Satellite images used to feel like magic postcards from space. Now they’re everyday tools. Farmers watch crops week by week, cities track floodwater at night, and insurers validate claims without sending trucks. You don’t need a space program to do this. You need a clear workflow, a few trustworthy data sources, and a basic sense of what those colorful bands actually mean.
This guide is simple by design. It focuses on practical steps you can use today with open data and common tools. We’ll cover which sensors fit your job, how to fetch and clean imagery, and how to build maps that answer real questions. You’ll learn enough to get reliable results without needing a PhD—or getting lost in jargon.
Start With the Question, Not the Satellite
Before you open a portal or load a map, write down the decision you want to support. “Is my field oversaturated?” “Which city blocks heated up most this month?” “Where did the river overflow yesterday?” The question points you to the right sensor and data product.
Match the question to the sensor
- Greenery and crops: Use optical imagery with red and near-infrared bands (e.g., Sentinel‑2, Landsat 8/9). These support vegetation indices like NDVI.
- Water, floods, and coastlines: Optical works when skies are clear, but SAR (radar) like Sentinel‑1 cuts through clouds and darkness. That’s crucial during storms.
- Urban heat and hotspots: Thermal bands from Landsat can estimate land surface temperature and urban hot zones.
- Nighttime activity: VIIRS night lights can show outages, blackout recovery, or shifting human activity at large scales.
Two rules frame almost every choice you’ll make:
- Resolution vs. revisit: Images with finer detail often arrive less often or cost money. Open data gives you medium detail with frequent revisits.
- Optical vs. radar: Optical looks like photos. Radar measures how rough or smooth a surface is and doesn’t care about clouds or daylight.
The Core Sensors You’ll Actually Use
You can do a lot with a few open missions. Knowing what they actually provide saves time and headaches.
Sentinel‑2 (optical)
What it is: A pair of European satellites capturing 13 spectral bands from visible to shortwave infrared. Resolution is 10–60 m depending on the band. Typical revisit: ~5 days near the equator with both satellites.
Why it’s useful: Clear, consistent surface reflectance products; great for vegetation, water quality, and land change mapping.
Landsat 8/9 (optical + thermal)
What it is: USGS/NASA’s long-running mission. Optical bands at 30 m and thermal bands at 100 m (resampled to 30 m). Revisit ~8 days when you combine 8 and 9.
Why it’s useful: Reliable, well-documented, and includes thermal. Helpful for heat islands, burn severity, and long time series.
Sentinel‑1 (SAR)
What it is: Radar at C‑band with VV/VH polarizations. Resolution commonly 10 m. Independent of cloud cover and daylight.
Why it’s useful: Flood detection, forest monitoring, and infrastructure coherence checks—even in storms and at night.
VIIRS (night lights)
What it is: Nighttime visible bands that capture human lighting patterns at coarse resolution.
Why it’s useful: Power outage monitoring, economic activity proxies, disaster recovery signals at national or regional scale.
Getting the Data Without Wrestling It
Open imagery is broadly accessible, but the easiest path depends on your skills and time.
Portals and platforms
- USGS EarthExplorer: A direct way to download Landsat and more. Simple, but you manage files yourself.
- Copernicus portals: Access Sentinel‑1 and Sentinel‑2 products, often with surface reflectance and cloud masks ready.
- Google Earth Engine: Compute near the data with a massive public catalog. Good for analysis at scale. It’s powerful but has a learning curve.
- Microsoft Planetary Computer: Similar idea with open datasets, STAC, and hosted compute. Plays nicely with Python and Jupyter.
- Sentinel Hub: Commercial service for streaming and processing imagery with handy APIs and ready-to-use layers.
Learn one key idea: STAC
The SpatioTemporal Asset Catalog (STAC) is a common way to describe and find imagery. Think of it as a searchable card catalog for scenes and bands. Many modern APIs speak STAC, which means you can switch providers without rewriting everything.
Cloud-optimized GeoTIFFs (COGs)
A COG is a GeoTIFF that supports partial reads on cloud storage. You don’t have to download the entire image to render a small tile in your map. Tools like QGIS and cloud tile servers can stream these efficiently.
Cleaning Data: Clouds, Atmosphere, and All the Unseen Stuff
If you want trustworthy numbers, do just enough pre-processing. Over-processing wastes time; under-processing gives you noisy results.
Surface reflectance vs. top-of-atmosphere
Whenever possible, use surface reflectance products. They correct for the atmosphere so vegetation and water indices are more consistent. Sentinel‑2 L2A and Landsat Collection 2 Level‑2 are your friends.
Clouds and shadows
Cloud masks are included in many datasets. In Sentinel‑2, the QA60 band flags clouds. Some services also provide scene classification layers that mark clouds, shadows, water, and more. Mask them out before calculating indices.
