Glossary

Earth Observation & AI/ML Terms

Date

November 17, 2025

How to Use This Glossary

This glossary defines key terms used throughout the CoPhil EO AI/ML Training Programme. Terms are organized alphabetically within categories for easy reference.

Categories: - Earth Observation Terms - AI/ML Terms - Geospatial Data Terms - Satellite & Sensor Terms - Philippine EO Organizations - Technical Acronyms


Earth Observation Terms

Absorption Band

Wavelength region where atmospheric gases (water vapor, oxygen, CO2) absorb electromagnetic radiation, limiting remote sensing capabilities.

Backscatter

The portion of radar energy reflected back to the sensor from a target. Used in SAR imaging to detect surface properties and moisture.

Cloud Masking

The process of identifying and removing cloud-contaminated pixels from optical satellite imagery to improve data quality.

Composite Image

A single image created by combining multiple images from different dates, often using statistical methods (median, mean) to reduce noise and clouds.

Earth Observation (EO)

The gathering of information about Earth’s physical, chemical, and biological systems through remote sensing technologies, primarily satellites.

False Color Composite

An image display where spectral bands are assigned to RGB colors differently than natural vision (e.g., NIR-Red-Green), revealing features invisible to the human eye.

Ground Truth

Field-collected reference data used to validate remote sensing classifications and train machine learning models.

Image Collection

A set of satellite images covering the same geographic area at different times, used for time series analysis.

Pixel

The smallest unit in a raster image, representing a specific ground area (spatial resolution) and spectral value.

Preprocessing

Steps taken to correct raw satellite data before analysis, including atmospheric correction, geometric correction, and radiometric calibration.

Remote Sensing

The science of obtaining information about objects or areas from a distance, typically using sensors on satellites or aircraft.

Revisit Time

The frequency with which a satellite can observe the same location on Earth (e.g., Sentinel-2 has 5-day revisit with 3 satellites).

Spectral Signature

The unique reflectance pattern of an object across different wavelengths, used to identify materials and land cover types.

True Color Composite

An image display using red, green, and blue bands to create a natural-looking image similar to human vision.


AI/ML Terms

Activation Function

Mathematical function in neural networks that introduces non-linearity (e.g., ReLU, Sigmoid, Tanh), enabling learning of complex patterns.

Artificial Intelligence (AI)

Computer systems capable of performing tasks that typically require human intelligence, including perception, reasoning, and decision-making.

Backpropagation

Algorithm for training neural networks by calculating gradients of loss and adjusting weights to minimize error.

Batch Size

Number of training samples processed before updating model weights. Smaller batches = more updates but noisier; larger batches = smoother but fewer updates.

Classification

Supervised learning task of assigning input data to predefined categories (e.g., forest, water, urban).

Clustering

Unsupervised learning technique that groups similar data points together without predefined labels (e.g., K-means, DBSCAN).

Confusion Matrix

Table showing predicted vs. actual classifications, used to calculate accuracy, precision, recall, and F1-score.

Convolutional Neural Network (CNN)

Deep learning architecture specialized for image analysis, using convolutional layers to detect spatial patterns.

Data Augmentation

Technique to artificially increase training data by applying transformations (rotation, flipping, scaling) to existing samples.

Deep Learning

Subset of machine learning using multi-layer neural networks to learn hierarchical representations of data.

Epoch

One complete pass through the entire training dataset during model training.

Feature Engineering

The process of creating new input variables from raw data to improve model performance.

Feature Extraction

Identifying and extracting relevant patterns or characteristics from raw data for use in machine learning models.

Ground Truth Labels

Verified, accurate labels for training data, typically from field surveys or expert interpretation.

Hyperparameter

Model configuration setting chosen before training (e.g., learning rate, number of layers) that affects model performance.

Loss Function

Mathematical function measuring the difference between predicted and actual values, used to guide model training.

Machine Learning (ML)

Subset of AI enabling systems to learn and improve from experience without explicit programming.

Neural Network

Computing system inspired by biological brains, consisting of interconnected nodes (neurons) organized in layers.

Overfitting

When a model learns training data too well, including noise, resulting in poor performance on new data.

Precision

Proportion of positive predictions that are actually correct. Precision = TP / (TP + FP).

Random Forest

Ensemble learning method using multiple decision trees to improve prediction accuracy and reduce overfitting.

Recall (Sensitivity)

Proportion of actual positives correctly identified. Recall = TP / (TP + FN).

Regression

Supervised learning task of predicting continuous numerical values (e.g., crop yield, temperature).

