This resource includes eight temperature indices used to assess the accuracy of spatiotemporal variability and trends in temperature extremes in Colombia. T
These indices were calculated monthly for the 1997–2016 period.

* Hottest Day (TXx)
* Coldest Night (TNn)
* Diurnal Temperature Range (DTR)
* Mean Temperature (2TM)
* Percentile-Based Threshold Indices, which measure:
Number of days below the 10th percentile (Cold Nights—TN10p and Cold Days—TX10p)
Number of days above the 90th percentile (Warm Nights—TN90p and Warm Days—TX90p)

To estimate these indices, six high-resolution gridded datasets with varying spatial and temporal resolutions were used:

* ERA5 (0.25°)
* ERA5-Land (0.10°)
* AgERA5 (0.10°)
* MSWX (0.10°)
* CHELSA (0.01°)
* CHIRTS (0.05°)

The performance of these temperature-gridded products in calculating climate extremes was compared with observed data from selected ground weather stations for the period 1997–2016.
Initially, long-term climate data for daily minimum (TN) and maximum temperatures (TX) were assessed from 664 and 629 weather stations, respectively. These records, provided by the Colombian Institute of Hydrology, Meteorology, and Environmental Studies (IDEAM), spanned from 1997 to 2016. After filtering out stations with more than 20% missing data, 153 stations remained, which are located at elevations ranging from 0 to 3500 meters above sea level. Most of these stations are situated in the country's interior, particularly in the Andean region (74%), whereas the peripheral areas have limited data availability due to fewer in-situ stations.
To estimate the missing data in the observed monthly temperature indices, the nonlinear principal component analysis (NLPCA) method was employed. This technique has been previously applied to estimate missing data in various contexts, including precipitation series, extreme precipitation indices, and streamflow data on daily, monthly, and seasonal scales. NLPCA is a nonlinear extension of principal component analysis that utilizes Artificial Neural Networks (ANN), specifically employing the inverse NLPCA method for its calculations.