GISdevelopment.net --> Proceedings --> GISDECO
Remote Sensing and GIS for Good Governance: Analysis of High Spatial Resolution Ikonos Imagery for Surveying Agricultural activities in the city of Ouagadougou, Burkina Faso
Ilona Kemeling
Laboratory for Geo-information and Remote Sensing, Wageningen University
P.O. Box 47, 6700 AA Wageningen, The Netherlands
Tel: +31317 474 640; Fax: +31317 474 567;
Email: Ilona.Kemeling@Student.GIRS.WAU.NL
Steven M. de Jong & Pieter B.M. van Teeffelen
Faculty of Geographical Sciences, Utrecht University
P.O. Box 80.115, 3508 TC Utrecht The Netherlands
Tel: +3130 253 2749; Fax: +3130 253 1145;
Email: s.dejong@geog.uu.nl
Leo M. van den Berg & Gerbert J.Roerink
Alterra, Wageningen University & Reseacrh Centre
P.O. Box 47, 6700 AA Wageningen, The Netherlands
Email: l.m.vandenberg@alterra.wag-ur.nl, G.J.Roerink@Alterra.wag-ur.nl
Abstract
In West African cities, urban and peri-urban agricultural activities form a major economic activity for a significant group of inhabitants. Intensive and commercial-oriented horticulture is one of the most important activities in the cities such as Ouagadougou and Bamako. Horticulture activities form a fragile balance with waste management (city waste is used as manure for horticulture) and with other agricultural activities such as livestock raising. Local governments consider these agricultural activities as a loss of valuable city-space and often deny their economic importance. Consequently, horticulturists located within the city are often forced to stop their activities or to relocate their business to the urban fringe or beyond. Unfortunately, no well-organised resettlement programmes for these activities exist in the West African countries. The government does often not recognise the importance of such programmes due to a lack of awareness of the importance of horticultural activities in the cities as a widespread economic activity, in combination with city waste recycling. Often no reliable information is available on the extent of the horticultural activities in the city.
One of the objectives of this study is to use high-resolution satellite images to survey the agricultural activities in the cities and to use time-series of images to determine the spatial dynamics of these activities. Only since 1999, satellite images are available at a resolution of 1 to 4 m potentially enabling us to monitor these small scale agricultural lots. Previous available images (SPOT-XS and Landsat TM) having pixels of 20 meters or more do not provide sufficient detail to map these activities properly. In this study we investigated the value of high spatial resolution IKONOS imagery to survey agricultural activities within the city and at the urban fringe. IKONOS has a pixel size of 4 by 4 meters and has 4 spectral bands: blue, green, red and near infrared. Traditional spectral classification methods, based on ground truth sets and spectral differences between the crops, failed to identify individual activities. This was due to spectral overlap of the thematic land cover types. The agricultural activities occur mainly near water bodies because of irrigation requirements and near roads (pistes) to facilitate transport. Adding this type of GIS stored information to the classification process improved the results considerably.
1. Management of (peri-) urban agriculture and waste
The government does often not recognise the importance of (peri)- urban agricultural activities. Apart from being a major economic activity for a significant group of inhabitants, (peri)- urban agriculture plays an important role in the urban ecosystem in various ways. (Peri-) urban agriculture contributes to food variety and health of the urban population (Aldington, 1997; Leisinger and Schmitt, 1995). Nutrient recycling in the form of composting reduces the ecological footprints of towns. Some typical climate related problems faced in (semi-) arid tropical cities are air-pollution, caused by dust and flooding, caused by high intensity rainfall. Urban green can improve the microclimate by providing soil cover for soil and water conservation. Often no reliable information is available on the extent of the horticultural activities in the city.
