TEMPO Validation Appendices

Pandora vs TEMPO totHCHO

Barron H. Henderson (US EPA)

Pandora vs TEMPO totHCHO

generated by: Barron H. Henderson
institution: US EPA
last updated: 2025-12-15T01-0500

1: Report Description

This report has plots organized by analysis region for Pandora totHCHO. All data uses the version 3 (v3) TEMPO data products. Except when evaluating the effect of cloud fraction, all TEMPO data has been filtered for effective cloud fraction less than 0.15 based on SAO recommendation at the GeoXO ACX May meeting. All data has also been filtered for high quality and having solar zenith angle less than 70 degrees

Pandora intersections are any TEMPO pixel that overlaps a 0.03-degree radius. Only intersections within 15 minutes of a TEMPO measurement are used. Statistics (mean, std, quantiles) are all based on each plus/minus 15-minute period or pooled data.

TropOMI intersections include both Pandora locations and Nonattainment Areas (NAA). However, the NAA are larger and include more pixels. The NAA have more area in the West, so the time-series including all sites are skewed by the western performance. By comparison, the Pandora are more in the East, so the time-series are any TropOMI pixel that overlaps a TEMPO pixel. Statistics (mean, std, quantiles) are all based on all intersections within the region.

Use section headers to navigate by Headings to a specific analysis region.

2: Location Overview

Figure 2.1 shows the PGN stations used for NO2. Comparisons include only pixels within 0.03 degrees of each station. Only Pandora stations selected for high quality HCHO (solid markers) are included.

All Pandora stations used for NO2 analysis are shown, but only stations with a solid marker are used for HCHO.
Figure 2.1: All Pandora stations used for NO2 analysis are shown, but only stations with a solid marker are used for HCHO.

This section shows plots for all sites

3: All Sites by month

Box (25%, 50%, 75%) and whisker (min/max) plots by month at All Sites.
Figure 3.1: Box (25%, 50%, 75%) and whisker (min/max) plots by month at All Sites.

Scatter plot medians by month at All Sites with linear regression fit. Each point has all pixels in a month and error bars represent the 25% to 75% range of those pixels.
Figure 3.2: Scatter plot medians by month at All Sites with linear regression fit. Each point has all pixels in a month and error bars represent the 25% to 75% range of those pixels.

Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by month. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.
Figure 3.3: Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by month. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.

4: All Sites by week

Box (25%, 50%, 75%) and whisker (min/max) plots by week at All Sites.
Figure 4.1: Box (25%, 50%, 75%) and whisker (min/max) plots by week at All Sites.

Box (25%, 50%, 75%) and whisker (min/max) plots by week at All Sites for average weekday.
Figure 4.2: Box (25%, 50%, 75%) and whisker (min/max) plots by week at All Sites for average weekday.

Box (25%, 50%, 75%) and whisker (min/max) plots by week at All Sites for average weekend.
Figure 4.3: Box (25%, 50%, 75%) and whisker (min/max) plots by week at All Sites for average weekend.

Scatter plot medians by week at All Sites with linear regression fit. Each point has all pixels in a week and error bars represent the 25% to 75% range of those pixels.
Figure 4.4: Scatter plot medians by week at All Sites with linear regression fit. Each point has all pixels in a week and error bars represent the 25% to 75% range of those pixels.

Scatter plot medians by week at All Sites with linear regression fit for weekdays only. Each point has all pixels in a week and error bars represent the 25% to 75% range of those pixels.
Figure 4.5: Scatter plot medians by week at All Sites with linear regression fit for weekdays only. Each point has all pixels in a week and error bars represent the 25% to 75% range of those pixels.

Scatter plot medians by week at All Sites with linear regression fit weekends only. Each point has all pixels in a week and error bars represent the 25% to 75% range of those pixels.
Figure 4.6: Scatter plot medians by week at All Sites with linear regression fit weekends only. Each point has all pixels in a week and error bars represent the 25% to 75% range of those pixels.

5: All Sites by loc

Median by loc at All Sites.
Figure 5.1: Median by loc at All Sites.

