Site Navigation

Section Navigation

The Urban Transition in Ghana and Its Relation to Land Cover and Land Use Change Through Analysis of Multi-scale and Multi-temporal Satellite Image Data

Interdisciplinary Research in Earth Science (IDS) program
NASA Award #: G00009708

Background

Most countries in sub-Saharan Africa remain in the midst of a major transformation from being predominantly rural to having a majority of people living in cities. This has enormous consequences for urban and rural land cover and land use (LCLU). As urbanization proceeds, food security concerns grow not simply in tandem, but at a higher rate because urban development implies a higher caloric intake per person meaning that agricultural output (or income to buy another country’s output) must grow at a higher rate than the population (Weeks 2011). Urban growth is likely to lead to sprawl, which will convert good agricultural land to urban purposes. Land beyond the former urban periphery may then transition to agriculture. This land cover and land use change (LCLUC) transition sequence of natural vegetation to agriculture to low density built environment to high density is likely to perpetuate sprawl for the major urban centers. The need to increase agricultural productivity demands that traditional subsistence agriculture be modernized to increase per hectare yield. This typically leads, somewhat paradoxically, to a lower demand for rural labor because higher yield requires mechanization rather than human labor.

Map of regional study area
Figure 1. Map of regional study area and study cities in Ghana.

In this research project, we expand the range and scope of the project team’s current work in Accra, Ghana, leveraging our existing resources to explicitly incorporate an analysis of LCLUC in the southern and most urban portion of Ghana, as shown in Figure 1. Based on the reasoning described above, the UN projects that Ghana's rural population will decline over the next several decades. All of the population growth in rural areas is likely to be absorbed by cities, as the redundant rural population seeks jobs in urban places. This rapid growth of cities, which are already characterized by a high degree of informality and a low level of infrastructure, means that LCLU in urban areas is constantly evolving, and that this transformation of the built environment will be associated with inequalities in the social, economic, and health conditions.

We hypothesize, then, that demographic and health outcomes in Ghana are strongly impacted by LCLUC. First, we hypothesize that the rapid increase in land consolidation, driven by international investments and measured remotely by a signature pattern shift from mottled mixed land use small farms to large industrial plantations is a key independent driver of urbanization vis a vis rural farm labor displacement. We also hypothesize that the quality of life in major Ghanaian cities, measured in terms of socio-economic and health variables, is driven largely by intra-urban LCLUC. The nominal study period is 1986 to 2010, with an emphasis on 2000 to 2010 for which demographic and census data are more complete. The realities of remote sensing and demographic data collection and data availability determine the actual period of analysis, which will vary slightly across the study area. The primary geographic unit of analysis will be the enumeration area (EA); equivalent to a census tract in the US, EAs range from less than a hectare in urban areas to several hundred square kilometers in rural areas.

Objectives

  1. To identify, map, and quantify LCLUC within an extensive study area of Ghana over 25 years (1986 through 2010),
  2. to understand the relationship between rural-to-urban migration as an outcome of LCLUC and concomitant drivers for the 2000 through 2010 period, and
  3. to assess LCULC and its effect on demographic and quality of life factors for four major Ghanaian urban centers during this time period.

Hypothesis

Virtually all of the major LCLUC in both rural and urban areas of Ghana are directly or indirectly driving patterns of inter-regional population change

Team

Our interdisciplinary team consists of remote sensing, human-environment, landscape, ecological, and social scientists from:

San Diego State University (lead)

Douglas Stow (PI), John Weeks (Co-PI), Lloyd (Pete) Coulter (Project Manager), Li An (Co-Investigator), Magdalena Benza-Fiocco (Post-doc), Sory Toure (PhD Student), Nicholas Ibanez (MS student), Helena Taflin (MA student), Ace Shih (MS student), Sean Taugher (MA student), Milo Verjaska (MS student)

The George Washington University

Ryan Engstrom (PI), Qin Yu (Post-doc), Avery Sandborn (MA student)

University of California Santa Barbara

David Lopez-Carr (PI), Cascade Tuholske (PhD student)

University of Ghana

Samuel Agyei-Mensah and Foster Mensah (CERSGIS) (Collaborators)

Study Area

There are four regions of interest for which LCLUC is being identified, quantified, and mapped: Greater Accra, Central, Ashanti, and Eastern. With these regions, LCLUC is being be mapped with the greatest detail for the cities of Accra, Kumasi, Obuasi, and Cape Coast. The cities are shown in red in Figure 1, and the four regions are shown are shown as light colored polygons, with Greater Accra at lower right, Central at lower left, Ashanti at upper left, and Eastern at upper right.

