Code and data for interpretable bounded-tail neural extreme-value regression with dependence-robust inference: daily maximum temperature in Jammu, Kashmir and Ladakh (1979–2025)

Code, input data and a Google Colab notebook that reproduce all figures and tables of the paper by N. Siddiqui and M. A. Khanday (submitted to Stochastic Environmental Research and Risk Assessment), which models daily maximum temperature extremes in Jammu & Kashmir and Ladakh, western Himalaya. The model (IB-EVT) is a neural extreme-value regression with a continuous spliced Gaussian–generalized Pareto distribution, an end-point parameterization of the bounded upper tail, interpretable regime-specific warming effects, and year-cluster sandwich (cluster-robust) inference. The archive also contains the simulation studies and the baseline forecast models used for comparison. Input data cover 14 sites in Jammu & Kashmir and Ladakh, India, 1979–2025. They are derived from Copernicus C3S (ERA5-Land; ERA5 via Open-Meteo), NCEP/NCAR Reanalysis 1 (NOAA PSL), NOAA CPC (Oceanic Niño Index), NOAA PSL (Dipole Mode Index)