Possible problems of (peri)- urban agriculture like health problems related to water quality and the use of city wastes as organic fertiliser make planning and integration of agricultural and waste management even more important. For urban planning up-to-date maps of the urban area are essential (DGUT, 1993). Due to the fast growth of the cities, there is a lack of accurate information. Remote sensing in combination with field checks forms a cost-effective way to map (peri-) urban agriculture. Remote sensing data can be applied in a standardised way using each time the same method and the same land cover legend, it can be used to update maps frequently and it can easily be expanded to other areas. Although the use of satellite information in developing countries seems very promising, case studies in the particular developing countries are scarce but very necessary (Pollé and Boogaard, 1996). Burkina Faso is a Sahel country with a tropical semi-arid climate. Ouagadougou is the capital of Burkina Faso and the population was estimated in 1996 to exist of 1.8 million people but continues to grow very fast. As a result the farmers located in the inner fringe of the city are forced to move out (Club Du Sahel, 2000; Lompo et al., 2000). The Greater Ouagadougou lies in the main cereal production area of the West African Savannah, producing millet and sorghum (Kassam, 1976).
Annual precipitation is not very reliable but volumes are generally between 600 and 900 mm per year. The project 'Recycling Urban Waste in Urban Agriculture Production; Participatory Technology Development In Bamako and Ouagadougou' (APUGEDU) is an EU-funded urban development program. The APUGEDU project requires maps of (peri-) urban agricultural activities to estimate its total surface area and locations to compute compost needs derived from domestic waste (Van den Berg et al., 2001). The first objective of this study is to answer the question of the APUGEDU project: what surface area of Ouagadougou is used for (peri-) urban agriculture? To answer this question a hybrid classification method is developed using land cover data extracted the IKONOS-2 image and other spatial data sets. A second objective of this study is to investigate and evaluate the use of remote sensing, more specifically the use of the IKONOS-2 image in a development project. The accuracy of the resulting map is evaluated using land cover data collected in the field. The technical and human resources are described and possibilities to repeat the study at local institutes are assessed. Specific advantages and disadvantages of IKONOS-2 imagery in the context of the APUGEDU-project are discussed.
2 Using IKONOS-imagery to map (peri-) urban agriculture
In the framework of the APUGEDU project an IKONOS-2 image of 7 June 2000 was purchased to evaluate its value for land use analysis and urban planning. The resolution of IKONOS-2 data is spatially high with pixels of 4m x 4m, but spectrally relatively low. Ikonos has four spectral bands in blue, green, red and near infrared. The first three bands are highly correlated. The IKONOS image is thought to be the most suitable for areas with short-range spatial variation, like a city area. To complement the IKONOS-2 data a SPOT-XS image, acquired in May 1997 is used. The spectral window of SPOT-XS is comparable to that of IKONOS-2 with three bands in the visible green and red and one in the near infrared part of the electromagnetic spectrum. The spatial resolution is much coarser with each pixel representing an area of 20m x 20m. Other spatial data used for this include (1) a tourist map, (2) a digital land register from Ouagadougou, (3) a map giving the sections for waste collection, and (4) an urban horticulture map. As part of the APUGEDU-project, INERA carried out a detailed characterisation study in four selected agricultural sites in Ouagadougou (Lompo et al., 2000).
Collection of field data is carried out one year after the registration date of the IKONOS-2 image. The field data is collected in Ouagadougou, according to a prepared fieldwork plan in co-operation with the remote sensing-centre of INERA. The basis of this study is that the spectral and spatial variation in the image corresponds with the actual physical variations found in the field. At 279 locations in the field thematic information on the land situation is collected using a GPS to determine the geographical co-ordinates. The accuracy of the GPS device is measured at the time of registration and lies between ___ six and nine meter. The thematic data filled in on the registration form are physical characteristics that may have an effect on the reflectance values such as landuse, degradational state, vegetation cover and type, waste cover, soil type, soil colour and soil management.
The extent of the IKONOS-2 image of 11 by 11 km does not cover the total city area and some parts of the IKONOS-2 image are not usable due to cloud contamination. Merging the IKONOS-2 image with the available SPOT-XS image improves the extent and provides data in the areas excluded due to cloud contamination. The merging method applied in this study is the Principal Component Analysis (PCA) and reverse PCA (Lillesand and Kiefer, 2000). The combined-image, which is used for further analysis in this research, is the result of this data integration of IKONOS-2 and SPOT-XS. The combined-image gives the same information as the original IKONOS-2 image with additional information on places where data was missing. The spatial resolution of the combined-image image enables visual recognition of objects larger than four meter in most of the image and enables visual recognition of objects larger than twenty meters in the filled up patches. Roads and river beds can easily be recognised and are digitised manually on-screen for further analysis. A vector file of roads and house was created and will be used later to compute spatial buffers in a GIS environment to assess the extent of certain agricultural activities. Houses appeared to be clearly visible on an IKONOS image when a spectral transformation was applied normally referred to as the Tasseld Cap brightness.
