This article explores technical aspects of using satellite data. It requires basic technical skills. If you are a farmer and are interested in how you can access satellite data analysis for your farm then you may be interested in this article “How Can I Use Satellite Data for My Farm.”
reading time: 80min – demo: https://dolifarm.com/satelite/
Introductions
These are the topics we will cover:
- Understand the Basics of Remote Sensing
- Learn About Satellite Data Sources
- Get Familiar with Data Formats
- Learn Data Processing and Analysis Techniques
- Master Programming for Remote Sensing
- Resources on the Net
1. Understand the Basics of Remote Sensing
Remote sensing is the science of obtaining information about objects or areas from a distance, typically from aircraft or satellites. Unlike in-situ data collection, where sensors are placed directly on or near the subject, remote sensing involves capturing data from afar.
Electromagnetic Spectrum and its Role in Remote Sensing
Electromagnetic Radiation (EMR): This is the energy that remote sensing instruments detect. It includes visible light, infrared, microwave, and radio waves.
Wavelength and Frequency:
- Wavelength: The distance between successive crests of a wave, typically measured in meters (m).
- Frequency: The number of waves that pass a point in one second, measured in hertz (Hz).
- These two properties are inversely related: Wavelength×Frequency=Speed of Light\text{Wavelength} \times \text{Frequency} = \text{Speed of Light}Wavelength×Frequency=Speed of Light.
Electromagnetic Spectrum and its Role in Remote Sensing
- Visible Spectrum (0.4 – 0.7 µm): The range of wavelengths that the human eye can see. This is used in optical remote sensing to create true-color images.
- Infrared Spectrum:
- Near-Infrared (NIR) (0.7 – 1.3 µm): Reflects vegetation strongly, useful for studying plant health.
- Shortwave Infrared (SWIR) (1.3 – 3 µm): Sensitive to moisture content and soil.
- Thermal Infrared (TIR) (8 – 14 µm): Measures emitted radiation from Earth’s surface, useful for temperature analysis.
- Microwave Spectrum:
- Passive Microwave: Detects naturally emitted microwave radiation, useful in soil moisture and sea surface temperature measurements.
- Active Microwave (Radar): Sends out pulses and measures the return signal, useful for topography, surface roughness, and moisture content.
Types of Remote Sensing
- Active Remote Sensing: The sensor emits its radiation and measures the reflection (e.g., Radar, LiDAR).
- Passive Remote Sensing: The sensor detects natural radiation that is emitted or reflected by the object or surrounding areas (e.g., sunlight reflection in optical sensors).
Types of Satellite Data
Understanding the different types of data satellites collect is crucial for selecting the right dataset for your analysis.
- Optical Imagery:
- Function: Captures images using visible and near-infrared light.
- Examples:
- Landsat Series: Provides multispectral data in the visible, NIR, SWIR, and TIR regions.
- Sentinel-2: Offers high-resolution optical imagery with additional spectral bands for detailed analysis.
- Radar Imagery:
- Function: Uses microwave signals to penetrate clouds and is independent of sunlight, providing data on surface structure and moisture.
- Examples:
- Sentinel-1: Provides C-band SAR (Synthetic Aperture Radar) data, useful for surface deformation, forest monitoring, and flood mapping.
- ALOS PALSAR: L-band SAR for deeper penetration, useful for forestry and biomass studies.
- Thermal Imagery:
- Function: Captures the emitted thermal radiation from the Earth, used to measure surface temperature.
- Examples:
- Landsat 8 Thermal Infrared Sensor (TIRS): Provides thermal infrared data for land surface temperature analysis.
Applications of Remote Sensing
The utility of remote sensing data spans various domains:
- Land Use and Land Cover (LULC) Mapping:
- Definition: Classifying the Earth’s surface into categories like urban areas, forests, water bodies, and agricultural lands.
- Example: Using Sentinel-2 data to monitor urban expansion and deforestation.
- Agriculture:
- Crop Monitoring: Monitoring crop health and predicting yield using vegetation indices like NDVI (Normalized Difference Vegetation Index).
- Soil Moisture Content: Analyzing soil moisture using SAR data to optimize irrigation practices.
- Disaster Monitoring:
- Flood Mapping: Using Sentinel-1 radar data to map flood extents even under cloud cover.
- Forest Fire Detection: Using MODIS or VIIRS thermal infrared data to detect and monitor wildfires.
- Climate and Environmental Studies:
- Climate Change: Tracking long-term changes in glaciers, sea level, and vegetation patterns using multi-temporal datasets.
- Urban Heat Islands: Analyzing urban temperature patterns using thermal infrared data to study urban heat islands.
- Oceanography:
- Sea Surface Temperature: Monitoring ocean temperatures using thermal infrared sensors, crucial for studying climate change.
- Chlorophyll Concentration: Analyzing ocean productivity and health using satellite data like MODIS or Sentinel-3.
- Forestry:
- Deforestation Monitoring: Using time-series optical and radar data to detect deforestation activities.
- Biomass Estimation: Estimating forest biomass using radar data, crucial for carbon accounting.
Resolution in Remote Sensing
Understanding resolution types is critical for selecting the appropriate data for analysis.
- Spatial Resolution:
- Definition: The smallest object that can be detected by the sensor. High spatial resolution (e.g., 10m) allows for detailed mapping, while low spatial resolution (e.g., 1km) is used for broader, regional studies.
- Temporal Resolution:
- Definition: The frequency at which a sensor revisits the same location. High temporal resolution is vital for monitoring changes over time, such as crop growth or deforestation.
- Spectral Resolution:
- Definition: The ability of a sensor to distinguish between different wavelengths. Higher spectral resolution allows for more detailed analysis of materials and land cover types.
- Radiometric Resolution:
- Definition: The sensitivity of a sensor to detect small differences in energy. Higher radiometric resolution means the sensor can detect finer differences in reflectance or emittance.
Satellite Orbits
The orbit of a satellite affects the type of data it can collect:
- Geostationary Orbit:
- Description: The satellite remains fixed over one point on the Earth’s surface, providing continuous coverage. Used for weather monitoring (e.g., GOES satellites).
- Polar Orbit:
- Description: The satellite passes over the Earth’s poles, allowing for global coverage as the Earth rotates underneath. Most Earth observation satellites (e.g., Landsat, Sentinel) are in near-polar orbits.
- Sun-Synchronous Orbit:
- Description: A type of polar orbit where the satellite passes over the same part of the Earth at approximately the same local solar time, ensuring consistent lighting conditions for optical sensors (e.g., Landsat, Sentinel-2).
2. Learn About Satellite Data Sources
Understanding where to find satellite data and how to access it is fundamental to working with remote sensing. Accessing and manipulating these datasets will require familiarity with the platforms mentioned below, as well as an understanding of the associated licensing and data formats.
