GIS4035C: Remote Sensing of Environment

Course Description

GIS4035C Remote Sensing of Environment teaches how to extract information about the Earth's surface from imagery collected without touching it — aerial photographs, satellite sensors, thermal and radar systems — and how to know when the answer you have extracted is wrong.

The statewide inventory records the course at four institutions, including Florida Atlantic University, Florida State University, the University of South Florida and the University of West Florida.

UWF's description is the fullest: the course "familiarize[s] students with the fundamentals of remote sensing and photo interpretation through hands-on techniques with aerial photographs and satellite imagery based on real-world applications. Both active and passive sensors will be discussed." ⚠ It states its own structure explicitly: "broken up into two distinct sections — interpreting digital aerial photographs and examining and manipulating digital data from satellites and other remote sensors." Its prerequisite is GIS 4043*/L* — the GIS course and its laboratory, both permitted concurrently. The course sits in the Department of Earth and Environmental Sciences.

FSU's covers "remote sensing foundations and the use of remote sensing for environmental and cultural applications… aerial photography and photogrammetry, characteristics of various sensing systems." ⚠ "Cultural applications" is worth noting — it signals archaeological and land-use work alongside the environmental core.

The physical foundation, which is where the course starts and where students underestimate it. ⚠ Remote sensing is applied physics before it is image processing. Electromagnetic radiation interacts with the atmosphere and with surfaces in ways that depend on wavelength — and the whole discipline rests on the fact that different materials reflect and emit differently across the spectrum. Healthy vegetation is dark in red and extremely bright in near-infrared, which is why a simple ratio of those two bands measures plant vigour; water absorbs near-infrared almost completely, which is why it appears black and why shorelines are easy to map. The spectral signature is the thing being measured; the picture is just how it is displayed.

The four resolutions, which govern everything a sensor can and cannot answer.

• Spatial — how small an object can be resolved.
• Spectral — how many bands and how narrow.
• Radiometric — how finely brightness is quantised.
• Temporal — how often the sensor revisits.

⚠⚠ These trade against each other, and understanding the trade-off is the course's central practical skill. A sensor with very fine spatial resolution has a narrow swath and revisits rarely; one that images the whole Earth daily is coarse. Choosing the wrong sensor for a question is the most common and most expensive error in applied remote sensing — and it is made before any analysis begins.

Passive and active. ⚠ UWF names both, and the distinction matters in Florida more than in most places. Passive sensors measure reflected sunlight or emitted heat — and they cannot see through cloud. Active sensors supply their own energy: radar (SAR) penetrates cloud and works at night; lidar measures distance directly and produces elevation. ⚠⚠ In a state that is cloudy through the entire wet season and whose most urgent imaging needs arise immediately after hurricanes, radar is not an advanced topic — it is the sensor that works when the optical one cannot.

Learning Outcomes

Required Outcomes

• Explain the physical basis of remote sensing — electromagnetic radiation, the spectrum, atmospheric windows, scattering and absorption.
• Explain spectral signatures and how surfaces differ in their interaction with radiation across wavelengths.
• Distinguish passive and active sensing systems and explain the advantages of each.
• Explain the four resolutions and analyse the trade-offs among them for a given application.
• ⚠ Select an appropriate sensor and imagery source for a stated question, and justify the choice.
• Describe the major satellite programmes — Landsat, Sentinel, MODIS, commercial high-resolution constellations — and their characteristics.
• Perform visual photo interpretation using tone, texture, shape, size, pattern, shadow, site and association.
• Explain photogrammetry fundamentals — scale, relief displacement, stereo viewing, orthorectification.
• Apply geometric correction and georeferencing to imagery.
• Apply radiometric and atmospheric correction, and explain when each is necessary.
• Perform image enhancement — contrast stretching, filtering, band combinations.
• Compute and interpret spectral indices such as NDVI, NDWI and burn indices.
• Perform unsupervised classification and interpret the resulting clusters.
• Perform supervised classification, including training sample selection and signature evaluation.
• ⚠ Conduct an accuracy assessment using a confusion matrix, and report overall, producer's and user's accuracy and kappa.
• ⚠ Explain why a classified map without an accuracy assessment is not a result.
• Perform change detection between image dates and distinguish real change from artefact.
• Integrate remote sensing products with GIS data and analysis.
• Explain lidar and its use in elevation and vegetation structure mapping.
• Explain radar fundamentals and its applications in cloud-prone and post-disaster settings.
• Acquire imagery from public archives and document its provenance and processing level.
• Produce a professional map and technical report presenting method, results and limitations.