First, a straightforward spectral classification of the IKONOS image was carried out using a conventional maximum likelihood classifier and ground truth collected in the field. Unfortunately, this approach failed to separate the agricultural areas and individual agricultural crops. Therefore, alternative methods to analyse and classify the image were investigated. One approach that proved to be successful comprised to include GIS stored information about 'distance to water', distance to roads (piste)' in the analysis and classification of the image. Four assumptions derived from field observations were included in the alternative classification procedure and are shown in figure 1. Based on these four 'rules it was possible to restrict the potential agricultural area within the city of Ouagadougou based on the following conditions (figure 1):
3. (Peri-) urban agriculture in Ouagadougou
The research and approached followed in this study resulted in various types of land use maps. The APUGEDU project demands a map giving specific information on the location and size of waste application in peri- (urban) agriculture, either unimproved waste or compost. As it was not possible to classify agricultural land use and type of waste used on these fields in a very reliable way on the basis of IKONOS, we decided to produce a probability map with four classes representing the use of waste in urban and peri-urban agricultural activities in Ouagadougou. All four classes are 'agricultural area using waste' resulting from the described spectral classification in combination with the GIS-based classification classified (table 1). The first division is made by separating the area inside the buffer around the rivers and barrages. These areas have access to irrigation during the wet season contrary to the extensive agricultural areas. The second division is based on spectral classification. Areas that have a spectral signature of vegetation, meaning that they are covered by green plants, are classified as 'certain'. Areas with a spectral signature indicating no presence of vegetation are classified as 'potential'. The reason why areas without vegetation features are marked as 'potential' is the nature of agriculture. With one IKONOS image, it is impossible to determine whether a bare plot is a fallow field or permanent bare soil. Overall accuracy of the map lies between 54.7% if the certain and potential classes are merged and 65.5% if the potential classes are considered as non-agricultural. This leads to the four classes presented in table 1.
Table 1: Description of the four classes of map "Utilisation des déchets en agriculture urbaine et péri
urbaine ŕ Ouagadougou", Use of waste in urban and peri-urban agriculture in Ouagadougou.
Figure 2: Map representing the use of waste for agriculture activities (see also table 1)
4. Discussion and conclusions
Conventional spectral classification of IKONOS images proved not to result in reliable maps of agricultural activities and 'use of waste maps' within Ouagadougou despite the high spatial resolution of this sensor. Therefore, we developed a new and advanced hybrid approach to include GIS-stored information in the classification process. This information included distance to irrigation water, distance to roads and distance to houses. This approach resulted in a set of probability maps of agricultural activities and use of waste as shown in figure 2. A disadvantage of the new proposed hybrid classification method is that it does not make optimal use of the information captured by the IKONOS-2 data. In the procedure only reflectance information is used. A promising approach might be to include the morphological properties of agricultural land such as shape, size, and patterns in the classification procedure. Conventional remote sensing methods classify an image on the basis of spectral data without accounting for spatial patterns. In an urban environment, not only spectral patterns are important but also spatial patterns are useful to identify certain objects (De Jong et al., 2000; Van Deursen et al., 1999). To make full use of the information content of an IKONOS image in an urban environment it might be useful to use classification methods based on spatial patterns, such as eCognition (Bauer and Steinnocher, 2001).
Other conclusions that came out of this study are that for a cost-effective use of the IKONOS image there are two important conditions. Firstly, the institute assigned to work with the image should have access to sufficient hardware and software and sufficient human resources for analysis and fieldwork. The large size of the IKONOS image file formed a limitation for processing in Ouagadougou.