Land Cover and Land Use Classification

Land cover and land use is being mapped for c.2000 and c.2010 periods, and changes between these LCLU maps will be utilized (in conjunction with direct image-based change detection products) to identify LCLU changes of interest. Inter-regional identification of LCLUC is based on moderate spatial resolution, multi-temporal image data primarily from Landsat TM/ETM+ and LCDM OLI optical satellite systems (Figure 2). Where/when required, data from Terra ASTER, SPOT HRV, AWiFS and DMCii., and ERS-2 synthetic aperture radar (SAR) satellite systems will be also used. In addition, a map created by the Center for Remote Sensing and Geographic Information Services (CERSGIS) representing land cover conditions for 2000 is being utilized in conjunction with imagery for identifying c.2000 land cover conditions (Figure 3). The inter-regional LCLU classification scheme below is being used to classify moderate spatial resolution imagery across the four regions. Example inter-regional LCLU products are shown in Figures 4-6. These draft products are considered Version 1.0, and Version 2.0 (final) products are expected to be completed by February 2015.

Land cover image of Accra, Ghana
Figure 3. Center for Remote Sensing and Geographic Information Services (CERSGIS) 2000 Ghana land cover map. Cloud covered “no data” areas within the CERSGIS map were filled with MODIS land cover product (version 5.1) with classes defined by the 2000 International Geosphere Biosphere Programme (IGBP). A crosswalk was used to transform MODIS IGBP land cover classes to CERSGIS land cover classes.
LandSAT image of Accra, Ghana
Figure 2. Regional Landsat data for the Greater Accra region. This Landsat 7 image is from 26 Dec. 2002, and is one of the only cloud-free Landsat 7 images available during the study period for the Greater Accra region. The outer extent of the yellow polygons represent the Greater Accra region.

Inter-regional LCLU classification scheme

  1. Water
  2. Urban/built (city core, suburban, peri-urban, village)
  3. Forest
  4. Degraded Forest
  5. Agriculture
  6. Savanna Natural Vegetation
  7. Commercial Agriculture
  8. Barren
  9. Mining
LCLU for 2000 and 2010
Figure 4. Draft inter-regional land cover and land use (LCLU) maps for c. 2000 and c. 2010 for the four region study area.
LCLU for Accra, 2000 
          and 2010
Figure 5. Draft inter-regional land cover and land use (LCLU) maps for c. 2000 and c. 2010 for Accra and surrounding areas.
LCLU for Kumasi, 
          2000 and 2010
Figure 6. Draft inter-regional land cover and land use (LCLU) maps for c. 2000 and c. 2010 for Kumasi and surrounding areas.
High resolution imagery example
Figure 7. Example of high spatial resolution satellite imagery for the Accra area. IKONOS and QuickBird satellite images collected in 2000 and 2002 are shown.

High spatial resolution image data from QuickBird, IKONOS, GeoEye-1, and WorldView-2 commercial satellite systems are being utilized primarily for intra-urban mapping and analysis of LCLUC (Figure 7). The intra-urban LCLU classification scheme below is being used to classify the high spatial resolution imagery corresponding to the extents of the four cities. Example intra-urban LCLU image classification products are shown in Figure 8. These products are considered to be draft products as well. Final products will be published when available.

Intra-urban LCLU classification scheme

  1. Water
  2. Undeveloped
  3. Urban Non-residential
  4. Urban Residential
  5. Urban Agriculture
LCLU classification for Accra, 2010
Figure 8. Draft c. 2010 intra-urban land cover and land use (LCLU) map for a portion of Accra.

Analysis of Regional-scale Impacts of LCLUC on Migration, Demography, and Health

Unlike most developing countries, demographic, socioeconomic and health data for Ghana exist in higher quality and quantity than for most other developing countries, with national and nongovernment organization data collection programs starting in the early 1990s. We are utilizing quantitative spatial analysis techniques to examine relationships between LCLUC and magnitudes and changes of demographic, socioeconomic, and health variables. While EAs are the basic spatial analytical unit, analyses are also being conducted at multiple spatial scales, including: District (first level of agglomeration above EAs), Sub-regions (generated by merging EA units using socio-economic similarity and spatial autocorrelation measures, or field verified neighborhoods as in Figure 9), and Regions (state-level units). Relationships of LCLUC with demographic/socio-economic variables will be explored for EAs sampled in the 1993, 1998, 2003 and 2008 Demographic and Health Surveys, as well as for the 2000 and 2010 census data for all EAs within the study area.

Levels of 
          analysis
Figure 9. Hierarchical levels of analyses are shown, including: districts, neighborhoods, and enumeration areas (EAs).

Outcomes

Project Reports

Conference Papers & Presentations

Posters

Doctoral Dissertations

Master's Theses

Manuscripts Accepted and in Preparation

Data: Demographic, Socio-economic, and Health

Data: Imagery and Geographic Information Systems (GIS) Products

Last Updated January 12, 2015

This page is supported by San Diego State University International Population Center.
For more information please contact John R. Weeks at john.weeks@sdsu.edu.