Mean difference
The average Landsat 9 minus Landsat 7 value among the paired clear-land pixels.
PYTHON · GEOSPATIAL AUTOMATION · LANDSAT
A Python workflow for Landsat temperature comparison.Brevard County, Florida
The script retrieves, checks, aligns, compares, summarizes, and maps Landsat surface-temperature data.
PythonRasterioNumPySTAC APIQA bit masking
01 / GENERATED OUTPUT
Output from the script
Only locations with valid, quality-screened temperature values in both periods are displayed and summarized.
02 / RUN SUMMARY
Results from this pair of February datasets
The values describe this specific pair of February datasets. They are not estimates of regional climate change.
Median difference
The midpoint of the paired-pixel differences, less sensitive to extreme values than the mean.
Above +2°C
The share of paired pixels more than 2°C warmer in the later observation.
Pixels compared
Clear land cells remaining after the county clip, QA masks, alignment, and paired-data requirement.
03 / AUTOMATION
What the script does
The five stages cover data discovery, raster processing, quality checks, calculation, and output.
- 01
Search and select
Query the Landsat Collection 2 Level-2 STAC catalog for the specified February acquisitions and scene identifiers.
- 02
Read only what is needed
Stream the Brevard County windows from cloud-optimized GeoTIFFs instead of downloading every full Landsat band.
- 03
Apply quality controls
Decode QA bits to remove fill, clouds, cloud shadows, snow, water, saturation, and Landsat 7 dropped pixels.
- 04
Align and compare
Reproject the quality-screened mosaics to a common 60-meter grid and calculate Landsat 9 minus Landsat 7 in Celsius.
- 05
Report reproducibly
Generate the finished map and a machine-readable JSON summary from the same script used for the analysis.
QA masking
rejected_bits = sum(1 << bit
for bit in (0, 1, 3, 4, 5, 7))
valid = ((qa_pixel & rejected_bits) == 0)
valid &= qa_radsat == 0Temperature and difference
temperature_c = dn * 0.00341802
+ 149.0 - 273.15
difference = landsat_9 - landsat_704 / DATA INTEGRITY
Quality checks and limits
Landsat 7 scan-line gaps remain visible. The workflow rejects dropped and fill pixels rather than interpolating them into apparently complete data.
Acquisition conditions matter. Weather, moisture, atmosphere, overpass timing, emissivity, and sensor differences can affect the result.
A longer record is needed. Multiple years, additional dates, and environmental controls would be required before making a long-term trend claim.
05 / TEACHING AND REPRODUCIBLE MATERIALS
Code students can inspect and rerun
The Python file and JSON summary could support a lesson on QA masking, reprojection, raster alignment, paired-pixel comparison, and reproducible reporting.
Source data: USGS Landsat Collection 2 Level-2, accessed through Microsoft Planetary Computer. Boundary: U.S. Census Bureau TIGERweb.