Analysis and application of such a large data set requires considerable technological resources as well as human resources. Secondly, the data should be used for as many applications as is permitted by the license and the results should be made accessible to all potential users. To make data accessible a system is required to store and order these data, such a system is called a Spatial Data Infrastructure (SDI). A properly functioning SDI requires long-term investments. Collection of spatial data useable for environmental research is still a very costly and time-consuming process. A hardware/software system, organisational structure, human resources and data sets have to be established.
The role that local partners such as INERA can play in the development of an infrastructure for environmental data for Burkina Faso is large since it is a governmental institute responsible for collection, analysis and distribution of environmental data. Unfortunately, lack of resources is the reason many other institutes have invested in data collection and processing. These data are not regarded as a common resource for the country. This leads to isolated research, duplication of work and unnecessary costs. Data collected and processed with fiscal or foreign aid resources should be used to benefit the country. Data and resources like computers should be shared and be available to the person that needs it most for research. Users and producers of spatial data should form an information community sharing common digital language and data feature definitions and working together on problems (Prévost, et al., 1997). The benefit of using an IKONOS image is now mainly determined by the activities in The Netherlands, since the local institutes lacked human and technical resources to work with the image. Long-term investments are required to train the local staff and provide hardware and software systems to work with satellite images. Data and information can then be used more efficiently and easier be accessed. When these conditions are not assured, the use of IKONOS to classify (peri)-urban agriculture will not lead to significantly better results than aerial photography. In fact, the latter will be a cheaper solution and have benefits over the IKONOS image.
Another problem encountered during this study was the dynamic characteristics of agricultural land use (FAO, 1981). This dynamic character made it very difficult to identify (peri-) urban agriculture using a single, mono-temporal image. (Peri-) urban agriculture is often not characterised by a certain spectral or spatial pattern but by the temporal dynamics in both spectral as spatial pattern. In future, images taken during different moments in the growing season should be combined into one classification.
Acknowledgements
The research project presented in this paper is financially supported by the INCO programme of the European Commission under contract no. Contract ERBIC18-CT98-0288 (APUGEDU).
References
Remote Sensing and GIS for Good Governance: Analysis of High Spatial Resolution Ikonos Imagery for Surveying Agricultural activities in the city of Ouagadougou, Burkina Faso
Ilona Kemeling
Laboratory for Geo-information and Remote Sensing, Wageningen University
P.O. Box 47, 6700 AA Wageningen, The Netherlands
Tel: +31317 474 640; Fax: +31317 474 567;
Email: Ilona.Kemeling@Student.GIRS.WAU.NL
Steven M. de Jong & Pieter B.M. van Teeffelen
Faculty of Geographical Sciences, Utrecht University
P.O. Box 80.115, 3508 TC Utrecht The Netherlands
Tel: +3130 253 2749; Fax: +3130 253 1145;
Email: s.dejong@geog.uu.nl
Leo M. van den Berg & Gerbert J.Roerink
Alterra, Wageningen University & Reseacrh Centre
P.O. Box 47, 6700 AA Wageningen, The Netherlands
Email: l.m.vandenberg@alterra.wag-ur.nl, G.J.Roerink@Alterra.wag-ur.nl
Abstract
In West African cities, urban and peri-urban agricultural activities form a major economic activity for a significant group of inhabitants. Intensive and commercial-oriented horticulture is one of the most important activities in the cities such as Ouagadougou and Bamako. Horticulture activities form a fragile balance with waste management (city waste is used as manure for horticulture) and with other agricultural activities such as livestock raising. Local governments consider these agricultural activities as a loss of valuable city-space and often deny their economic importance. Consequently, horticulturists located within the city are often forced to stop their activities or to relocate their business to the urban fringe or beyond. Unfortunately, no well-organised resettlement programmes for these activities exist in the West African countries. The government does often not recognise the importance of such programmes due to a lack of awareness of the importance of horticultural activities in the cities as a widespread economic activity, in combination with city waste recycling. Often no reliable information is available on the extent of the horticultural activities in the city.
