Urban and rural landuse mapping of parts of Ekiti State using NigeriaSat1 imagery

1
Urban and rural landuse mapping of parts of Ekiti
State using NigeriaSat-1 imagery

• National Workshop
• on
• Satellite Remote Sensing (NigeriaSat-1) and GIS
  A solution to sustainable National Development
  Challenges
• 15th -17th June, 2004 Le Meridien Hotel, Abuja
• Ademola Omojola, Ph.D

2
1. Introduction

• Landuse and landcover (LU/LC) data constitutes a
  vital environmental data whether for urban or
  rural areas
• The data set has been implicated in the worlds
  major environmental issues and constitutes one of
  the best indicators for environmental monitoring
• It is equally a vital component of any nations
  Spatial Data Infrastructure (SDI)
• The dearth of this data in the nation (spatially,
  temporally, non-standardised) is well known and
  one can easily appreciate its effect on
  environmental decision making
• Although the data set could be generated through
  sources such as ground surveys and assessment
  rolls, the ease and practicability of use of
  remotely sensed data, most especially the
  satellite data sets, is well acknowledged.

3
Introduction .. contd .

• The ease of RS use for LU/LC mapping was
  responsible for the formulation of several
  classification schemes that could be used with
  remotely sensed data nationally and regionally
  around the world e.g. U.S.A, U.K, The
  Netherlands, South Africa
• Although Nigeria is yet to nationally formulate
  or adopt a classification scheme for LU/LC
  mapping (for standardisation and comparisons),
  efforts at having an NSDI in place, and the
  launching of the NgeriaSat-1 makes the generation
  of this vital data more promising
• Thus, there is to assess the potentials of
  NigeriaSat-1 data in generating this data set.

4
2. Objectives

• An evaluation of NigeriaSat-1 in the generation
  of urban and rural landuse and landcover data in
  parts of Ekiti State

5
3. The Study Area

• Ekiti State is located in the south western part
  of Nigeria
• About 5,860 sq km
• The test data however covers mostly 4 LGAs
  Ijero, Ekiti West, Efon and Ekiti South-West

Nigeria Showing Ekiti State
Hill shading of Ekiti State derived from 150,000
map contours
6
4a. Materials

• NigeriaSat-1 satellite data (32X32m S.R) -
• primary image data
• 150,000 Topographical map sheets intended for
  geo-rectification and as interpretation
• 1250,000 Vegetation and Landuse map - collateral
  data
• Standard Garmin 12 GPS
• geo-rectification and field study
• Image Processing and GIS software

7
4b. Method

• LU/LC mapping from remotely-sensed data is well
  documented. While it may be easy to clearly
  differentiate between digital and visual mapping
  procedures, the initial digital image processing
  procedure and the ease and frequent use of
  heads-up delineation of classes makes even the
  visual mapping more of a hybrid process in
  practice.
• Thus, in evaluating NigerSat-1 data in parts of
  Ekiti State for LU/LC mapping, the major primary
  level LU/LC classes -- Built-up Area, Wetland,
  Water Body, Agricultural, Natural/Semi-Natural
  cover and Bare surface -- were investigated. The
  following analysis were carried out
• Urban and rural landuse and landcover review in
  Ekiti State
• Digital Image Analysis
• Visual landuse landcover mapping
• Digital landuse landcover mapping
• Unsupervised classification
• Supervised classification

8
5. RESULTS AND DISCUSSION

• Urban and rural landuse and landcover review in
  Ekiti State reconnaissance

9
5. RESULTS AND DISCUSSION ..contd..

• (b) Digital Image Analysis
• The NigerSat-1 data used was subjected to
  standard image pre-processing procedures visual
  inspection and image statistics display,
  enhancement and geometric correction.
• The visual inspection review of the data revealed
  an inherent banding or stripping in the image
  supplied. This stripping greatly reduced the
  visual quality of the data.
• The lack of a very convincing pattern of the
  stripes with the image statistics computation and
  display could not allow an appropriate
  de-striping technique so far (now described as
  error of band co-registration shift?)
• The low contrast on the image covering the study
  area was equally subjected to a standard linear
  stretching to aid good visual study of the data
• The dated nature of the topographical map series
  made them almost unsuitable for the selection of
  Ground Control Points (GCPs) for image
  geo-rectification. The recognition and choice of
  control points on the images were also hindered
  by the visual quality of the image used.
• In the absence of any material (image or map) for
  geo-rectification, a standard Garmin 12 GPS was
  used in the field to collect ground coordinates
  for the geo-rectification . The justification
  was based on the fact that their accuracy are
  mostly within the sub-pixel resolution of the
  satellite data if carefully used

10
The Coverage of the NigerSat-1 data for the study
in Ekiti State Georectified and linearly-stretched
11
(No Transcript)
12
5. RESULTS AND DISCUSSION ..contd..

• (c) Visual landuse landcover mapping
• The data used was a cubic convolution re-sampled
  data product of the geo-rectification.

Water
Gallery/Fringing Forest
The mapping (vector delineation )done heads-up
within a GIS for a built-up area as an example
Bare surface
Mosaic of Forest and tree crop agric
Agric land open
13
5. RESULTS AND DISCUSSION ..contd..

• (d) Digital landuse landcover mapping
• The data used was the nearest-neighbour
  re-sampled data product of the geo-rectification.
  
• Analysis was restricted to an unsupervised
  classification.
• Previous experience using supervised
  classification in this area for agricultural
  mapping has not been very successful. This due to
  mixed pixel effect attributable to the nature of
  small, non-contiguous agric land holdings, a
  mosaic of tree-cropping within forests and and
  the different phenological stages of plants
  culminating in complex mix-pixels relative to the
  medium spatial resolution satellites.
• Other primary classes could also be generated by
  post clustering regrouping of the unsupervised
  classes

14
A post-clustering grouping (from 16 classes) to
generate built-up areas
15
6. CONCLUSIONS

• Image geo-rectification may be difficult using
  topo map sheets as most features that could be
  easily recognised and used for rectification are
  new eg road network and not on the nearly 40
  year-old maps. Handheld GPS sets if used with
  care can be utised.
• LU/LC theme visual mapping at the primary class
  level is derivable manually from the image even
  up to a scale of 1150,000 scale. This may
  however be improved to 1100,000 scale if
  adequate image pre-processing is performed on the
  image to fully exploit its inherent spatial
  resolution
• Unsupervised classification of the primary level
  proves promising in digital mapping of landuse
  and landcover using the data sets in the study
  area
• There is still the need to perfect image
  pre-processing techniques to enhance or remove
  the visible stripping effect currently on the
  image challenge to local image processing
  practitioners
• This type of study will definitely help in the
  formulation of standardised LU/LC schemes that
  will make environmental monitoring and change
  detection more operational