Here we focus on remote sensing data services worldwide that provide open and free access, specifically tailored for agricultural applications
| Service | Provider | Access Platform | Applications |
|---|---|---|---|
| Sentinel-2 (Copernicus Program) | European Space Agency (ESA) | Copernicus Open Access Hub | Crop health monitoring, land cover classification, precision agriculture |
| Landsat Program | NASA/USGS | USGS Earth Explorer, Google Earth Engine | Crop mapping, yield prediction, soil moisture estimation |
| MODIS (Moderate Resolution Imaging Spectroradiometer) | NASA | NASA Earthdata, LP DAAC, Google Earth Engine | Vegetation index tracking, drought assessment, crop phenology studies |
| SMAP (Soil Moisture Active Passive) | NASA | NASA Earthdata, NSIDC, Google Earth Engine | Soil moisture monitoring, irrigation planning, drought monitoring |
| Copernicus Global Land Service (CGLS) | European Commission | Copernicus Global Land Service portal | Monitoring crop conditions, precision agriculture |
| GODAN (Global Open Data for Agriculture and Nutrition) | GODAN | GODAN website | Access to various open data resources for agriculture |
| CropScape (NASS Cropland Data Layer) | USDA National Agricultural Statistics Service (NASS) | CropScape website | Crop-specific land cover monitoring, land use analysis (USA) |
| OpenLandMap | OpenLandMap Consortium | OpenLandMap portal | Agricultural land assessment, soil fertility analysis, crop planning |
| World Soil Information (ISRIC) | ISRIC – World Soil Information | ISRIC Soil Data Hub | Soil management, agricultural planning, soil fertility assessment |
| Africover (FAO) | Food and Agriculture Organization (FAO) | FAO GeoNetwork | Land cover mapping, agricultural statistics (Africa) |
| JAXA Global ALOS Land Use/Land Cover Map | Japan Aerospace Exploration Agency (JAXA) | JAXA Global ALOS portal | Monitoring agricultural land use changes, land use planning |
| PRISM Climate Group | Oregon State University | PRISM Climate Group portal | Climate data for agricultural planning, crop modeling |
This table provides a concise overview of the open and free remote sensing data services that are specifically applicable to agriculture.
3. Get Familiar with Data Formats
Satellite data can come in various formats depending on the type of data (e.g., raster or vector), the satellite mission, and the processing level. Here’s an in-depth look at the most common data formats you’ll encounter
| Data Format | Description | Common Use Cases | Pros | Cons |
|---|---|---|---|---|
| GeoTIFF | A georeferenced version of the TIFF (Tagged Image File Format) that includes spatial metadata. | Satellite imagery, DEMs (Digital Elevation Models), and other geospatial raster data. | Widely supported, includes spatial metadata, allows for compression. | Can be large in size if uncompressed. |
| HDF (Hierarchical Data Format) | A file format for storing multidimensional data, often used in scientific datasets. | MODIS, VIIRS, and other NASA satellite data. | Supports large, complex datasets, hierarchical data organization. | Can be complex to work with, requires specific tools for manipulation. |
| NetCDF (Network Common Data Form) | A format for array-oriented scientific data, used widely in climate and atmospheric research. | Climate data, model outputs, environmental data. | Self-describing, platform-independent, efficient for large datasets. | Requires specific libraries for processing, can be complex. |
| Shapefile | A vector data format used for representing geographical features. A Shapefile is a collection of files, with .shp, .shx, and .dbf being the main ones. | Vector data, such as points, lines, and polygons for GIS applications. | Simple, widely supported, easy to use. | Limited attribute data storage, can become fragmented with many files. |
| KML (Keyhole Markup Language) | An XML-based format for representing geographic data, often used with Google Earth. | Sharing geographic data on web platforms like Google Earth. | Easy to share, integrates well with web-based applications. | Limited to simpler datasets, not suitable for large or complex data. |
| GeoJSON | A format based on JSON (JavaScript Object Notation) for encoding a variety of geographic data structures. | Web-based mapping applications, APIs, vector data. | Human-readable, easy to use with web technologies, lightweight. | Not ideal for large datasets, limited support for complex geometries. |
| GRIB (Gridded Binary) | A format commonly used for meteorological data, particularly model output. | Weather and climate models, real-time forecasting. | Compact, optimized for gridded data, widely used in meteorology. | Complex to work with, requires specialized tools. |
| BUFR (Binary Universal Form for the Representation of meteorological data) | A binary format for encoding meteorological data, used by the World Meteorological Organization (WMO). | Atmospheric data, weather observations. | Efficient for transmitting large datasets, handles complex data structures. | Requires specialized libraries for decoding and encoding. |