One of the objectives of this study is to use high-resolution satellite images to survey the agricultural activities in the cities and to use time-series of images to determine the spatial dynamics of these activities. Only since 1999, satellite images are available at a resolution of 1 to 4 m potentially enabling us to monitor these small scale agricultural lots. Previous available images (SPOT-XS and Landsat TM) having pixels of 20 meters or more do not provide sufficient detail to map these activities properly. In this study we investigated the value of high spatial resolution IKONOS imagery to survey agricultural activities within the city and at the urban fringe. IKONOS has a pixel size of 4 by 4 meters and has 4 spectral bands: blue, green, red and near infrared. Traditional spectral classification methods, based on ground truth sets and spectral differences between the crops, failed to identify individual activities. This was due to spectral overlap of the thematic land cover types. The agricultural activities occur mainly near water bodies because of irrigation requirements and near roads (pistes) to facilitate transport. Adding this type of GIS stored information to the classification process improved the results considerably.
1. Management of (peri-) urban agriculture and waste
The government does often not recognise the importance of (peri)- urban agricultural activities. Apart from being a major economic activity for a significant group of inhabitants, (peri)- urban agriculture plays an important role in the urban ecosystem in various ways. (Peri-) urban agriculture contributes to food variety and health of the urban population (Aldington, 1997; Leisinger and Schmitt, 1995). Nutrient recycling in the form of composting reduces the ecological footprints of towns. Some typical climate related problems faced in (semi-) arid tropical cities are air-pollution, caused by dust and flooding, caused by high intensity rainfall. Urban green can improve the microclimate by providing soil cover for soil and water conservation. Often no reliable information is available on the extent of the horticultural activities in the city.
Possible problems of (peri)- urban agriculture like health problems related to water quality and the use of city wastes as organic fertiliser make planning and integration of agricultural and waste management even more important. For urban planning up-to-date maps of the urban area are essential (DGUT, 1993). Due to the fast growth of the cities, there is a lack of accurate information. Remote sensing in combination with field checks forms a cost-effective way to map (peri-) urban agriculture. Remote sensing data can be applied in a standardised way using each time the same method and the same land cover legend, it can be used to update maps frequently and it can easily be expanded to other areas. Although the use of satellite information in developing countries seems very promising, case studies in the particular developing countries are scarce but very necessary (Pollé and Boogaard, 1996). Burkina Faso is a Sahel country with a tropical semi-arid climate. Ouagadougou is the capital of Burkina Faso and the population was estimated in 1996 to exist of 1.8 million people but continues to grow very fast. As a result the farmers located in the inner fringe of the city are forced to move out (Club Du Sahel, 2000; Lompo et al., 2000). The Greater Ouagadougou lies in the main cereal production area of the West African Savannah, producing millet and sorghum (Kassam, 1976).
Annual precipitation is not very reliable but volumes are generally between 600 and 900 mm per year. The project 'Recycling Urban Waste in Urban Agriculture Production; Participatory Technology Development In Bamako and Ouagadougou' (APUGEDU) is an EU-funded urban development program. The APUGEDU project requires maps of (peri-) urban agricultural activities to estimate its total surface area and locations to compute compost needs derived from domestic waste (Van den Berg et al., 2001). The first objective of this study is to answer the question of the APUGEDU project: what surface area of Ouagadougou is used for (peri-) urban agriculture? To answer this question a hybrid classification method is developed using land cover data extracted the IKONOS-2 image and other spatial data sets. A second objective of this study is to investigate and evaluate the use of remote sensing, more specifically the use of the IKONOS-2 image in a development project. The accuracy of the resulting map is evaluated using land cover data collected in the field. The technical and human resources are described and possibilities to repeat the study at local institutes are assessed. Specific advantages and disadvantages of IKONOS-2 imagery in the context of the APUGEDU-project are discussed.
