MODIS: derived greenness maps to better monitor Desert Locust habitats
The project design, its components and partners
In 2007, FAO DLIS partnered with the Université catholique de Louvain (UCL) and the Flemish Institute for Technological Research (VITO), both in Belgium, to develop more efficient remote sensing products. The project was funded by the Belgian Science Policy Office under the Research Programme for Earth Observation - STEREO II. It started in October 2007 and a second phase concludes in summer 2011. The developers met with DLIS and identified the need for a new product that would show changes over time in vegetation conditions in the Desert Locust recession area. UCL designed a new algorithm and validated the prototype. Experts at the European Commission's Joint Research Centre (JRC) refined the prototype. VITO applies the new algorithm to 250m resolution MODIS imagery within an automated process chain to make greenness maps available every ten days to DLIS and national locust centres. The result of these efforts is a dynamic green vegetation map that includes both the spatial and temporal information in one geotif-formatted file suitable for analysis within a GIS. The map shows the evolution of vegetation development for the previous 11 decades (i.e. eleven 10-day periods) for each 250m pixel, indicated in varying shades of red, orange and green. The product informs the user of the spatial-temporal variations of the green vegetation and indirectly of the rainfall distribution through vegetation development. This allows the onset of green vegetation, ephemeral vegetation (false starts), and the disappearance of vegetation at the end of its developmental cycle to be identified. It also indicates the location of evergreen vegetation (values above 11 decades until 36 decades) that is of less importance to Desert Locust.
In 2007, FAO DLIS partnered with the Université catholique de Louvain (UCL) and the Flemish Institute for Technological Research (VITO), both in Belgium, to develop more efficient remote sensing products. The project was funded by the Belgian Science Policy Office under the Research Programme for Earth Observation - STEREO II. It started in October 2007 and a second phase concludes in summer 2011. The developers met with DLIS and identified the need for a new product that would show changes over time in vegetation conditions in the Desert Locust recession area. UCL designed a new algorithm and validated the prototype. Experts at the European Commission's Joint Research Centre (JRC) refined the prototype. VITO applies the new algorithm to 250m resolution MODIS imagery within an automated process chain to make greenness maps available every ten days to DLIS and national locust centres. The result of these efforts is a dynamic green vegetation map that includes both the spatial and temporal information in one geotif-formatted file suitable for analysis within a GIS. The map shows the evolution of vegetation development for the previous 11 decades (i.e. eleven 10-day periods) for each 250m pixel, indicated in varying shades of red, orange and green. The product informs the user of the spatial-temporal variations of the green vegetation and indirectly of the rainfall distribution through vegetation development. This allows the onset of green vegetation, ephemeral vegetation (false starts), and the disappearance of vegetation at the end of its developmental cycle to be identified. It also indicates the location of evergreen vegetation (values above 11 decades until 36 decades) that is of less importance to Desert Locust.
A standard MODIS image (Indo-Pakistan border, area of interest circled in red)
Before the project, national locust information officers in affected countries and FAO DLIS were using 250m MODIS imagery, produced every 16 days, to try to identify green vegetation in desert areas that should be checked for locusts by ground survey teams. Although MODIS represented a significant improvement in resolution compared to earlier remote sensing imagery such as SPOT-VGT (1 km), it remained difficult to identify with precision green vegetation and to understand the conditions of these areas.
Before the project, national locust information officers in affected countries and FAO DLIS were using 250m MODIS imagery, produced every 16 days, to try to identify green vegetation in desert areas that should be checked for locusts by ground survey teams. Although MODIS represented a significant improvement in resolution compared to earlier remote sensing imagery such as SPOT-VGT (1 km), it remained difficult to identify with precision green vegetation and to understand the conditions of these areas.
Greenness map derived from MODIS image (Indo-Pakistan border)
The new greenness map developed under the project is overlaid on top of the standard MODIS image. In the area of interest (red circle), the yellow pixels in the greenness map indicate that vegetation has been green in some places for several decades (i.e. at least one month). This information is not available from a single standard MODIS image. The other pixels that are green in the standard MODIS image are probably false positives, i.e. areas where the image indicates the presence of green vegetation but in reality there is no green vegetation present.
The new greenness map developed under the project is overlaid on top of the standard MODIS image. In the area of interest (red circle), the yellow pixels in the greenness map indicate that vegetation has been green in some places for several decades (i.e. at least one month). This information is not available from a single standard MODIS image. The other pixels that are green in the standard MODIS image are probably false positives, i.e. areas where the image indicates the presence of green vegetation but in reality there is no green vegetation present.
Satellite imagery can be misleading (Tademait Plateau, central Algerian Sahara)
The image on the left suggests that vegetation is becoming green (indicated in red) over a larger area than in the right-side image. The actual situation on the ground is closer to that indicated in the right-side image, which is the new greenness map product developed by UCL. Great efforts were spent during the project in reducing the various errors associated image processing and the sensor.
The image on the left suggests that vegetation is becoming green (indicated in red) over a larger area than in the right-side image. The actual situation on the ground is closer to that indicated in the right-side image, which is the new greenness map product developed by UCL. Great efforts were spent during the project in reducing the various errors associated image processing and the sensor.
Commission errors (Hadramaut region, Yemen)
An illustration of the confusion between bare soils and vegetation, known as Commission Errors, observed with a NDVI-based method (left) compared with the new methodology employed in the greenness maps (right). Since remote sensing imagery is used by locust-affected countries to help guide its survey and control teams and by FAO DLIS to assess and forecast the current situation, commission errors can lead to poor and costly decision-making and inaccurate advice. In this example, Yemeni teams could have been mislead into spending considerable effort to survey a large area of the plateau south and north of Wadi Hadhramaut only to find bare soil and no vegetation, rather than concentrating in the wadi itself where green vegetation is present.
An illustration of the confusion between bare soils and vegetation, known as Commission Errors, observed with a NDVI-based method (left) compared with the new methodology employed in the greenness maps (right). Since remote sensing imagery is used by locust-affected countries to help guide its survey and control teams and by FAO DLIS to assess and forecast the current situation, commission errors can lead to poor and costly decision-making and inaccurate advice. In this example, Yemeni teams could have been mislead into spending considerable effort to survey a large area of the plateau south and north of Wadi Hadhramaut only to find bare soil and no vegetation, rather than concentrating in the wadi itself where green vegetation is present.
Improved detection of green vegetation (southeast Air Mountains, Niger)
Commission errors such as bare soil and other artifacts that are thought to be green vegetation in standard MODIS imagery (top) are reduced significantly in the new greenness map product (bottom). This is most evident in rocky and open areas but also, to a lesser degree, in wadis and drainage areas. Consequently, national ground surveys can be targeted and prioritized better.
Commission errors such as bare soil and other artifacts that are thought to be green vegetation in standard MODIS imagery (top) are reduced significantly in the new greenness map product (bottom). This is most evident in rocky and open areas but also, to a lesser degree, in wadis and drainage areas. Consequently, national ground surveys can be targeted and prioritized better.
Large survey areas can be reduced (Al Hammadah Al Hamra plateau, NW Libya)
One challenge of monitoring Desert Locust that contributes to effective early warning is the large remote areas that need to be checked by national ground teams. As it is not possible to check all of these areas, a prioritization scheme must be used. Many locust-affected countries rely on remote sensing to help reduce the size of the areas that require monitoring. Although the use of 250m resolution MODIS imagery can help, it is believed that regular use of greenness maps can contribute even more to guiding survey teams in the field. In this example, the standard MODIS image (top) suggests that large areas of green vegetation are present on the Al Hammadah Al Hamra plateau in northwest Libya; whereas, the greenness map (bottom) suggests that very little if any green vegetation is present.
One challenge of monitoring Desert Locust that contributes to effective early warning is the large remote areas that need to be checked by national ground teams. As it is not possible to check all of these areas, a prioritization scheme must be used. Many locust-affected countries rely on remote sensing to help reduce the size of the areas that require monitoring. Although the use of 250m resolution MODIS imagery can help, it is believed that regular use of greenness maps can contribute even more to guiding survey teams in the field. In this example, the standard MODIS image (top) suggests that large areas of green vegetation are present on the Al Hammadah Al Hamra plateau in northwest Libya; whereas, the greenness map (bottom) suggests that very little if any green vegetation is present.
More accurate estimate of green vegetation in between sand dunes (Empty Quarter, Saudi Arabia)
Good ecological conditions, including green vegetation, can occur in the relatively narrow depressions and low-lying areas in between sand dunes. As these areas are often remote and extremely difficult to survey, it is important to have a reliable estimate of ecological conditions in such habitats. The new greenness maps (bottom) offer a better estimate of green vegetation than standard MODIS imagery (top). In this example, national locust centres using greenness maps would not survey this sand sea while those centres who rely on standard MODIS imagery might decide differently.