| LAS/LAZ | Formats used for LiDAR point cloud data, with LAZ being the compressed version. | LiDAR data processing, 3D modeling, terrain analysis. | High precision, supports large datasets, LAZ offers compression. | Requires specialized tools for processing and visualization. |
| HDF-EOS (Earth Observing System) | A variation of HDF specifically designed for storing Earth science data. | Satellite observations from NASA missions. | Combines the advantages of HDF with geospatial metadata, widely used in NASA datasets. | Complex, requires specific tools for manipulation. |
| ENVI | A proprietary format used by the ENVI software for storing raster data, including multispectral imagery. | Remote sensing image analysis, multispectral data. | Rich metadata support, integrates well with ENVI software. | Proprietary, less support outside the ENVI environment. |
| GML (Geography Markup Language) | An XML-based format for expressing geographic information, developed by the OGC (Open Geospatial Consortium). | Data interchange between GIS systems, web mapping services. | Flexible, supports complex data structures, widely adopted. | Verbose, can be inefficient for large datasets, requires XML parsing. |
| JP2 (JPEG 2000) | A raster image format that supports lossy and lossless compression, with geospatial metadata. | Satellite imagery, especially when storage space is a concern. | High compression efficiency, supports large images, includes geospatial metadata. | Not as widely supported as GeoTIFF, more complex compression algorithms. |
| Zarr | A newer format for chunked, compressed, N-dimensional arrays, increasingly used in large-scale scientific datasets. | Climate data, high-performance computing, cloud storage. | Optimized for cloud storage and parallel processing, scalable. | Still emerging, less support in traditional GIS software. |
4. Learn Data Processing and Analysis Techniques
Image preprocessing is crucial for ensuring that satellite data is accurately aligned, corrected for atmospheric disturbances, and ready for further analysis. Here are some of the key steps involved:
Once the data is preprocessed, various analytical techniques can be applied to extract meaningful information from the satellite imagery.
4.1 Image Preprocessing
Image preprocessing is crucial for ensuring that satellite data is accurately aligned, corrected for atmospheric disturbances, and ready for further analysis. Here are some of the key steps involved:
4.1.1 Atmospheric Correction
Atmospheric correction is necessary to remove the effects of the atmosphere on satellite images, which can distort the true reflectance values of the Earth’s surface.
- Purpose: To convert raw satellite data (top-of-atmosphere reflectance) to surface reflectance, which represents the true reflectivity of surface materials.
- Tools:
- Sen2Cor: A processor integrated with the Sentinel Application Platform (SNAP) that corrects Sentinel-2 data.
- 6S (Second Simulation of the Satellite Signal in the Solar Spectrum): A radiative transfer code used for atmospheric correction.
- ATCOR: A commercial tool for atmospheric correction of optical satellite data.
4.1.2 Georeferencing
Georeferencing ensures that satellite images align with geographic coordinates, allowing for accurate spatial analysis.
- Purpose: To map satellite images to a coordinate system, making them usable in GIS applications.
- Steps:
- Identify Ground Control Points (GCPs): Use known coordinates from a map or GPS.
- Apply Transformation: Fit the image to the coordinate system using transformation algorithms (e.g., polynomial, spline).
- Resampling: Interpolate pixel values to fit the new grid using nearest neighbor, bilinear, or cubic methods.
- Tools:
- QGIS: Provides a georeferencing plugin for aligning images.
- ArcGIS: Also has georeferencing capabilities.
4.1.3 Radiometric Correction
Radiometric correction adjusts the pixel values in an image to accurately represent the scene’s observed radiance or reflectance.
- Purpose: To correct sensor-related errors (e.g., striping, banding) and adjust for sun angle, sensor angle, and distance from Earth.
- Types:
- Absolute Radiometric Correction: Converts digital numbers (DNs) to physical units (radiance or reflectance).
- Relative Radiometric Correction: Adjusts for sensor noise and systematic errors between different bands or images.
- Tools:
- ENVI: Provides tools for radiometric calibration.
- IDL: Used for custom radiometric correction scripts.
4.2 Data Analysis
Once the data is preprocessed, various analytical techniques can be applied to extract meaningful information from the satellite imagery.
4.2.1 Vegetation Indices
Vegetation indices are algorithms applied to satellite imagery to highlight certain features, like vegetation health.