2 Using IKONOS-imagery to map (peri-) urban agriculture
In the framework of the APUGEDU project an IKONOS-2 image of 7 June 2000 was purchased to evaluate its value for land use analysis and urban planning. The resolution of IKONOS-2 data is spatially high with pixels of 4m x 4m, but spectrally relatively low. Ikonos has four spectral bands in blue, green, red and near infrared. The first three bands are highly correlated. The IKONOS image is thought to be the most suitable for areas with short-range spatial variation, like a city area. To complement the IKONOS-2 data a SPOT-XS image, acquired in May 1997 is used. The spectral window of SPOT-XS is comparable to that of IKONOS-2 with three bands in the visible green and red and one in the near infrared part of the electromagnetic spectrum. The spatial resolution is much coarser with each pixel representing an area of 20m x 20m. Other spatial data used for this include (1) a tourist map, (2) a digital land register from Ouagadougou, (3) a map giving the sections for waste collection, and (4) an urban horticulture map. As part of the APUGEDU-project, INERA carried out a detailed characterisation study in four selected agricultural sites in Ouagadougou (Lompo et al., 2000).
Collection of field data is carried out one year after the registration date of the IKONOS-2 image. The field data is collected in Ouagadougou, according to a prepared fieldwork plan in co-operation with the remote sensing-centre of INERA. The basis of this study is that the spectral and spatial variation in the image corresponds with the actual physical variations found in the field. At 279 locations in the field thematic information on the land situation is collected using a GPS to determine the geographical co-ordinates. The accuracy of the GPS device is measured at the time of registration and lies between ___ six and nine meter. The thematic data filled in on the registration form are physical characteristics that may have an effect on the reflectance values such as landuse, degradational state, vegetation cover and type, waste cover, soil type, soil colour and soil management.
The extent of the IKONOS-2 image of 11 by 11 km does not cover the total city area and some parts of the IKONOS-2 image are not usable due to cloud contamination. Merging the IKONOS-2 image with the available SPOT-XS image improves the extent and provides data in the areas excluded due to cloud contamination. The merging method applied in this study is the Principal Component Analysis (PCA) and reverse PCA (Lillesand and Kiefer, 2000). The combined-image, which is used for further analysis in this research, is the result of this data integration of IKONOS-2 and SPOT-XS. The combined-image gives the same information as the original IKONOS-2 image with additional information on places where data was missing. The spatial resolution of the combined-image image enables visual recognition of objects larger than four meter in most of the image and enables visual recognition of objects larger than twenty meters in the filled up patches. Roads and river beds can easily be recognised and are digitised manually on-screen for further analysis. A vector file of roads and house was created and will be used later to compute spatial buffers in a GIS environment to assess the extent of certain agricultural activities. Houses appeared to be clearly visible on an IKONOS image when a spectral transformation was applied normally referred to as the Tasseld Cap brightness.
First, a straightforward spectral classification of the IKONOS image was carried out using a conventional maximum likelihood classifier and ground truth collected in the field. Unfortunately, this approach failed to separate the agricultural areas and individual agricultural crops. Therefore, alternative methods to analyse and classify the image were investigated. One approach that proved to be successful comprised to include GIS stored information about 'distance to water', distance to roads (piste)' in the analysis and classification of the image. Four assumptions derived from field observations were included in the alternative classification procedure and are shown in figure 1. Based on these four 'rules it was possible to restrict the potential agricultural area within the city of Ouagadougou based on the following conditions (figure 1):
- Area less than 200 meter from a waterbody (dam or river)
Irrigation is necessary for the intensive horticulture that takes place in the urban area (Hoogmoed, 1999). The local observed method of irrigation is by collecting water directly from the waterbody or from waterholes and manually distributing the water in the field. This is very labour intensive and restricts the workable area to a range of maximum 200 meters. - Area within the IGB landregister-map
The landregister-map derived from the IGB gives the area that is build-on or has a certain purpose. This excludes that area from being potentially used for (peri-) urban agriculture. In the peri-urban area of Ouagadougou agricultural fields require compost as fertilizer and are consequently not randomly located but match the following conditions. Areas are excluded from potential agricultural regions when: - Area less than 500 meter from a piste
Agriculture takes place even more than 500 meter from a piste. That area is not identified in this research as agricultural, since it does not make use of urban waste. The trucks and charts that distribute waste for agricultural use in the peri-urban areas are restricted to the pistes to reach the fields. Observations have confirmed the use of urban on fields waste up to 500 meters from the pistes. - Area less than 500 meter from a habitation (residence).