Optional Outcomes

• Explain hyperspectral imaging and its applications.
• Apply object-based image analysis rather than pixel-based classification.
• Apply machine learning classifiers — random forest, support vector machines — to imagery.
• Use Google Earth Engine or a cloud platform for large-area analysis.
• Plan and conduct a drone (UAS) survey and process the imagery.
• Explain thermal remote sensing and urban heat applications.
• Explain bathymetric and coastal remote sensing.
• Explain archaeological and cultural applications.
• Write scripts in Python for image processing workflows.

Major Topics

Required Topics

• Electromagnetic radiation and energy interactions.
• Spectral signatures and reflectance curves.
• Sensors and platforms; the four resolutions.
• Aerial photography and photogrammetry.
• Visual image interpretation.
• Digital image structure and display.
• Geometric correction and georeferencing.
• Radiometric and atmospheric correction.
• Image enhancement and band combinations.
• Spectral indices.
• Unsupervised and supervised classification.
• Accuracy assessment.
• Change detection.
• Lidar and radar.
• Integration with GIS.
• Data sources, archives and metadata.
• Map production and technical reporting.

Optional Topics

• Hyperspectral imaging.
• Object-based image analysis.
• Machine learning classification.
• Cloud platforms and Google Earth Engine.
• Drones and UAS.
• Thermal applications.
• Coastal and bathymetric remote sensing.
• Archaeological applications.
• Python scripting.

Resources & Tools

• Jensen, Introductory Digital Image Processing and Remote Sensing of the Environment — ⚠ Jensen is the standard author in this field and the second title is where the statewide course name comes from; Lillesand, Kiefer and Chipman, Remote Sensing and Image Interpretation — the other classic, strong on photo interpretation; Campbell and Wynne, Introduction to Remote Sensing.
• ⚠⚠ The imagery is free, and this is the single most important practical fact about the field. Landsat — over fifty years of continuous global coverage, free from USGS EarthExplorer; Sentinel-1 (radar) and Sentinel-2 (optical) — free from the Copernicus Data Space, with better resolution and revisit than Landsat; MODIS and VIIRS for daily coarse coverage; NASA Earthdata as the general portal. An undergraduate can do real analysis at no cost, which is not true of most technical fields.
• Software: ArcGIS Pro with the Image Analyst extension and ERDAS IMAGINE where the department licenses them (⚠ Esri offers a student licence and both are what job postings name); ENVI. ⚠⚠ Free and fully capable alternatives: QGIS with the Semi-Automatic Classification Plugin, SNAP (ESA's toolbox, and the standard tool for Sentinel-1 radar), Orfeo ToolBox, and Google Earth Engine — ⚠ Earth Engine is free for research and education and lets you process decades of imagery over a whole state without downloading anything. Learn it if the course allows.
• Python: rasterio, GDAL, numpy, scikit-learn, earthengine-api. ⚠ Scripting is what separates a technician from an analyst in this field, and it is worth starting even if the course does not require it.
• ⚠⚠ Florida-specific data, all free: the Florida Geographic Data Library (FGDL) at UF; the Florida Department of Environmental Protection and the five Water Management Districts, which publish extensive imagery and land cover; the Florida Fish and Wildlife Conservation Commission's land cover and habitat data; NOAA's Digital Coast — ⚠ lidar-derived elevation and sea level rise viewers for the entire Florida coast, free; and FDOT aerial imagery. Florida is one of the best-mapped states in the country and a student here has an unusual advantage on project work.
• Professional: ASPRS (the American Society for Photogrammetry and Remote Sensing) — ⚠ its Certified Mapping Scientist and Certified Photogrammetrist credentials are the field's recognised certifications; the GISP certification from GISCI; ⚠⚠ and the FAA Part 107 Remote Pilot Certificate if drones are involved — it is inexpensive, achievable as a student, and named in job postings.