- NDVI (Normalized Difference Vegetation Index):
- Formula: NDVI=(NIR−RED)(NIR+RED)NDVI = \frac{(NIR – RED)}{(NIR + RED)}NDVI=(NIR+RED)(NIR−RED)
- Purpose: Measures vegetation health by comparing near-infrared (NIR) and red light reflectance.
- Applications: Agricultural monitoring, drought assessment, and forest cover analysis.
- Other Indices:
- EVI (Enhanced Vegetation Index): Improves sensitivity to high biomass and reduces atmospheric effects.
- SAVI (Soil Adjusted Vegetation Index): Adjusts for soil brightness in areas with sparse vegetation.
- Tools:
- QGIS and ArcGIS: Can calculate NDVI using raster calculator.
- Python (rasterio, NumPy): For custom vegetation index calculations.
- Google Earth Engine: Allows rapid computation of vegetation indices on large datasets.
4.2.2 Change Detection
Change detection techniques identify changes in the Earth’s surface over time by comparing satellite images taken at different times.
- Techniques:
- Image Differencing: Subtract pixel values of two images taken at different times.
- Change Vector Analysis (CVA): Analyzes the magnitude and direction of change in spectral space.
- Post-Classification Comparison: Classify each image separately and then compare the results.
- Applications:
- Urban Expansion: Monitor growth of cities.
- Deforestation: Detect loss of forest cover.
- Disaster Impact: Assess damage from natural disasters like floods or earthquakes.
- Tools:
- QGIS and ArcGIS: Provide tools for raster differencing and classification.
- SNAP: Useful for change detection in Sentinel data.
- Python (NumPy, rasterio): For custom change detection scripts.
4.2.3 Image Classification
Image classification categorizes pixels in an image into different land cover classes, such as water, forest, urban areas, etc.
- Types:
- Supervised Classification: Requires user-defined training data (e.g., Maximum Likelihood, Random Forest).
- Unsupervised Classification: The algorithm groups pixels into clusters without user-defined classes (e.g., K-means, ISODATA).
- Object-Based Image Analysis (OBIA): Segments the image into meaningful objects before classification.
- Steps:
- Collect Training Data: For supervised classification, select representative pixels for each class.
- Apply Classification Algorithm: Use statistical or machine learning methods to classify pixels.
- Accuracy Assessment: Evaluate classification accuracy using metrics like the confusion matrix and overall accuracy.
- Tools:
- QGIS: Semi-Automatic Classification Plugin (SCP) for supervised classification.
- ArcGIS: Has a range of tools for both supervised and unsupervised classification.
- Python (scikit-learn, TensorFlow, PyTorch): For advanced machine learning-based classification.
- Google Earth Engine: Offers easy implementation of classification algorithms on large datasets.
4.2.4 Spectral Unmixing
Spectral unmixing decomposes a pixel’s spectrum into its constituent materials, which is useful when the pixel represents a mixture of different surface types.
- Purpose: To estimate the proportion of different land cover types within a pixel.
- Steps:
- Select Endmembers: Identify pure spectral signatures of materials in the image (e.g., water, vegetation, soil).
- Apply Unmixing Algorithm: Decompose the mixed pixel spectrum into its component endmembers.
- Tools:
- ENVI: Provides built-in tools for spectral unmixing.
- Python (NumPy): For implementing linear spectral unmixing manually.
