Choosing satellite imagery for environmental change detection is a trade-off between detail, frequency, coverage, and the ability to see through clouds. This guide explains satellite revisit time and resolution, compares widely used public missions, and provides a repeatable workflow for selecting imagery that matches the landscape, time scale, and change you need to measure.
Overview
No single satellite is best for every monitoring project. A 10-metre image may reveal narrow roads, field boundaries, or small clearings, but it may not be collected often enough to capture a short-lived event. A sensor that observes the same area every day can support wildfire or flood monitoring, yet its coarser pixels may conceal small features.
The most useful starting point is to define the change before choosing the sensor. Are you measuring the expansion of a mining site, the seasonal condition of crops, the spread of a wildfire, the loss of wetland, or a gradual change in shoreline position? Each question has a different minimum spatial detail, observation frequency, spectral requirement, and tolerance for cloud cover.
Four types of resolution matter most:
- Spatial resolution: the ground area represented by each pixel. Smaller pixels generally show finer detail.
- Temporal resolution: how frequently a satellite can observe the same location, often described by revisit interval.
- Spectral resolution: the number and width of wavelength bands available. Near-infrared and shortwave-infrared bands, for example, can help distinguish vegetation, water, burn scars, and moisture.
- Radiometric resolution: the sensor's ability to record differences in energy or brightness. Higher radiometric detail can help identify subtle changes, although atmospheric effects and calibration also matter.
Swath width is another practical consideration. A wider swath covers more ground in one pass, which can improve the chance of obtaining a scene over a large or fragmented study area. The trade-off may be coarser spatial detail or less flexibility for small sites.
What to track
Satellite imagery resolution comparison
The table below is a planning guide rather than a substitute for checking the current mission catalog. Products can differ by band, processing level, acquisition mode, and archive availability. Revisit figures also describe nominal or typical opportunities, not a guarantee that a cloud-free image will exist on a particular date.
| Mission or sensor | Spatial detail | Typical revisit pattern | Spectral and radiometric strengths | Coverage and cloud limits | Common uses |
|---|---|---|---|---|---|
| Landsat series | About 30 m for many multispectral bands; finer detail is available in selected bands | About 16 days for an individual satellite; combined missions can improve observation opportunities | Long-running, calibrated multispectral record suitable for consistent change analysis | Optical imagery is blocked or degraded by cloud, smoke, haze, and shadows; moderate swath | Long-term land-cover change, forests, water bodies, urban growth, and surface disturbance |
| Sentinel-2 | 10 m, 20 m, and 60 m depending on band | Roughly several days with the constellation, subject to latitude, orbit, and acquisition planning | Multiple visible, near-infrared, and shortwave-infrared bands support vegetation, water, and burn analysis | Optical observations are affected by cloud and smoke; relatively wide field of view | Field monitoring, vegetation condition, flood and burn-scar mapping, and detailed land-cover change |
| MODIS | Approximately 250 m, 500 m, or 1 km depending on product and band | Daily or near-daily observation opportunities for many locations | Many bands and frequent products support regional to global environmental monitoring | Coarse pixels; optical observations remain vulnerable to cloud and atmospheric conditions | Fire activity, vegetation seasonality, snow and ice, aerosols, and broad regional trends |
| VIIRS | Approximately 375 m or 750 m for many land products | Daily or near-daily observation opportunities | Useful for broad-area environmental indicators, including fire and nighttime observations in relevant products | Coarser than Landsat or Sentinel-2; cloud and atmospheric screening still matter | Active fires, regional vegetation, snow, air-quality indicators, and large-scale change |
| Sentinel-1 radar | Often around 10 m in commonly used land products, depending on mode and processing | Several days to more than a week depending on location, acquisition mode, and constellation availability | Radar can provide information day or night and is less affected by cloud; sensitive to surface structure and moisture | Interpretation is less intuitive than optical imagery; speckle, geometry, vegetation effects, and acquisition mode require attention | Flood extent, soil or surface moisture indicators, deformation, and change in cloudy regions |
For long-term climate data analysis, consistency often matters more than maximum sharpness. A stable processing method applied to a moderate-resolution archive can be more useful than combining irregular high-detail scenes. For small sites, however, a 10-metre product may be preferable when the feature being measured is only a few pixels wide at coarser resolution.