Career Pathways

• Cartographers and photogrammetrists (SOC 17-1021) — the field's own occupational code.
• Geographers (SOC 19-3092) and geoscientists (SOC 19-2042).
• GIS analysts, specialists and developers (SOC 15-1299, 17-1021) — ⚠ the largest employer category, and remote sensing is a differentiating skill within it.
• Remote sensing scientists and image analysts (SOC 19-2099, 17-1021).
• Environmental scientists (SOC 19-2041) — wetland delineation, land cover change, water quality.
• Urban and regional planners (SOC 19-3051) — ⚠ land use change monitoring is standard practice in Florida's fast-growing counties.
• Foresters and conservation scientists (SOC 19-1032, 19-1031).
• Precision agriculture specialists (SOC 19-1013, 11-9013) — ⚠ Florida's citrus, sugar and vegetable sectors use imagery operationally.
• Emergency management and damage assessment (SOC 11-9161, 19-2041) — ⚠⚠ a genuine and recurring Florida role. After every major hurricane, imagery-based damage assessment drives federal disaster declarations and insurance response, and analysts are mobilised for it.
• Defence and intelligence imagery analysts (SOC 33-3021, 17-1021) — ⚠ the National Geospatial-Intelligence Agency and its contractors are major employers; requires US citizenship and clearance. MacDill Air Force Base in Tampa and the Space Coast contractors give Florida a real presence in this sector.
• Coastal and marine scientists (SOC 19-2042, 19-1023) — shoreline change, seagrass and reef mapping, harmful algal bloom detection.
• Archaeologists (SOC 19-3091) — ⚠ lidar has transformed site detection under vegetation, and this is the "cultural applications" FSU names.

Special Information

⚠ Register for both halves if your institution splits them

At FSU, GIS 4035L is a corequisite and a separate registration. ⚠ Enrolling in the lecture alone leaves the requirement unmet, and the pair is normally offered on a fixed rotation, so the error costs a term. Check the degree audit rather than the course listing.

⚠ Prerequisites — and note what UWF's notation means

UWF requires GIS 4043*/L* — the introductory GIS course and its laboratory, with both permitted concurrently.

• ⚠ The /L notation is itself worth reading carefully: it means the prerequisite is a split family too. You need both halves of GIS 4043, not just the lecture.
• ⚠⚠ Take GIS before this course if you can, rather than alongside. Remote sensing outputs are consumed in GIS — the classification you produce is a raster you then analyse, clip, overlay and map — and a student meeting projections, georeferencing and map layout for the first time in the same term is doing two courses' learning at once.
• Unlisted but assumed: ⚠ enough physics to be comfortable with wavelength, energy and the idea of a spectrum; basic statistics for the classification and accuracy material; and comfort with large files and computer housekeeping, which sounds trivial and is not — satellite scenes are large, and students routinely run out of disk space mid-assignment.

Course format and workload

3 credits, 60 contact hours in the integrated C form — lecture plus laboratory. In FSU's split form, 3 credits lecture plus a separately enrolled laboratory, 4 credits total.

Expect 8–10 hours per week outside class. ⚠⚠ The laboratory work is the workload and it is unpredictable: a processing chain that runs cleanly takes an hour and the same chain with a projection mismatch takes an evening. Start laboratory assignments early — not for virtue, but because software failures are not compressible.

Assessment typically includes examinations, weekly laboratory exercises with map deliverables, and an independent project applying the workflow to a question of your choosing — ⚠ which is the assignment that becomes a portfolio piece, and the reason to choose a real local question rather than a tutorial dataset.

⚠⚠ Where students struggle, and the one habit that prevents most of it

• ⚠⚠ Projections and coordinate systems. The single largest source of lost time in this course and in the profession. Data in different coordinate systems will not overlay, will overlay wrongly, or will appear correct and produce wrong measurements. ⚠ The habit that prevents it: check the coordinate system of every layer before doing anything, and define a project projection deliberately rather than accepting whatever loads first.
• Training sample selection in supervised classification. ⚠ Classification quality is determined by the training samples, not by the algorithm. Students blame the classifier; the problem is nearly always samples that are too few, unrepresentative, or drawn from mixed pixels.
• ⚠⚠ Believing the classified map. A classification always produces a map. It looks authoritative. It can be substantially wrong. The accuracy assessment is not a formality appended at the end — it is the only thing that turns a picture into a result, and a map presented without one should not be trusted, including your own.
• Mixed pixels. A 30-metre Landsat pixel over a Florida suburb contains roof, road, grass and tree. It is not any one class, and understanding that explains most classification error.
• Change detection artefacts. ⚠ Two images from different seasons differ because of phenology, sun angle, tide and water level — not because the landscape changed. In Florida, wet-season and dry-season imagery of the same wetland can look like catastrophic land cover change and represent nothing at all.
• File and data management. Unglamorous, and the difference between a reproducible project and an unrecoverable one.