Median Bias in #/cm2 by loc at All Sites.
Figure 5.2: Median Bias in #/cm2 by loc at All Sites.

Normalized Median Bias as a percent plots by loc at All Sites.
Figure 5.3: Normalized Median Bias as a percent plots by loc at All Sites.

Mean by loc at All Sites.
Figure 5.4: Mean by loc at All Sites.

Mean Bias in #/cm2 by loc at All Sites.
Figure 5.5: Mean Bias in #/cm2 by loc at All Sites.

Normalized Mean Bias as a percent plots by loc at All Sites.
Figure 5.6: Normalized Mean Bias as a percent plots by loc at All Sites.

Correlation at each loc at All Sites.
Figure 5.7: Correlation at each loc at All Sites.

Box (25%, 50%, 75%) and whisker (min/max) plots by loc at All Sites.
Figure 5.8: Box (25%, 50%, 75%) and whisker (min/max) plots by loc at All Sites.

Scatter plot medians by loc at All Sites with linear regression fit. Each point has all pixels in a loc and error bars represent the 25% to 75% range of those pixels.
Figure 5.9: Scatter plot medians by loc at All Sites with linear regression fit. Each point has all pixels in a loc and error bars represent the 25% to 75% range of those pixels.

Count by loc at All Sites.
Figure 5.10: Count by loc at All Sites.

Correlation (R) and index of agreement (IOA) by loc at All Sites.
Figure 5.11: Correlation (R) and index of agreement (IOA) by loc at All Sites.

Mean bias (MB), mean absolute bias (MAB), and root mean squared error (RMSE) by loc at All Sites.
Figure 5.12: Mean bias (MB), mean absolute bias (MAB), and root mean squared error (RMSE) by loc at All Sites.

Mean bias (MBP), mean absolute bias (MABP), and root mean squared error (RMSEP) as percents by loc at All Sites. Percents are normalized by the reference mean.
Figure 5.13: Mean bias (MBP), mean absolute bias (MABP), and root mean squared error (RMSEP) as percents by loc at All Sites. Percents are normalized by the reference mean.

6: All Sites by cld

Box (25%, 50%, 75%) and whisker (min/max) plots by cld at All Sites.
Figure 6.1: Box (25%, 50%, 75%) and whisker (min/max) plots by cld at All Sites.

Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by cld. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.
Figure 6.2: Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by cld. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.

7: All Sites by tempo

Count by tempo at All Sites.
Figure 7.1: Count by tempo at All Sites.

Mean bias (MB), mean absolute bias (MAB), and root mean squared error (RMSE) by tempo at All Sites.
Figure 7.2: Mean bias (MB), mean absolute bias (MAB), and root mean squared error (RMSE) by tempo at All Sites.

Mean bias (MBP), mean absolute bias (MABP), and root mean squared error (RMSEP) as percents by tempo at All Sites. Percents are normalized by the reference mean.
Figure 7.3: Mean bias (MBP), mean absolute bias (MABP), and root mean squared error (RMSEP) as percents by tempo at All Sites. Percents are normalized by the reference mean.

8: All Sites by sza

Box (25%, 50%, 75%) and whisker (min/max) plots by sza at All Sites.
Figure 8.1: Box (25%, 50%, 75%) and whisker (min/max) plots by sza at All Sites.

Scatter plot medians by sza at All Sites with linear regression fit. Each point has all pixels in a sza and error bars represent the 25% to 75% range of those pixels.
Figure 8.2: Scatter plot medians by sza at All Sites with linear regression fit. Each point has all pixels in a sza and error bars represent the 25% to 75% range of those pixels.

Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by sza. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.
Figure 8.3: Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by sza. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.

9: All Sites by lsth

Box (25%, 50%, 75%) and whisker (min/max) plots by lsth at All Sites.
Figure 9.1: Box (25%, 50%, 75%) and whisker (min/max) plots by lsth at All Sites.

Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by lsth. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.
Figure 9.2: Count, correlation (R), index of agreement (IOA), mean bias (MB), mean absolute bias (MAB), root mean squared error (RMSE) at All Sites by lsth. MB, MAB, and RMSE are also sown as percents normalized to the reference mean.