Santiago Andes

Dictionary of Key Concepts

In this section, you will find definitions and explanations of the most relevant terms in statistical modeling, GIS, and remote sensing. The goal is to provide you with tools to make informed decisions. We want you to understand the applications and limitations of each concept, and to see how they can help you solve problems and improve your projects. Our commitment is to offer reliable and transparent information, explained in a simple yet rigorous way. We believe that an informed client knows exactly what they are hiring and why they need it. We hope this "Dictionary of Key Concepts" helps you become familiar with the technical language and make the most of the services we offer.

Precision Agriculture
An approach that applies information technologies, remote sensing, and GPS to manage specific areas of the field, optimizing inputs and reducing environmental impact.
Algorithm
A set of logical steps that transform input data into results; its efficiency determines the speed and scalability of a system.
API (Application Programming Interface)
A set of definitions that allows different applications or services to communicate with each other, facilitating integration and separation of responsibilities.
Model Fitting
The process of training a model by finding the best parameters to describe relationships between variables or to classify/predict accurately.
Back-end (Web Development)
The part of an application that runs on the server: business logic, databases, and security, responding to requests from the front-end.
Relational vs NoSQL Databases
Relational databases use tables and SQL; NoSQL databases use documents, graphs, or wide columns, offering flexibility and scalability depending on the use case.
Supervised vs Unsupervised Classification
Supervised classification is trained using labeled examples; unsupervised classification groups similar data without labels.
Cloud Computing
Access to computing resources via the internet without managing your own infrastructure, enabling scalability and cost savings.
Bias Correction
Adjusting data or models to eliminate systematic errors that distort conclusions, improving accuracy and fairness.
Training Data
A portion of data used to teach a model; if the data is biased, the model will learn incorrectly.
Test Data
A separate set of data used to objectively evaluate a trained model's performance and detect overfitting.
Spatial Data
Information that includes location or geometry; essential for mapping and geospatial analysis.
Data Science
A collection of techniques for extracting knowledge from large datasets, combining statistics, programming, and visualization.
Deep Learning
A subfield of machine learning that uses deep neural networks capable of discovering complex patterns in data.
DEM (Digital Elevation Model)
A raster that represents the elevation of the Earth's surface; useful in topography and hydrology.
Downscaling
A technique used to increase the spatial or temporal resolution of data or models, refining climate projections.
Extrapolation
Applying a model to areas or ranges outside of its training data, often leading to less reliable predictions.
Front-end (Web Development)
The visual part of an application that users interact with in their browsers; includes interface design and logic.
GCM (Global Climate Model)
A mathematical model that simulates global climate dynamics for long-term projections.
Geostatistics
A branch of statistics focused on spatial phenomena, using tools like variograms and kriging for prediction.
Geomatics
Integrates methods for collecting, analyzing, and representing geographic information.
Geospatial Data Science
Applies data science to georeferenced information, combining remote sensing, GIS, and programming.
Google Earth Engine (GEE)
A cloud platform with a vast satellite archive and tools for processing it at a global scale.
Hyperspectral
Sensors that capture dozens or hundreds of narrow bands, enabling precise material identification.
Satellite Imagery
Visual capture of the Earth's surface from orbital sensors at various resolutions.
Spectral Index
Band combinations used to highlight properties such as vegetation (NDVI) or moisture (NDWI).
Artificial Intelligence (AI)
A field that develops systems capable of tasks requiring human intelligence; depends on quality data and expertise to be useful.
Model Interpretability
The degree to which a human can understand how a model makes decisions; greater in simpler models.
Interpolation
Estimating values at unsampled points within a region, assuming spatial continuity.
Landsat
A NASA-USGS program capturing multispectral images since 1972 at 30 m resolution.
LIDAR (Light Detection and Ranging)
Technology that uses laser pulses to obtain high-precision 3D point clouds.
LLM (Large Language Model)
An AI model trained on large text corpora, capable of generating coherent language (e.g. GPT).
Machine Learning
Methods that learn patterns from data for predictions without explicit rules.
Accuracy Metrics
Indicators (accuracy, precision, recall) that evaluate a model's correctness in classification.
Error Metrics
Quantify the deviation between prediction and reality in regression (MSE, RMSE, MAE).
Spatial Predictive Model
A model that estimates variables over an area using environmental data and statistical techniques.
Model (in AI / Statistics)
A mathematical representation learned from data; not to be confused with the algorithm that trains it.
MODIS
A multispectral sensor on the Terra and Aqua satellites, with 36 bands and daily global revisit.
Pixel
The smallest unit of an image; in remote sensing, it represents a ground area with spectral values.
Python
A popular programming language for data science, ML, and automation, with libraries like NumPy and Pandas.
R
A language and environment for statistical analysis and data science with powerful libraries (ggplot2, sf).
Random Forest
An ensemble algorithm based on multiple decision trees, robust and effective with noisy data.
Raster
A data format based on a grid of cells; each cell stores continuous values such as reflectance or elevation.
Neural Network
A brain-inspired algorithm with layers of connected neurons for detecting complex patterns.
Spatial Resolution
The ground area covered by each pixel in an image; determines the level of detail.
Spectral Resolution
The number and width of bands a sensor captures; hyperspectral sensors have many narrow bands.
Radiometric Resolution
The ability to distinguish levels of energy; measured in bits (8 bits = 256 levels).
Temporal Resolution
The frequency at which a sensor captures images of the same area, for example every 5 days (Sentinel-2).
Statistical Significance
The probability that a result is not due to chance; usually measured with a p-value.
SAR (Synthetic Aperture Radar)
A microwave sensor that operates day and night and penetrates clouds, measuring backscatter.
Sentinel-1
SAR satellites from the Copernicus programme with 5-20 m resolution and 6-12 day revisit.
Sentinel-2
Multispectral optical satellites (13 bands) with 10-60 m resolution and ~5 day revisit.
Time Series
A sequence of data measured at successive intervals, useful for detecting trends and patterns.
Bias
A tendency or distortion in data that produces unrepresentative results; must be corrected for reliable predictions.
GIS (Geographic Information System)
Tools for collecting, storing, analyzing, and visualizing geographic data with coordinates.
Overfitting
When a model memorizes the training set and fails to generalize to new data.
Remote Sensing
Obtaining information about the Earth's surface through remote sensors such as satellites or drones.
Vector
A GIS format that represents points, lines, or polygons; useful for discrete entities.
XGBoost
A high-performance gradient boosting algorithm based on decision trees.