Remote sensing is the acquisition and analysis of information at a distance, without direct contact with the phenomenon. In our services it rests on two main sources: satellite imagery and aerial imagery. The discipline is used for wide-area coverage, periodic monitoring, comparison of a region’s condition across time, and extraction of land-cover layers. This page describes the path from sensor selection and image acquisition through pre-processing, index computation, classification, and time-series analysis. It also sets out the deliverables and the boundaries of the method, including the limits imposed by cloud, shadow, and the accuracy of reference data.
Foundations and the role of data in analysis
In remote sensing, each pixel is a measurement of reflected or emitted energy in a defined wavelength range. The quality of the analysis depends on two factors: matching the sensor to the phenomenon under study, and the quality of the corrections applied before analysis. Sensor and scene selection is therefore a technical decision, not a simple choice of data source.
The client’s question determines the level of spatial detail, the reliability of cover-type discrimination, and the temporal repeat rate that is required. For vegetation monitoring at regional scale, moderate spatial resolution and a short revisit interval are suitable; finer-scale investigation needs imagery of higher resolution.
For this reason, the work begins by recording the list of questions, the geographic extent, and the study period, and on that basis the sensor family and the time window are chosen.
Sensor families: optical, thermal, and radar
Optical sensors record solar energy reflected in the visible, near-infrared, and short-wave infrared bands, and they underpin vegetation and water indices. Thermal sensors record emitted energy in the thermal infrared range and are used for surface temperature and for some investigations linked to surface moisture.
Radar sensors belong to the active group: they transmit their own signal and record the return. Synthetic aperture radar (SAR) does not depend on sunlight and images through cloudy conditions, which makes it suitable for monitoring in cloudy seasons and for interferometric applications (InSAR).
- Multispectral optical: land cover, vegetation and water indices, cover classification.
- Thermal: surface temperature and investigations linked to moisture and heat.
- Radar (SAR): sunlight-independent imaging, moisture monitoring, and interferometric applications.
- Aerial imagery: higher detail over smaller areas.
Resolution classes and revisit
Four kinds of resolution govern remote sensing. Spatial resolution refers to the pixel size on the ground, spectral resolution to the number and width of bands, radiometric resolution to the capacity to distinguish intensity levels, and temporal resolution to the interval between two images of the same area.
Revisit expresses the time gap between two acquisitions and is decisive for monitoring rapid change. Viewing angle, local overpass time, and atmospheric conditions each affect how comparable images taken at different times are.
When sources are combined, the resolution gap between satellite and aerial imagery must be stated explicitly; the combination can be useful in some applications, provided the scale limitation of each layer is recorded in metadata.
Pre-processing, geometric correction, and orthorectification
Geometric correction ties the image to a defined spatial reference and carries out transformations between references with their parameters recorded. Atmospheric correction reduces the effect of scattering and absorption in the atmosphere so that surface reflectance values come as close as possible to a real measurement.
Orthorectification removes distortion caused by terrain slope and by the displacement of relief features, using a digital elevation model, and converts the image into one with a uniform scale; this step is a precondition for accurate use of the image in GIS. Cloud and shadow masking handles invalid pixels so that no spurious value enters later analysis.
Where needed, the panchromatic band is merged with multispectral bands (pansharpening) to give imagery with high spatial detail and spectral colour; in that case any artefacts of the fusion are noted in the method report.
Spectral indices and classification
Indices are combinations of several bands that highlight a surface property. The normalised difference vegetation index (NDVI), together with indices such as SAVI and EVI, targets vegetation; the water index (NDWI) targets water bodies, and built-up indices such as NDBI target constructed surfaces.
Classification maps pixels to land-cover classes and runs in two forms, supervised and unsupervised. In the supervised form, training samples are defined for each class and validation is carried out against independent reference data; in the object-based approach (OBIA) the image is first segmented into homogeneous objects and then classified.
- Vegetation, water, and built-up indices based on image bands.
- Supervised and unsupervised pixel-based classification.
- Object-based image analysis (OBIA) with preliminary image segmentation.
- Validation with independent samples and reporting of accuracy measures.
Change detection and time-series analysis
Change detection compares the cover or condition of an area between two or more dates. It can be based on differencing indices, differencing classifications, or direct comparison of bands. The chosen method has a direct consequence for the client’s question.
Time-series analysis, using multiple images and the revisit of an area, reveals the trend of change over a longer interval and is used to monitor vegetation, water bodies, and the expansion of built-up areas. In this analysis, consistency of the corrections applied across images is central to the validity of the result.
For applications related to ground displacement, interferometric methods based on radar imagery are applied; these methods need suitable data and control of several error sources, and the limitation to state is that the result depends strongly on the chosen image pair.
Deliverables and practical limitations
The deliverables of this line are usually raster and vector layers: land-cover maps, index and difference maps, and time-series outputs. Formats such as GeoTIFF, COG, and GeoPackage are used for raster and vector data, and metadata describing the method is provided with the product.
The limitations must be stated plainly. Cloud cover and cloud shadow in humid regions can restrict access to a cloud-free image; classification accuracy depends on the quality of the training samples; and the pixel-size difference between sources means that direct comparison of layers requires adjustment.
- A written method report, metadata, and delivery list.
- Data delivered in standard formats suited to the client’s system.
- Explicit statement of accuracy, image date, and data source for each layer.
Summary
Remote sensing, used correctly, provides a quantitative and repeatable picture of land cover and change, provided the sensor choice matches the client’s question and the full pre-processing chain is executed. Indices and classification extract information layers from the raw scene, and time-series analysis turns those layers into a monitoring tool.
The value of these services is not confined to the spatial output; recording the method, metadata, and limitations lets the client evaluate and reproduce the results. Awareness of the method’s boundaries, including cloud, shadow, and reference-data quality, is a precondition for correct use of the product.
To begin, it is enough to define the client’s question, the geographic extent, and the study period; on that basis the right sensor combination, processing level, and deliverable outputs can be determined.