Metadata for "Machine learning enables reconstruction of past fire regimes from charcoal-derived fire intensity and fuel composition"

In ‘AllTempConf.csv’, this presents data in a confusion matrix that reports the predicted test values from the deep learning temperature model. Each value equals the number of predictions made for specific temperature. Diagonal represents the correct classification. In ‘AllSPConf.csv’, this presents data in a confusion matrix reporting the predicted test values of the deep learning species model. Each value equals the number of predictions made for each species. Diagonal represents the correct classification. In ‘Ancient_cut.csv’, this presents reflectance FTIR data collected from ancient charcoal. Each ancient charcoal particle (Column A: Charcoal Particle Number) has a depth associated with it from where in the sediment core it was collected from (Column B: Depth cm), which sieve size it was collected from during charcoal analysis (Column C: Sieve Size), and which scan number the reflectance data was collected from using a gridded pattern centered on each charcoal particle (Column D: Scan Number). To create this ‘cut’ dataset, wavenumber values >3500 and <950 cm−1 were excluded (Columns E-AYA). In ‘Ancient_zstandard.csv’, this presents reflectance FTIR data collected from ancient charcoal. Each ancient charcoal particle (Column A: Charcoal Particle Number) has a depth associated with it from where in the sediment core it was collected from (Column B: Depth cm), which sieve size it was collected from during charcoal analysis (Column C: Sieve Size), and which scan number the reflectance data was collected from using a gridded pattern centered on each charcoal particle (Column D: Scan Number). To create this ‘zstandard’ dataset, the data was truncated to 3500-950 cm−1 and then a baseline correction was performed with a 2nd-degree polynomial baseline subtraction, then the data was smoothed using a Savitzky-Golay filter with a window length of 10 and a polynomial order of 3, and finally normalized via Z-standardization (Columns E-AYA). In ‘Reflectance_reference_cut.csv’, this presents reflectance FTIR data collected from modern reference charcoal. For both the oxygen and nitrogen atmospheric experiments (Column C; Atmosphere), two of each plant tissue were charred (Column D: Sample Number), per plant species (Column E: Species Name), per plant tissue type (Column E: Tissue Type) at each temperature (Column A: Temperature). For each plant tissue sample, four locations (Column F: Location) on the sample were imaged, with 16 scans (Column G: Scan Number) taken at that specified location in a gridded pattern across the 250 x 250 um area. To create this ‘cut’ dataset, wavenumber values >3500 and <950 cm−1 were excluded (Columns H-AYD). In ‘Reflectance_reference_zstandard.csv’, this presents reflectance FTIR data collected from modern reference charcoal. For both the oxygen and nitrogen atmospheric experiments (Column C; Atmosphere), two of each plant tissue were charred (Column D: Sample Number), per plant species (Column E: Species Name), per plant tissue type (Column E: Tissue Type) at each temperature (Column A: Temperature). For each plant tissue sample, four locations (Column F: Location) on the sample were imaged, with 16 scans (Column G: Scan Number) taken at that specified location in a gridded pattern across the 250 x 250 um area. To create this ‘zstandard’ dataset, the data was truncated to 3500-950 cm−1 and then a baseline correction was performed with a 2nd-degree polynomial baseline subtraction, then the data was smoothed using a Savitzky-Golay filter with a window length of 10 and a polynomial order of 3, and finally normalized via Z-standardization (Columns H-AYD). In ‘Transmission_reference_cut.csv’, this presents transmission FTIR data collected from modern reference charcoal. For both the oxygen and nitrogen atmospheric experiments (Column C; Atmosphere), two of each plant tissue were charred (Column D: Sample Number), per plant species (Column E: Species Name), per plant tissue type (Column: E Tissue Type) at each temperature (Column A: Temperature). For each plant tissue sample, four locations (Column F: Location) on the sample were imaged, with 16 scans (Column G: Scan Number) taken at that specified location in a gridded pattern across the 250 x 250 um area. To create this ‘cut’ dataset, wavenumber values >3500 and <950 cm−1 were excluded (Columns H-AYD). In ‘Transmission_reference_zstandard.csv’, this presents transmission FTIR data collected from modern reference charcoal. For both the oxygen and nitrogen atmospheric experiments (Column C; Atmosphere), two of each plant tissue were charred (Column D: Sample Number), per plant species (Column E: Species Name), per plant tissue type (Column E: Tissue Type) at each temperature (Column A: Temperature). For each plant tissue sample, four locations (Column F: Location) on the sample were imaged, with 16 scans (Column G: Scan Number) taken at that specified location in a gridded pattern across the 250 x 250 um area. To create this ‘standard’ dataset, the data was truncated to 3500-950 cm−1 and then a baseline correction was performed with a 2nd-degree polynomial baseline subtraction, then the data was smoothed using a Savitzky-Golay filter with a window length of 10 and a polynomial order of 3, and finally normalized via Z-standardization (Columns H-AYD). This dataset was published under LA-UR-26-24024.