Around the residence found in the peri-urban area (figure 1), agricultural fields are observed. These fields are well protected by the presence of the farmers and well fertilised by domestic waste. These areas are identified in this study as agricultural.
3. (Peri-) urban agriculture in Ouagadougou
The research and approached followed in this study resulted in various types of land use maps. The APUGEDU project demands a map giving specific information on the location and size of waste application in peri- (urban) agriculture, either unimproved waste or compost. As it was not possible to classify agricultural land use and type of waste used on these fields in a very reliable way on the basis of IKONOS, we decided to produce a probability map with four classes representing the use of waste in urban and peri-urban agricultural activities in Ouagadougou. All four classes are 'agricultural area using waste' resulting from the described spectral classification in combination with the GIS-based classification classified (table 1). The first division is made by separating the area inside the buffer around the rivers and barrages. These areas have access to irrigation during the wet season contrary to the extensive agricultural areas. The second division is based on spectral classification. Areas that have a spectral signature of vegetation, meaning that they are covered by green plants, are classified as 'certain'. Areas with a spectral signature indicating no presence of vegetation are classified as 'potential'. The reason why areas without vegetation features are marked as 'potential' is the nature of agriculture. With one IKONOS image, it is impossible to determine whether a bare plot is a fallow field or permanent bare soil. Overall accuracy of the map lies between 54.7% if the certain and potential classes are merged and 65.5% if the potential classes are considered as non-agricultural. This leads to the four classes presented in table 1.
Figure 2: Map representing the use of waste for agriculture activities (see also table 1)
4. Discussion and conclusions
Conventional spectral classification of IKONOS images proved not to result in reliable maps of agricultural activities and 'use of waste maps' within Ouagadougou despite the high spatial resolution of this sensor. Therefore, we developed a new and advanced hybrid approach to include GIS-stored information in the classification process. This information included distance to irrigation water, distance to roads and distance to houses. This approach resulted in a set of probability maps of agricultural activities and use of waste as shown in figure 2. A disadvantage of the new proposed hybrid classification method is that it does not make optimal use of the information captured by the IKONOS-2 data. In the procedure only reflectance information is used. A promising approach might be to include the morphological properties of agricultural land such as shape, size, and patterns in the classification procedure. Conventional remote sensing methods classify an image on the basis of spectral data without accounting for spatial patterns. In an urban environment, not only spectral patterns are important but also spatial patterns are useful to identify certain objects (De Jong et al., 2000; Van Deursen et al., 1999). To make full use of the information content of an IKONOS image in an urban environment it might be useful to use classification methods based on spatial patterns, such as eCognition (Bauer and Steinnocher, 2001).
Other conclusions that came out of this study are that for a cost-effective use of the IKONOS image there are two important conditions. Firstly, the institute assigned to work with the image should have access to sufficient hardware and software and sufficient human resources for analysis and fieldwork. The large size of the IKONOS image file formed a limitation for processing in Ouagadougou.
Analysis and application of such a large data set requires considerable technological resources as well as human resources. Secondly, the data should be used for as many applications as is permitted by the license and the results should be made accessible to all potential users. To make data accessible a system is required to store and order these data, such a system is called a Spatial Data Infrastructure (SDI). A properly functioning SDI requires long-term investments. Collection of spatial data useable for environmental research is still a very costly and time-consuming process. A hardware/software system, organisational structure, human resources and data sets have to be established.
The role that local partners such as INERA can play in the development of an infrastructure for environmental data for Burkina Faso is large since it is a governmental institute responsible for collection, analysis and distribution of environmental data. Unfortunately, lack of resources is the reason many other institutes have invested in data collection and processing. These data are not regarded as a common resource for the country. This leads to isolated research, duplication of work and unnecessary costs. Data collected and processed with fiscal or foreign aid resources should be used to benefit the country. Data and resources like computers should be shared and be available to the person that needs it most for research. Users and producers of spatial data should form an information community sharing common digital language and data feature definitions and working together on problems (Prévost, et al., 1997). The benefit of using an IKONOS image is now mainly determined by the activities in The Netherlands, since the local institutes lacked human and technical resources to work with the image. Long-term investments are required to train the local staff and provide hardware and software systems to work with satellite images. Data and information can then be used more efficiently and easier be accessed. When these conditions are not assured, the use of IKONOS to classify (peri)-urban agriculture will not lead to significantly better results than aerial photography. In fact, the latter will be a cheaper solution and have benefits over the IKONOS image.