Composites beat individual scenes
To reduce noise, create a composite over a window—say, one to two weeks. A median or percentile composite smooths out clouds and small anomalies. For change detection after a disaster, pick a clean pre-event composite and a post-event composite from the earliest usable window.
Simple Indices That Do Serious Work
Most applied EO maps boil down to a handful of simple band ratios or differences. Keep it simple before you reach for a deep model.
Vegetation and burn severity
- NDVI: (NIR − Red) / (NIR + Red). Greenery vigor. Good for crop monitoring and drought stress.
- NBR: (NIR − SWIR2) / (NIR + SWIR2). Burn severity when you compare pre/post fire values.
Water
- NDWI: Uses Green and NIR to highlight water bodies. Works best in cloud-free optical data.
- SAR flooding: Lower backscatter over smooth water surfaces. Use change detection to reduce false positives.
Urban heat
- Landsat thermal (TIRS): Estimate land surface temperature with emissivity assumptions. You don’t need exact °C to rank hotspots.
Water quality proxies
- Chlorophyll-like indices: Ratios using Red Edge and NIR bands can indicate algal blooms in lakes and coasts (Sentinel‑2 shines here). Treat results as screening, not lab-grade measurements.
Crash Course in Radar (SAR) Without the Headache
Radar measures how the surface reflects microwave pulses back to the satellite. Smooth water returns little energy; rough surfaces return more. The two common polarizations in Sentinel‑1, VV and VH, help you read targets. More VV usually indicates roughness; VH is sensitive to vegetation structure.
Key SAR concepts you can use
- Incidence angle: Backscatter changes with the angle of the beam. Comparing scenes with similar angles improves consistency.
- Change detection: Subtract a clean “before” scene from “after.” Floodwater shows up as a sharp drop in backscatter.
- Coherence: Measures stability between passes. Drops in coherence can indicate landslides, construction, or damage.
Three Practical Workflows You Can Run This Week
1) Rapid flood mapping with Sentinel‑1
Goal: Identify areas likely underwater after a storm, even with clouds.
- Get data: Search a STAC catalog for Sentinel‑1 GRD scenes covering your area one month before the event and within 24–72 hours after.
- Preprocess: Use calibrated and terrain-corrected products (many platforms provide this). Clip to your area of interest. Normalize if needed.
- Compute change: For VV polarization, compute a difference image: pre − post. Pixels with a large drop are candidates for flooding.
- Refine: Mask steep slopes or high ground using a digital elevation model (DEM). Smooth small speckles with a simple filter to reduce noise.
- Validate: Cross-check with any optical windows that are cloud-free, or with ground photos and reports.
Tip: Avoid treating a single threshold as truth. Compare to known water masks and adjust. The goal is a rapid, directionally accurate map for response, not a final hydrology study.
2) Urban heat screening with Landsat thermal
Goal: Rank neighborhoods by relative heat to target shade, cool roofs, or tree planting.
- Get data: Landsat 8/9 Level‑2 scenes during hot, clear days. Mask clouds and shadows.
- Thermal band: Use the surface temperature product if available or convert radiance to brightness temperature with the provided metadata.
- Composite: Average multiple days to reduce noise. Convert to z-scores or deviations from citywide mean.
- Overlay context: Add land cover, impervious surfaces, and tree canopy layers to explain the pattern.
- Publish: A simple color ramp showing the hottest quartile by block group is actionable for planners.
Tip: Focus on relative heat zones. Exact temperature requires detailed emissivity and atmospheric correction, which may be overkill for targeting interventions.
3) Algal bloom watch with Sentinel‑2
Goal: Flag potential harmful algal blooms in lakes and reservoirs.
- Get data: Sentinel‑2 Level‑2A scenes with low cloud cover. Focus on red, red-edge, and NIR bands.
- Mask water: Create a water mask using NDWI or a scene classification layer to exclude land.
- Index: Use a simple chlorophyll-like ratio (e.g., Red Edge to Red) as a proxy. Track changes week to week.
- Ground truth: Pair with a few low-cost water samples. The resulting map is a screening tool that guides testing, not a replacement for lab results.
Tip: Wind, sun glint, and suspended sediment can confuse optical signals. Consistency over time matters more than absolute numbers.
Visualization Without the Pain
Once you have metrics, you need a clear map. Good visualization tells a story.
Desktop: QGIS
QGIS is free and powerful. It reads cloud-optimized GeoTIFFs and remote STAC assets with plugins. For most users, it’s enough to layer a base map, your composite raster, and some administrative boundaries, then publish as images or PDFs.
Web: Stream tiles from COGs
If you need a simple web map, serve tiles directly from COGs through a tiler and display them with a lightweight library. MapLibre GL JS renders vector and raster tiles; it’s easy to integrate into a website and works well with open data stacks.
Legend and color discipline
- Use intuitive schemes: blue for water, green for vegetation, red for high change or heat.