⚠ Florida makes this course unusually concrete

Applications a Florida student can work on with free data:

• Hurricane damage assessment — ⚠ radar is the operational sensor here, because the days after landfall are cloudy, and this is exactly why UWF's description names active sensors.
• Coastal change and shoreline erosion; sea level rise exposure using NOAA lidar-derived elevation.
• Wetland and Everglades monitoring, including restoration tracking.
• Harmful algal blooms and red tide — ⚠ ocean colour sensors detect blooms operationally, and Florida's bloom events are a recurring public health and economic problem.
• Urban growth in Central and South Florida, which is among the fastest land cover change in the United States.
• Seagrass and coral reef mapping, and the shallow-water bathymetry problem.
• Wildfire burn severity mapping in the pine flatwoods.
• Invasive species detection — melaleuca, Brazilian pepper, hydrilla.

⚠ Choose your independent project from this list. The data are free, the questions are real, local agencies care about the answers, and it makes a far better portfolio piece than a generic exercise.

Articulation and transfer

⚠ The suffix and credit count are the transfer issue, as with every split family: GIS4035C (3 credits, integrated) versus GIS4035 + GIS4035L (4 credits). SCNS equivalency does not cross the suffix automatically. ⚠ Compounding it here, the titles differ at every institution — Remote Sensing of Environment, Introduction to Remote Sensing, Photo Interpretation and Remote Sensing — so a title-based match will fail. Search by number and keep the syllabus and lab exercises.

This is a 4000-level upper-division course; Florida College System institutions generally do not offer it, though some teach introductory GIS at the lower division and those courses transfer.

Prefix note. GIS is geographic information science; GEO geography; GLY geology; SUR surveying and mapping; EVR environmental science; FOR forestry. ⚠ Remote sensing is taught under GIS, GEO, SUR and FOR at different Florida institutions depending on which department owns it — search by subject rather than prefix.

AI Integration

⚠⚠ Remote sensing has been a machine learning field for decades — longer than most students realise — and it is now one of the areas where deep learning has changed practice most.

What is genuinely standard practice:

• Supervised classification IS machine learning. ⚠ Maximum likelihood classification, taught in this course as a foundational technique, is a statistical classifier — the field has been doing this since the 1970s and simply did not call it AI.
• Random forests and support vector machines now routinely outperform the classical classifiers and are available in QGIS, ArcGIS and Earth Engine.
• Deep learning for object detection and segmentation — buildings, roads, vehicles, individual trees, damaged structures — ⚠ convolutional networks substantially outperform pixel-based methods because they use spatial context, which is what a human interpreter uses too.
• Cloud and shadow masking, atmospheric correction and gap filling, now largely automated.
• Change detection at scale across continental archives, which is only feasible computationally.
• Foundation models trained on satellite imagery are an active and fast-moving research area.

⚠⚠ Where it fails, and the failures are the course's own lessons restated:

• ⚠⚠ Training data quality still determines everything. A deep model trained on poor labels produces a confident, detailed, wrong map. The accuracy assessment is MORE important with a complex model, not less, because the output looks more convincing.
• Transferability. ⚠ A model trained on imagery from one region, season or sensor frequently fails on another. A building detector trained on temperate cities does poorly on Florida's vegetation-obscured suburbs, and a classifier trained on dry-season imagery misclassifies the wet season.
• Physical implausibility. ⚠ Models produce results that violate physics — vegetation where reflectance says water, elevation that does not drain. Knowing the spectral physics is what lets you catch it, which is the argument for learning the foundations before the tools.
• Explaining a language model's role specifically: it is useful for writing and debugging processing scripts, explaining a concept, and drafting a report's method section. ⚠ It is not useful for stating sensor specifications, band designations or data availability — those are fabricated confidently and are trivially checkable in the mission documentation.

The professional point. ⚠⚠ Remote sensing outputs are used to make consequential decisions — disaster declarations, insurance payouts, wetland permitting, habitat designation, and in defence contexts, targeting. An analyst is responsible for the accuracy and the stated limitations of what they produce. The professional habit the course should leave you with is publishing the accuracy assessment alongside the map, every time — a map without stated accuracy invites a confidence it has not earned, and automation makes producing such maps faster, not safer.

Academic integrity. Follow the course policy. Submitting generated work as your own violates every Florida institution's policy — and in this course the laboratory work is the portfolio, which makes it a poor thing to outsource.