Digital Earth Basics

DEvision: Overview of module sections

Geospatial approaches across disciplines and industries

Remotely sensed imagery is one of the 'macroscope' instruments to monitor the state and dynamics of Earth's surface.

As a first step we will explore basic online platforms (beyond Google) for working with satellite imagery, and get to appreciate the diversity of imagery characteristics as well as application domains.

Re-view >  this introduction  ? Following this section overview, now continue below with study materials in the learning platform! Checking the Activities / Tasks section beforehand might be helpful.


Measuring global dimensions and local spaces

Remote Sensing measures Earth surface characteristics from a distance. It therefore needs a way for contact-free measurements, which is provided by electromagnetic (em) radiation.

Re-view >  this introduction  ? Following this section overview, now continue below with study materials in the learning platform! Checking the Activities / Tasks section beforehand might be helpful.


Spatial reference systems

In this section, we will explore the basics of documenting location on the surface of Earth with cartesian coordinate system - moving from longitude / latitude to x'es and y's.

Globe to plane ...

...

Why flatten Earth?

Right, a virtual globe avoids many problems of flat maps, distortions and more ...

Still, if we want to see Earth's surface entirely in one view, or need to represent it on a flat map, screen or projections - while globe views today are easily available, we do have use cases where a spherical surface needs flattening.

Thus, this section helps us explore basics of map projections.

Why do we need more than one map projection?

'Projections' use geometrical surfaces like a plane ('azimutal projection', or those which can easily be spread out on a plane without further distortions, like a cone or a cylinder.

In addition, 'mathematical projections' are purely numerical transformations without a geometrical model or metapher.

Repurposing a phrase: 'the great thing about projections is that we have so many of them' - serving different regions, scales and purposes!

We can't have it all ...

Map projections serve different purposes, thus different characteristics need to be emphasized. Projections can truthfully represent at scale either

  • areas - equal area projections
  • distances - equidistant projections
  • angles - conformal projections

but never all of these at the same time. How to check projection characteristics? Well, either you simply know, or interpret Tissot's indicatrix ... find out!

Universal Transversal Mercator

The widely used 'UTM' projection deserves our special attention and therefore will be discussed in more detail ...

Re-view >  this introduction  ? Following this section overview, now continue below with study materials in the learning platform! Checking the Activities / Tasks section beforehand might be helpful.


Collecting georeferenced data

Imagery allows a broad spectrum of interpretations, depending on 'readers' interests, experience and tasks to be completed.

Image enhancements, like improving contrast for better perception of details, support interpretation and human-based image analysis.

More importantly, though, we need to find algorithmic approaches to make the extraction of semantic information reproducible. Essentially, we aim at using imagery for the automatic generation of thematic layers.

Image correction and pre-processing

Individual imagery always is influenced by current lighting angles (for passive RS), atmospheric conditions and differences between sensors. To make images comparable - e.g. for change detection, and to allow smooth mosaicking over larger areas, several adjustments are necessary:

  • geometric correction - rectification
  • atmospheric correction, incl top-of-atmosphere correction
  • topographic normalisation

To a large degree, steps for pre-processing are taken by image suppliers providing different 'levels' of imagery. Compare e.g. Sentinel imagery level 1C vs 2A.

Topographic normalisation

In hilly and mountainous regions, different incidence angles of lighting on surface facets cause the biggest differences in pixel brightness.

This effect either can be explicitely removed with a kind of 'reverse hillshading', or by working with proportions between band values instead of band values >calculating spectral indices.

Indices

Spectral indices are using the ratio between bands to remove or at least reduce topographic and seasonal (solar) lighting effects.

A popular example is the Normalized Difference Vegetation Index:

NDVI: (NIR-Red) / (NIR+Red) Sentinel 2 NDVI: (Band 8 - Band 4) / (Band 8 + Band 4)

While NDVI provides a clear view on the amount of (photosynthetically) active vegetation in an AoI, other indices focus on water, urban/impervious surfaces or other topics of interest.

Thresholding

Any spectral index provides a continuous quantitative metric for a particular theme, a degree-of-something, but not yet definitive categories like the type or state of vegetation, the presence of surface water or a category of built-up land.

For that purpose, we need to find an index value threshold indicative of a particular category, e.g. allowing the delineation of water bodies and furthermore the monitoring of changes in e.g. flooded areas.

This is a first, simple approach to land cover classification. After exploring this, we will dive somewhat deeper in the following section.

Re-view >  this introduction  ? Following this section overview, now continue below with study materials in the learning platform! Checking the Activities / Tasks section beforehand might be helpful.


From maps to views: communicating and sharing spatial information

Maps used to be considered as definitive documents: transmitting information from the author, via cartographer to readers, from sender to recipient.

From a Digital Earth perspective, 'maps' are dynamic views on (geospatial) databases or data streams. These views support communication and interaction all around a location-centric paradigm.

Most views are generated with sharing in mind. To simply communicate, collect opinions, to direct, understand and support decisions.

Re-view >  this introduction  ? Following this section overview, now continue below with study materials in the learning platform! Checking the Activities / Tasks section beforehand might be helpful.


Getting started with the geospatial cloud

One essential characteristic of modern GIS as one core technology underlying 'Digital Earth' is its cloud-centric architecture. Locally installed desktop applications still are important in many professional settings, but browser-based access to the bulk of features faciliates learning, distributed setups and pervasive access to geospatial functionality.

Wayback imagery

Similar to the Web's  Wayback Machine , archived EO imagery is widely available through several portals, like e.g. through the  Living Atlas .

This allows us to explore and compare several snapshots along timelines, with full global coverage available already for more than half a century since the launch of the first  Landsat  mission in 1972.

Visual comparison

For exploratory analysis of temporal dynamics and to detect change, visual tools are widely available in remote sensing portal environments, like:

  • swipe
  • flicker
  • fade
  • time lapse animation
  • lens

These tools also allow comparisons with (base) maps, to detect differences between a definitive, document representation and the state at the time of image acquisition.

Single-theme monitoring

Depending on the question at hand, either all (land cover) changes in a target region are of interest, or the dynamics of a single thematic layer like water, urban area, or forest coverage are of interest.

The example at the right is provided by 'Global Forest Watch', a worldwide partnership monitoring forest coverage aiming at action.

Two different types of changes can be observed in each region, at each pixel: increase or decrease in forest cover. In a more general case, a transition matrix between two different points in time will summarize changes:

OBIA in monitoring changes

In all cases where change typically does not happen in a continuous regional environment but occurs related to structures or other discrete entities, an object-based approach to change detection is most adequate.

This would be the case with building damage assessment due to conflicts or natural disasters, urban sprawl monitoring, or informal settlement or refugee camp dynamics (example at the right).

Looking into the future

Detecting change by comparing past with present conditions is important to understand the causes underlying historic and current dynamics. These also inform about the future, though - we definitely want and need to project changed factors and processes into the future.

Development of scenarios and assessment of potential trends are indispensable instruments for making policy decisions.

Looking at the past must only be a means of understanding more clearly what and who they are so that they can more wisely build the future. - Paulo Freire

Re-view >  this introduction  ? Following this section overview, now continue below with study materials in the learning platform! Checking the Activities / Tasks section beforehand might be helpful.