- Include a clear legend and a “how to read this map” note. Even experts appreciate clarity.
- Show confidence or quality layers when you can. Trust improves when people see uncertainty.
Analyzing Near the Data: Pick Your Battle
Large scenes are heavy. Moving them around is slow and costly. Instead, compute where the data already lives.
Earth Engine
Great for long time series, composites, and rapid prototyping. You write concise scripts, and the platform handles scaling. It has many datasets pre-loaded and updated.
Planetary Computer
Python-first, with STAC search and analysis-ready COGs. Works well with xarray and cloud-native geospatial tools. For teams already on Azure, it reduces friction.
Sentinel Hub
Strong for operational setups that need predictable APIs, dynamic mosaicking, and easy integration into apps or dashboards.
Common Pitfalls and Simple Fixes
Most bad maps come from the same few mistakes.
Comparing apples to oranges
Don’t compare images with very different sun angles, seasons, or atmospheric conditions. Use surface reflectance and composites over similar windows. In SAR, align incidence angles and orbits when possible.
Ignoring masks
Clouds, shadows, and snow create false signals. Always check and apply masks. If your map looks “too crispy,” it probably is—add smoothing and quality filters.
Forgetting licensing
Open missions are fine, but many commercial scenes prohibit public redistribution. Read the license before publishing. If in doubt, use derived maps rather than raw imagery tiles.
Overconfidence
Indices are proxies. Pair them with ground truth or secondary datasets. Use simple thresholds and publish confidence bands where possible. Stakeholders trust maps that admit uncertainty.
From Map to Decision: Who Uses This and How
Satellite imagery becomes valuable when it changes what people do next. Aim for decisions you can track.
Agriculture
- Spot stress and irrigation leaks with NDVI drops.
- Plan scouting routes: visit the parcels with largest week-over-week change first.
- Document events for agronomic advice or claims.
Cities and utilities
- Map heat islands and target shade trees and cool roofs where it counts most.
- Monitor flood-prone blocks with SAR, then adjust drainage maintenance schedules.
- Track green space gains or losses against policy goals.
Insurance and finance
- Validate wind or flood claims with change maps and clear date stamps.
- Screen portfolios for wildfire risk with burn severity and vegetation recovery trends.
Environmental groups and researchers
- Watch wetlands and mangrove health with NDWI and NIR bands.
- Screen lakes for potential algal blooms to focus sampling budgets.
- Track reforestation projects with seasonal composites.
A Simple, Repeatable Workflow
Consistency beats complexity. Use a short checklist for every project.
- Question: Write the decision you’ll inform and the geography of interest.
- Sensor: Choose optical, thermal, or SAR—and which mission fits revisit and resolution needs.
- Data source: Pick Earth Engine, Planetary Computer, Sentinel Hub, or direct download.
- Preprocessing: Use surface reflectance, mask clouds/shadows/snow, and set a time window.
- Metric: Select a simple index or change method. Test it on known areas.
- Validation: Compare against ground photos, gauges, or trusted datasets.
- Visualization: Choose a clear color ramp, legend, and a short “how to read” note.
- Delivery: Export a map, share a lightweight web view, or schedule updates.
- Review: After use, record what worked and what confused readers. Improve the template.
What’s Next and Worth Watching
The open EO stack is advancing fast. A few trends matter for practical teams:
- Higher revisit, moderate resolution: More frequent public missions and partnerships mean better odds of a usable scene when you need it.
- Cloud-native by default: STAC, COGs, and scalable APIs reduce the need to download giant files. Expect more real-time tiles and analysis-on-demand.
- Blended sources: Combining SAR and optical improves reliability. Cheap ground sensors add context that satellites can’t see alone.
- Gentle ML: Pretrained models and lightweight classifiers will help clean clouds, segment water, and flag change without complex pipelines. Use them, but keep your sanity checks.
Short Playbook: Pick the Right Tool for the Job
- Fast disaster check: Sentinel‑1 SAR change map, no waiting for clear skies.
- Crops weekly: Sentinel‑2 surface reflectance NDVI composite.
- Heat planning: Landsat thermal composites and relative ranking.
- Night activity: VIIRS for national/regional patterns, not street-level insight.
- Publish to web: COG tiles + MapLibre for a simple shareable map.
Summary:
- Start with the decision, then pick the sensor: optical for color and vegetation, SAR for floods and clouds, thermal for heat.
- Rely on open missions (Sentinel‑1/2, Landsat, VIIRS) and cloud-native access via STAC and COGs to work faster.
- Use surface reflectance, masks, and composites so your maps are stable and repeatable.
- Stick to simple, proven indices and change detection before trying complex models.
- Validate with ground truth and be honest about uncertainty; publish clear legends and confidence cues.
- Compute near the data with platforms like Earth Engine or Planetary Computer, then stream results to the web.
- Make maps that guide action—routing, targeting, scheduling—not just pretty pictures.
