You are here: LearnEO! » Lessons » Lesson 9: Land Cover Mapping

Previous lesson Previous lesson

Classification with multi- and hyper-spectral data from satellites and airborne optical sensors

By F. Del Frate and M. Picchiani

Tor Vergata University, Italy

Land cover is a critical variable that links many parts of the human and physical environment. Accurate and up-to-date information on land cover is required for a plethora of applications, including land resource planning, studies of environmental change and biodiversity conservation. Realistically, remote sensing is the only feasible source of information on land cover over large areas that allows data to be acquired at a regularly repeatable manner. However, depending on spatial and spectral resolution; or multi-angle acquisition capabilities, rather different remote sensing systems exist. It is therefore important to understand the advantages and limitations of the system configuration being considered before deriving any information on land cover.

Scheme for land-cover classification from multi-band data.
Land cover classification generic scheme: from multi-spectral data to the map. The classification algorithm associates the values of each pixel in the available bands (i.e. Red, Green, Blue, Near-Infrared_1, Near_Infrared_2, Thermal-Infrared) with the land cover classes through a decision rule.

IMPORTANT NOTE: This lesson requires Bilko 3.4 from February 2013 or later, as earlier versions of the software can not open and display all data as described in the lesson.


Despite the huge potential of remote sensing as a source of information on land cover and the long history of research into the extraction of land cover information from remotely sensed imagery, many problems have been encountered, and the accuracy of land cover maps derived from remotely sensed imagery has often been viewed as not optimal for operational users. Many factors may be responsible for the problems encountered and the resulting inaccuracy. These include the data characteristic obtained by the remote sensing system (e.g., its spatial and spectral resolutions) or the methods used to extract the land cover information from the imagery (e.g., classification methods).

In this lesson attention is focused on both such issues. The methods of analysis and exploitation of multispectral and hyperspectral data for land cover mapping will be illustrated. In addition, information exploitable with multi-angular acquisitions will be considered. The lesson will also provide the concepts for the application of standard and Neural Network based land cover mapping techniques and the basics for carrying out validation procedures.

Lesson Overview

Aim and objectives

The lesson shows how to generate land cover maps from Earth Observation data taking into account the different characteristics of the data available such as spatial or spectral resolution.

At the end of the lesson you should be able to

Lesson content

Lesson is accordingly divided into 6 sections:

  1. Understanding pixel classification
  2. Using images at different spatial resolution
  3. Using images of different spectral resolution
  4. Information contained in multi-angular acquisitions
  5. Generate your land cover map
  6. Validate your map

Data and tools for this lesson

Files needed to complete the lesson activities

All data and tools needed to complete the lesson activities are listed below.

Data in their original format CHRIS.Proba level 1A data covering the South-East area of Rome, acquired on 28 February 2006. The image has 18 spectral bands acquired at five different viewing angles. The file contains the following data consisting of the different acquisition angles defined as Fly Zenith Angle (FZA):

MER_RR__2PQBCM20060202_093651_000002482044_00437_20531_0003.N1: MERIS Level 2 reduced resolution image of Italy, acquired on 02 February 2006. The Level 2 data are provided after atmospheric correction and geophysical processing in order to obtain the surface reflectances and some geophysical parameters. The image contains 13 spectral bands containing the surface reflectances, at 1.2 km spatial resolution, that will be analysed in the lesson.

Supplementary and pre-processed data

AHS_20060606a_b34.set: AHS image of agricultural areas surrounding the city of Demmin (Germany), acquired on 06 June 2006. The data is composed of 76 spectral bands from visible (441 nm) to thermal infrared (13.17 μm). The data have been pre-processed (cropped in order to include only the area of interest) and saved as Bilko set (provided as file) consisting 76 bz2-compressed images, with the same core file name as the set, and number from #001 to #076. These must be kept in the same folder as the set.
Note that Bilko can read bz2-compressed data, so there is not need to uncompress the individual files in the set.

LT51940222006162KIS00.set: Landsat 5 image of agricultural areas surrounding the city of Demmin (Germany), acquired on 12 July 2006; bands 1-6. The data have been pre-processed (cropped in order to include only the area of interest) and saved as Bilko set (provided as file) consisting six bz2-compressed images with the same core name as the set, and numbered from #001 to #006. These must be kept in the same folder as the set.

Software tools

There are no additional Bilko formulae or other tools required for this lesson. However, if you wish to compare the classification done in Bilko with other methods, you may download the Neumapper software as an 'extra' tool. This will allow you to compare different classification methods.

Downloading the lesson

The lesson downloads contain everything you need to complete the lesson. This includes the data and tools listed above, and three PDF documents:

To download the lesson and data you must be a registered user. If you have not yet done so you can register here. Once your registration is confirmed you can download the lesson after submitting your registered e-mail address below. This will take you to the download page where you can select the documents, tools and data you wish to download.

Please enter your registered e-mail address: