journal of Agrometeorology},
  year = {2026},
  volume = {28},
  number = {2},
  pages = {167--176},
  doi = {10.54386/jam.v28i2.3302}
}
Kostenlos starten →