a global in-situ daily temperature dataset preferentially for 0000-2400 UTC from 1981-2024

Large amounts of sub-daily temperature data are shared globally through the Global Telecommunication System (GTS) in near-real time and through international data exchanges. However, converting these data into a global daily temperature dataset with a uniform definition—especially for daily maximum (Tmax) and minimum (Tmin) temperatures—has proven challenging due to the independent observation schedules across the world. To address this issue, we developed a new method that decomposes sub-daily Tmax and Tmin records from the Integrated Surface Database (ISD) into finer intervals, subsequently reaggregating them into daily Tmax and Tmin based on prospective dateline. This new method increased the global daily Tmax and Tmin counts by 64% and 45%, respectively, compared to the original method, which relied on either two consecutive Tmax/Tmin records over 12 hours or a single Tmax/Tmin record over 24 hours. The Global Land Base Dataset-First Estimate Daily Data (GLBD-FED) was established for the period from 1981 to 2024, following corrections for misrecorded Tmax and Tmin and quality control. GLBD-FED includes daily maximum (Tmax), average (Tave), and minimum temperature (Tmin) from approximately 17,000 global sites, with daily data amounts reaching about 10,000 entries per day in the current decade. When compared to the Global Summary of the Day (GSOD) dataset, GLBD-FED exhibits more temperate performance over the last 40 years, showing a slightly lower daily Tmax (around -0.3°C) and a higher daily Tmin (around +0.3°C), with a nearly identical daily Tave (approximately +0.1°C). GLBD-FED identifies records that could or almost could represent the highest or lowest temperature in the 24-hour period as daily Tmax or Tmin, while GSOD selects the highest or lowest records within the 24 hours. This variance in definitions, combined with different preferences for meteorological and airport report sources, contributes to these observed biases.