Cadence and checkpoints
Use the following workflow when selecting the best satellite data for change detection.
- Define the measurable change. Write down the feature, expected direction of change, minimum size, and time scale. “Vegetation change” is broad; “seasonal loss of green cover across fields” is more operational.
- Set the spatial requirement. Estimate how many pixels should cover the smallest feature you want to measure. A feature represented by only one or two pixels is difficult to classify reliably, especially along boundaries.
- Set the temporal requirement. Decide whether you need daily alerts, observations every few days, monthly composites, or a multi-year record. Match the revisit interval to the speed of the process rather than collecting more images than you can interpret.
- Choose optical, radar, or both. Optical imagery is often easier to interpret and offers valuable color and infrared information. Radar can fill gaps caused by cloud and darkness, but it requires more specialized processing.
- Check spectral bands and product level. Confirm that the product includes the bands or derived measurements your method needs. Distinguish raw or minimally processed data from surface-reflectance, analysis-ready, or higher-level products.
- Inspect acquisition dates and cloud information. A nominal revisit is not the same as a usable observation. Review scene footprints, cloud masks, haze, smoke, snow, viewing geometry, and gaps around the study area.
- Document the trade-off. Record the mission, product name, processing level, band set, spatial resolution, acquisition dates, cloud-screening method, resampling method, and any missing observations.
For recurring monitoring, establish a fixed checkpoint. A monthly workflow might review new scenes, update a cloud-free composite, compare the result with the previous month, and flag unusual changes for closer inspection. A quarterly workflow can focus on seasonal summaries and reduce noise from short-term weather.
How to interpret changes
A visible difference between two images is not automatically environmental change. Illumination, sun angle, atmospheric conditions, soil moisture, tide stage, vegetation phenology, sensor calibration, and image registration can all alter the signal. Compare like with like whenever possible: use the same band definitions, processing approach, season, and projection.
Image alignment is especially important. If two scenes are shifted by even a small distance, boundaries may appear to move. Reprojecting data to a common grid and checking control features can reduce this problem. Resampling also changes the appearance of pixels, so record whether nearest-neighbour, bilinear, or another method was used.
Cloud masks are useful but imperfect. Cloud shadows, haze, smoke, and thin cirrus may remain after automated screening. Inspecting the images visually, or comparing multiple quality bands, can prevent a false change signal. When the observation is important, confirm it with another date, another sensor, field information, or an independent dataset.
Be cautious when comparing sensors with different spatial resolutions. A small wetland may be visible in Sentinel-2 but mixed with surrounding land in MODIS. Conversely, a regional MODIS time series may reveal a broad seasonal pattern that would be expensive or impractical to build from higher-resolution scenes. This is a resolution trade-off, not necessarily a disagreement.
For related environmental applications, see Satellite Imagery for Wildfires and Air Quality and How to Compare Climate Datasets. If your project concerns coastlines or water levels, How to Measure Sea-Level Rise explains why satellite measurements should be considered alongside tide gauges and land-motion data.
When to revisit
Revisit this selection process whenever the monitoring objective changes, a new season begins, or the current archive produces too many gaps. For routine environmental monitoring, check imagery monthly or quarterly according to the speed of change. After a wildfire, flood, storm, or other rapid event, inspect the archive as soon as practical and continue checking until cloud cover and acquisition timing provide a usable comparison.
Also review the workflow when a mission changes its product structure, processing terminology, band definitions, archive access, or availability. Mission specifications and catalog interfaces can evolve, so confirm current documentation before starting a new analysis or comparing a new dataset with an older one.
Practical checklist: define the change; choose the smallest useful pixel size; match revisit time to the process; check spectral bands and cloud limits; compare compatible dates; document every processing choice; and validate important findings with another observation. If you need sharper local detail, move toward Landsat or Sentinel-2. If you need frequent regional coverage, consider MODIS or VIIRS. If clouds are the main obstacle, evaluate radar alongside optical imagery.