Supervised Learning

Machine learning where models learn from labeled training data (input-output pairs).

Support Vector Machine (SVM)

Classification algorithm that finds the optimal hyperplane separating different classes in feature space.

Training Set

Portion of data used to train a machine learning model (typically 70-80% of total data).

Transfer Learning

Reusing a pre-trained model on a new but related task, reducing training time and data requirements.

Underfitting

When a model is too simple to capture the underlying patterns in data, resulting in poor performance.

Unsupervised Learning

Machine learning where models find patterns in unlabeled data without predefined categories.

Validation Set

Data used to evaluate model performance during training and tune hyperparameters (typically 10-15% of data).


Geospatial Data Terms

Affine Transformation

Mathematical operation describing the relationship between pixel coordinates and geographic coordinates in raster data.

Bounding Box

Rectangular area defined by minimum and maximum coordinates [minX, minY, maxX, maxY], used to specify geographic extents.

Coordinate Reference System (CRS)

System defining how coordinates relate to real-world locations, including datum and projection (e.g., WGS84, UTM).

Digital Elevation Model (DEM)

Raster representation of terrain elevation, with each pixel value representing height above a reference level.

Feature

In geospatial terms, a vector object (point, line, or polygon) with associated attributes.

GeoJSON

Open standard JSON format for encoding geographic data structures, widely used for web mapping.

GeoPackage

Open format for geospatial data storage in SQLite database, supporting both vector and raster data.

Geometry

The spatial component of a geographic feature, defining its shape and location (point, line, polygon).

GeoTIFF

Raster image format with embedded geographic metadata (CRS, extent, resolution), standard for geospatial raster data.

NoData Value

Special value in raster data indicating missing or invalid data (e.g., -9999, NaN).

Pixel Resolution (Spatial Resolution)

Ground area represented by one pixel (e.g., 10m means each pixel covers 10m × 10m on the ground).

Projection

Mathematical transformation converting 3D Earth coordinates to 2D map coordinates (e.g., Mercator, UTM).

Raster Data

Grid-based spatial data where each cell (pixel) contains a value, used for continuous phenomena (elevation, temperature, imagery).

Reproject

Converting geospatial data from one coordinate reference system to another.

Shapefile

Popular vector data format for GIS, consisting of multiple files (.shp, .shx, .dbf, .prj).

Spatial Join

Combining attributes from two geospatial datasets based on their spatial relationship (intersection, within, etc.).

Vector Data

Spatial data representing discrete features as points, lines, or polygons with associated attributes.

Well-Known Text (WKT)

Text markup language for representing vector geometry and spatial reference systems.


Satellite & Sensor Terms

Active Sensor

Sensor that emits its own energy and measures the reflected signal (e.g., SAR, LiDAR).

Aperture

Opening in a sensor that controls the amount of light collected, affecting image brightness and quality.

Atmospheric Correction

Processing step removing atmospheric effects (scattering, absorption) to retrieve surface reflectance.

C-band

Radar frequency band (4-8 GHz, wavelength 3.75-7.5 cm) used by Sentinel-1, good for vegetation and soil moisture.

Electromagnetic Spectrum

Range of all electromagnetic radiation wavelengths, from radio waves to gamma rays, including visible light.

Geometric Correction

Correcting image distortions caused by sensor viewing angle, terrain, and Earth’s curvature.

Level 1C (L1C)

Sentinel-2 product with Top-of-Atmosphere (TOA) reflectance, geometrically corrected.

Level 2A (L2A)

Sentinel-2 product with Bottom-of-Atmosphere (BOA) surface reflectance, atmospherically corrected.

LiDAR

Light Detection and Ranging - active sensor using laser pulses to measure distance, creating high-resolution 3D point clouds.

Multispectral

Imaging system capturing data in multiple (typically 3-15) wavelength bands across visible and infrared spectrum.

Near Infrared (NIR)

Electromagnetic radiation with wavelengths 0.7-1.4 μm, strongly reflected by healthy vegetation.

Optical Sensor

Passive sensor detecting reflected sunlight in visible and infrared wavelengths (e.g., Sentinel-2, Landsat).

Orbit

Path of a satellite around Earth, characterized by altitude, inclination, and period.

Panchromatic

Single-band imagery capturing all visible wavelengths, typically at higher spatial resolution than multispectral bands.

Passive Sensor

Sensor detecting naturally available energy, typically reflected sunlight (e.g., optical cameras).