Another problem encountered during this study was the dynamic characteristics of agricultural land use (FAO, 1981). This dynamic character made it very difficult to identify (peri-) urban agriculture using a single, mono-temporal image. (Peri-) urban agriculture is often not characterised by a certain spectral or spatial pattern but by the temporal dynamics in both spectral as spatial pattern. In future, images taken during different moments in the growing season should be combined into one classification.
Acknowledgements
The research project presented in this paper is financially supported by the INCO programme of the European Commission under contract no. Contract ERBIC18-CT98-0288 (APUGEDU).
References
- Aldington, T. , 1997, Urban and peri-urban agriculture: Some thoughts on the issue. Technical Report of the FAO Agriculture Department. FAO Rome.
- Bauer, T. and K. Steinnocher, 2001, Per-parcel land use classification in urban areas applying a rule-based technique. In: GeoBIT/GIS, 6(2001), pp. 24-27, 2001.
- Club du Sahel, 2000, Urbanization, rural-urban linkages and policy implications for rural and agricultural development: case study from West Africa, available on http://www.olis.oecd.org
- De Jong, S.M., A.Bagré, P.B.M. van Teeffelen, W.P.A. van Deursen, 2000, Monitoring Trends in Urban Growth and Surveying City Quarters in the City of Ouagadougou, Burkina Faso Using SPOT-XS. Geocarto International 15, No.2, pp.61-68.1
- DGUT (Direction Générale de l'Urbanisme et Transport), 1993, Strategie D'Amenagement du Grand Ouaga, Horizon 2010, Ministere des Traveaux Publics de l'Habitat et de l'Urbanisme, Burkina Faso.
- FAO, 1981, Agro-climatic suitability classification and yield (with constraints) in T/HA of crops by length of growing period, Report on the Agroecological Zone Project.Vol 3. FAO, Rome.
- Hoogmoed, W., 1999, Tillage for soil and water conservation in the semi-arid tropics, Doctoral Thesis, Wageningen University The Netherlands.
- Kassam, A.H., 1976, Crops of the West African semi-arid tropics. I.C.R.I.S.A.D., FAO, Rome. Leisinger, K. M., Schmitt, K.M. (Eds:), 1995, Survival in the Sahel An ecological and developmental challenge, International Service for National Agricultural Research (ISNAR), The Hague.
- Lillesand, T. M. and R.W. Kiefer, 2000, Remote sensing and image interpretation. Wiley & Sons, New York, 560pp.
- Lompo, F., Bonzi, M., Youl ,S., 2000, Caracterisation de l'agriculture urbaine et peri-urbaine de la ville de Ouagadougou: resultats des diagnostics participatifs dans les sites agricoles urbaines et peri-urbaines du projet APUGEDU, Burkina Faso. Mid-term Report to the EU, Brussels.
- Prévost A. and P. Gilruth, 1997, Environmental Information Systems in Sub-Saharian Africa. Pollé, V.P.L., P. van den Boogaard, 1996, Urban management A feasibility study into embedding of remote sensing for improved urban management in developing countries. BCRS report 96-17, BCRS Delft, 35pp.
- Van den Berg, L.M., I. Kemeling, G.J. Roerink, 2001, Urban Agriculture in Ouagadougou and Bamako: Estimating its spatial distribution with the help of IKONOS high resolution satellite imagery. Wageningen University Report GIRS-2001-35, Wageningen, 49pp.
- Van Deursen, W. A. Bagré, S.M. de Jong, P.B.M. van Teeffelen, 1999, Monitoring Trends in Urban Growth: Ouagadougou, Burkina Faso. BCRS report 3.1/DE-11. BCRS, Delft 50pp.