4.3 Software Tools for Data Processing and Analysis
Various software tools are available for processing and analyzing satellite data. Each has its own strengths and is suited to different tasks. Here we provide a list of tools opensource or free that offer a wide range of functionalities for agricultural applications, including vegetation monitoring, crop classification, land cover analysis, and more
| Software Tool | Description | Key Features | Website |
|---|---|---|---|
| QGIS | Open-source GIS software for handling spatial data and raster analysis. | Vegetation index calculation, land cover classification, image preprocessing. | QGIS |
| GDAL | Geospatial Data Abstraction Library for reading and writing raster and vector data. | Data format conversion, raster manipulation, and processing. | GDAL |
| SAGA GIS | Free GIS software with extensive tools for geospatial data analysis. | Terrain analysis, vegetation classification, hydrological modeling. | SAGA GIS |
| GRASS GIS | Open-source GIS software for geospatial data management and analysis. | Raster and vector data processing, spatial modeling, vegetation analysis. | GRASS GIS |
| SNAP | Sentinel Application Platform for processing Sentinel satellite data. | Atmospheric correction, image classification, time-series analysis. | SNAP |
| Google Earth Engine (GEE) | Cloud-based platform for large-scale geospatial data processing. | Vegetation index analysis, change detection, agricultural monitoring. | Google Earth Engine |
| OpenCV | Open-source computer vision library for image processing. | Feature extraction, object detection, image classification. | OpenCV |
| R | Free programming language and software environment for statistical computing and graphics. | Statistical analysis, data visualization, remote sensing data processing. | R Project |
| Python (with libraries) | Free programming language with libraries for data processing. | Vegetation indices, machine learning models, image analysis. | Python |
| Jupyter Notebook | Open-source web application for creating and sharing documents with code, equations, and visualizations. | Data analysis, interactive visualization, code sharing. | Jupyter |
4.4 Advanced Techniques
Once you are comfortable with basic data processing and analysis, you can explore more advanced techniques to further enhance your capabilities.
4.4.1 Time-Series Analysis
Analyze temporal changes using multi-temporal satellite data to study phenomena like seasonal vegetation changes or long-term environmental trends.
- Techniques:
- Trend Analysis: Evaluate long-term trends in time-series data (e.g., linear regression).
- Seasonal Decomposition: Separate seasonal patterns from trends and noise.
- Harmonic Analysis: Model cyclic patterns in time-series data.
- Tools:
- Python (pandas, statsmodels): For time-series analysis.
- Google Earth Engine: For handling large-scale time-series data.
4.4.2 Machine Learning and AI
Apply machine learning algorithms to classify land cover, predict changes, or extract features from satellite images.
- Algorithms:
- Random Forest: Popular for supervised classification.
- Support Vector Machines (SVM): Effective for high-dimensional data.
- Deep Learning (CNNs): For complex tasks like object detection in high-resolution images.
- Tools:
- scikit-learn: For traditional machine learning algorithms.
- TensorFlow/Keras, PyTorch: For deep learning models.
- Google Earth Engine: Integrates with TensorFlow for scalable AI applications.
4.4.3 Data Fusion
Combine data from different sources (e.g., optical and radar) to improve analysis accuracy.
- Techniques:
- Pixel-Level Fusion: Combine data at the pixel level to create composite images.
- Feature-Level Fusion: Merge features extracted from different datasets before analysis.
- Decision-Level Fusion: Integrate results from separate analyses to make final decisions.
- Tools:
- SNAP: For Sentinel-1 and Sentinel-2 data fusion.
- ENVI: Provides tools for data fusion and analysis.
- Python (NumPy, rasterio): For custom fusion algorithms.
4.5 Practical Example Workflow
Here’s a practical workflow that incorporates several of the techniques discussed:
- Data Acquisition: Download Sentinel-2 data from the Copernicus Open Access Hub.
- Preprocessing:
- Apply atmospheric correction using Sen2Cor.
- Georeference the images if needed using QGIS.
- Perform radiometric correction to adjust for sensor errors.
- Analysis:
- Calculate NDVI to assess vegetation health.
- Use change detection techniques to monitor deforestation over time.
- Classify land cover using Random Forest in QGIS or Google Earth Engine.
- Advanced Techniques:
- Conduct a time-series analysis of NDVI values to study seasonal vegetation changes.
- Apply machine learning models to predict land cover changes.
- Perform data fusion with Sentinel-1 radar data for improved classification accuracy.
- Visualization and Reporting:
- Visualize results in QGIS or using Python libraries like Matplotlib.
- Share your findings using Jupyter Notebooks or through interactive maps on platforms like ArcGIS Online.