Polarization

Orientation of radar waves (HH, VV, HV, VH), providing information about surface structure and moisture.

Radiometric Calibration

Converting raw sensor digital numbers to physical units (radiance or reflectance).

SAR (Synthetic Aperture Radar)

Active microwave imaging system creating high-resolution images independent of sunlight and clouds.

Short-Wave Infrared (SWIR)

Electromagnetic radiation with wavelengths 1.4-3.0 μm, useful for detecting moisture and minerals.

Spectral Resolution

Ability to distinguish between different wavelengths, determined by number and width of spectral bands.

Sun-Synchronous Orbit

Satellite orbit maintaining constant local solar time, ensuring consistent illumination conditions.

Swath Width

Width of the ground strip imaged by a satellite in a single pass (e.g., Sentinel-2 has 290 km swath).

Temporal Resolution

Frequency of repeat observations over the same location (synonymous with revisit time).

Thermal Infrared (TIR)

Electromagnetic radiation with wavelengths 8-14 μm, measuring heat emitted by Earth’s surface.


Philippine EO Organizations

CoPhil Programme

EU-Philippines Copernicus Capacity Support Programme - partnership to strengthen Philippine EO capabilities, establish a Copernicus Mirror Site, and develop AI/ML capacity.

DENR

Department of Environment and Natural Resources - responsible for forest monitoring, land cover mapping, and natural resource management using EO data.

DOST

Department of Science and Technology - leads science and technology advancement, co-chairs CoPhil programme, operates PAGASA and ASTI.

DOST-ASTI

Advanced Science and Technology Institute - ICT research and development, AI platforms (SkAI-Pinas, DIMER, AIPI).

LiPAD

LiDAR Portal for Archiving and Distribution - repository of high-resolution elevation data for the Philippines.

NAMRIA

National Mapping and Resource Information Authority - official mapping agency, operates GeoPortal and Resource Data Analysis Center.

NDRRMC

National Disaster Risk Reduction and Management Council - coordinates disaster response, uses EO data for assessment and planning.

PAGASA

Philippine Atmospheric, Geophysical and Astronomical Services Administration - weather, climate, and astronomical services.

PhilGIS

Philippine GIS Data Clearinghouse - repository of geospatial datasets for the Philippines.

PhilSA

Philippine Space Agency - national space authority, co-chairs CoPhil programme, operates SIYASAT portal and develops Copernicus Mirror Site.

SIYASAT Portal

PhilSA’s secure data archive for NovaSAR-1 data and maritime monitoring products.


Technical Acronyms

AI

Artificial Intelligence - computer systems performing tasks requiring human intelligence.

API

Application Programming Interface - set of functions allowing software to interact with services (e.g., Earth Engine Python API).

ASTER

Advanced Spaceborne Thermal Emission and Reflection Radiometer - NASA sensor providing multispectral imagery.

BOA

Bottom of Atmosphere - surface reflectance after atmospheric correction (Level 2A products).

CCA

Climate Change Adaptation - strategies and actions to adjust to climate change impacts.

CNN

Convolutional Neural Network - deep learning architecture for image analysis.

CRS

Coordinate Reference System - defines how coordinates relate to Earth’s surface.

DEM

Digital Elevation Model - raster representation of terrain elevation.

DL

Deep Learning - subset of ML using multi-layer neural networks.

DRR

Disaster Risk Reduction - strategies to minimize disaster impacts.

EO

Earth Observation - gathering information about Earth through remote sensing.

ESA

European Space Agency - operates Copernicus Sentinel missions.

EVI

Enhanced Vegetation Index - vegetation index less sensitive to atmospheric effects than NDVI.

GEE

Google Earth Engine - cloud platform for planetary-scale geospatial analysis.

GIS

Geographic Information System - software for capturing, managing, and analyzing spatial data.

GPU

Graphics Processing Unit - hardware accelerating deep learning computations.

GNSS

Global Navigation Satellite System - satellite-based positioning (GPS, Galileo, GLONASS).

HDF

Hierarchical Data Format - file format for storing large scientific datasets.

IW

Interferometric Wide Swath - Sentinel-1 acquisition mode with 250 km swath.

JAXA

Japan Aerospace Exploration Agency - operates Japanese Earth observation satellites.

LiDAR

Light Detection and Ranging - laser-based remote sensing for elevation mapping.

LULC

Land Use Land Cover - classification of Earth’s surface into categories (forest, urban, agriculture, etc.).

ML

Machine Learning - algorithms enabling systems to learn from data.

