mrtag08/thane-weather-dataset-2022-2026

Thane Daily Microclimate & Weather Dataset (2022–2026)

An unbroken, single-station daily meteorological time series recorded at ~10:00 AM each morning in Thane, Maharashtra, India ($19.22^\circ\text{ N}, 72.98^\circ\text{ E}$) spanning March 17, 2022 to September 19, 2026 (1,467 observations).

[image]

Key Highlights & Characteristics

1. 1.Continuous 4.5-Year Horizon: Captures 4 complete Southwest Monsoon cycles, 5 summer pre-monsoon heatwave spikes, and 4 winter atmospheric inversion periods without missing numerical values.
2. 2.Fixed-Time Sampling: Every record was sampled systematically at approximately 10:00 AM local time, eliminating diurnal time-of-day sampling variance.
3. 3.Organic Air Quality Proxy: Includes granular tracking of atmospheric visibility/conditions (Haze, Smoke, Mist, Fog, Clear), providing an empirical proxy for MMR's winter particulate matter / inversion smog.
4. 4.Dual Units & Enriched Features: Retains metric and imperial values, along with engineered features such as diurnal temperature delta, calculated day length, meteorological season, and precipitation indicators.

Visual Summary

Dataset Schema

Quickstart

Using pandas

import pandas as pd

# Load CSV directly
url = "https://huggingface.co/datasets/mrtag08/thane-weather-dataset-2022-2026/raw/main/data/thane_weather_2022_2026.csv"
df = pd.read_csv(url, parse_dates=['date'])
print(df.head())

Using Hugging Face datasets

from datasets import load_dataset

dataset = load_dataset("mrtag08/thane-weather-dataset-2022-2026")
df = dataset['train'].to_pandas()

Applications & Use Cases

1. 1.Time-Series Forecasting: Benchmark deep learning architectures (e.g. PatchTST, TimesFM, DLinear, LSTM) and tree models (XGBoost/LightGBM) to forecast next-day maximum temperature and diurnal spread.
2. 2.Monsoon & Rain Classification: Predict wet days and rain events using humidity, wind direction shifts, and pressure precursors.
3. 3.Environmental Inversion & Smog Tracking: Model the transition between maritime Arabian Sea air masses and continental winter haze/smoke inversions.

Provenance & Attribution

• —Collection Method: Automated daily trigger via IFTTT logging Weather Underground / The Weather Channel feed.
• —Location: Thane, Maharashtra, India.
• —License: Creative Commons Attribution 4.0 International (CC-BY-4.0).