MODIS

Moderate Resolution Imaging Spectroradiometer - NASA sensor providing daily global coverage.

NASA

National Aeronautics and Space Administration - operates Landsat and other EO missions.

NDBI

Normalized Difference Built-up Index - spectral index for detecting built-up areas.

NDVI

Normalized Difference Vegetation Index - spectral index measuring vegetation health/density.

NDWI

Normalized Difference Water Index - spectral index for detecting water bodies.

NIR

Near Infrared - electromagnetic radiation beyond visible red (0.7-1.4 μm).

NRM

Natural Resource Management - sustainable management of natural resources using EO monitoring.

RGB

Red-Green-Blue - color model using three primary colors, or the corresponding image bands.

RF

Random Forest - ensemble machine learning algorithm using multiple decision trees.

RL

Reinforcement Learning - ML paradigm where agents learn through interaction with environment.

SAR

Synthetic Aperture Radar - active microwave imaging system.

SCL

Scene Classification Layer - Sentinel-2 cloud and land cover classification mask.

SNAP

Sentinel Application Platform - ESA’s free software for processing Sentinel data.

SWIR

Short-Wave Infrared - electromagnetic radiation with wavelengths 1.4-3.0 μm.

SVM

Support Vector Machine - classification algorithm finding optimal separating hyperplane.

TIR

Thermal Infrared - electromagnetic radiation measuring surface temperature (8-14 μm).

TOA

Top of Atmosphere - apparent reflectance before atmospheric correction (Level 1C products).

USGS

United States Geological Survey - distributes Landsat and other EO data.

UTM

Universal Transverse Mercator - widely-used projected coordinate system dividing Earth into zones.

WGS84

World Geodetic System 1984 - global geographic coordinate system (EPSG:4326).


Spectral Indices

NDVI (Normalized Difference Vegetation Index)

Formula: (NIR - Red) / (NIR + Red)

Range: -1 to +1

Interpretation: - High values (0.6-0.9): Dense vegetation - Moderate (0.2-0.5): Sparse vegetation - Near zero: Bare soil, rock - Negative: Water, clouds

Sentinel-2 Bands: (B8 - B4) / (B8 + B4)


NDWI (Normalized Difference Water Index)

McFeeters Formula: (Green - NIR) / (Green + NIR)

Gao Formula: (NIR - SWIR) / (NIR + SWIR)

Range: -1 to +1

Interpretation: - Positive values: Water bodies - Negative: Land surfaces

Sentinel-2 (McFeeters): (B3 - B8) / (B3 + B8)

Sentinel-2 (Gao): (B8 - B11) / (B8 + B11)


NDBI (Normalized Difference Built-up Index)

Formula: (SWIR - NIR) / (SWIR + NIR)

Range: -1 to +1

Interpretation: - Positive values: Built-up areas - Negative: Vegetation, water

Sentinel-2: (B11 - B8) / (B11 + B8)


EVI (Enhanced Vegetation Index)

Formula: 2.5 × ((NIR - Red) / (NIR + 6×Red - 7.5×Blue + 1))

Range: -1 to +1

Advantages: Less sensitive to atmospheric effects and soil background than NDVI

Sentinel-2: 2.5 × ((B8 - B4) / (B8 + 6×B4 - 7.5×B2 + 1))


Quick Reference Tables

Sentinel-2 Band Summary

Band Name Wavelength (nm) Resolution (m) Typical Use
B1 Coastal aerosol 443 60 Atmospheric correction
B2 Blue 490 10 Bathymetry, soil/vegetation
B3 Green 560 10 Peak vegetation sensitivity
B4 Red 665 10 Vegetation discrimination
B5 Red Edge 1 705 20 Vegetation health
B6 Red Edge 2 740 20 Vegetation stress
B7 Red Edge 3 783 20 Vegetation stress
B8 NIR 842 10 Biomass, water bodies
B8A Narrow NIR 865 20 Atmospheric correction
B9 Water vapor 945 60 Atmospheric correction
B11 SWIR 1 1610 20 Moisture, soil/vegetation
B12 SWIR 2 2190 20 Moisture, burned areas

Common CRS for Philippines

Name EPSG Code Type Use Case
WGS84 4326 Geographic Global datasets, web maps
UTM Zone 50N 32650 Projected Western Mindanao
UTM Zone 51N 32651 Projected Luzon, Visayas, most of Philippines
UTM Zone 52N 32652 Projected Eastern Mindanao
PRS92 4683 Geographic Philippine Reference System