Good ecological conditions, including green vegetation, can occur in the relatively narrow depressions and low-lying areas in between sand dunes. As these areas are often remote and extremely difficult to survey, it is important to have a reliable estimate of ecological conditions in such habitats. The new greenness maps (bottom) offer a better estimate of green vegetation than standard MODIS imagery (top). In this example, national locust centres using greenness maps would not survey this sand sea while those centres who rely on standard MODIS imagery might decide differently.
What is a greenness map?
The greenness map is a dynamic product that provides the location of green vegetation areas at 250m resolution on a 10-day basis. The value of each pixel corresponds to the number of decades since vegetation onset. This allows the user, in this case FAO DLIS and national locust information officers, to follow the spatial-temporal dynamic of green vegetation and identify: (1) ephemeral vegetation occurring after one isolated rainfall event, (2) seasonal (or annual) vegetation, and (3) evergreen (or perennial) vegetation such as palm groves. A colour table is used to inform the user about the vegetation stage, i.e. at the onset (red), close to the onset (orange) or far from the onset (yellow/green to dark green) after 11 decades of vegetation detection. A value of 0 represents No Vegetation; a value of 1 indicates New Vegetation; a value of 12 indicates that vegetation was detected nearly 4 months (11 decades) ago for the first time.
The greenness map is a dynamic product that provides the location of green vegetation areas at 250m resolution on a 10-day basis. The value of each pixel corresponds to the number of decades since vegetation onset. This allows the user, in this case FAO DLIS and national locust information officers, to follow the spatial-temporal dynamic of green vegetation and identify: (1) ephemeral vegetation occurring after one isolated rainfall event, (2) seasonal (or annual) vegetation, and (3) evergreen (or perennial) vegetation such as palm groves. A colour table is used to inform the user about the vegetation stage, i.e. at the onset (red), close to the onset (orange) or far from the onset (yellow/green to dark green) after 11 decades of vegetation detection. A value of 0 represents No Vegetation; a value of 1 indicates New Vegetation; a value of 12 indicates that vegetation was detected nearly 4 months (11 decades) ago for the first time.
Development of green vegetation over time (northeast Mali, September 2008)
The colours indicate the progressive development of vegetation in the field. In this example, green vegetation is present in the wadis of northeast Mali from earlier rain or runoff. For each pixel, its value is compared to the previous image. If green vegetation is detected in the current decade, then the value is increased by one, if not, then it is reset to zero. This helps the user to identify those areas where vegetation is becoming green (red), already green (orange-yellow), and drying out (green). This information is used to guide survey teams and make operational decisions. It is also incorporated into decadal and monthly locust bulletins prepared by national locust centres and FAO DLIS.
The colours indicate the progressive development of vegetation in the field. In this example, green vegetation is present in the wadis of northeast Mali from earlier rain or runoff. For each pixel, its value is compared to the previous image. If green vegetation is detected in the current decade, then the value is increased by one, if not, then it is reset to zero. This helps the user to identify those areas where vegetation is becoming green (red), already green (orange-yellow), and drying out (green). This information is used to guide survey teams and make operational decisions. It is also incorporated into decadal and monthly locust bulletins prepared by national locust centres and FAO DLIS.
Training of new technologies and products is essential
FAO has made significant efforts in training national locust information officers in affected countries in the use and interpretation of the new greenness maps. Users can download an updated map every ten days and incorporate it as a geo-referenced layer in the custom geographic information system, RAMSES, used by national locust centres to manage and analyze locust and environmental data. In this way, greenness maps are an additional tool in the existing arsenal used to identify potential areas where locusts may be present, increasing in number and perhaps require control.
FAO has made significant efforts in training national locust information officers in affected countries in the use and interpretation of the new greenness maps. Users can download an updated map every ten days and incorporate it as a geo-referenced layer in the custom geographic information system, RAMSES, used by national locust centres to manage and analyze locust and environmental data. In this way, greenness maps are an additional tool in the existing arsenal used to identify potential areas where locusts may be present, increasing in number and perhaps require control.
Further refinement of greenness maps is required (central Mauritania)
Although significant gains have been made in using MODIS imagery with the development and introduction of greenness maps, further work remains to be undertaken. FAO DLIS and national locust centres use MODIS at its technological limit for detecting the presence of very sparse, yet important, vegetation in the desert in order to reduce the vast areas that require regular monitoring. It is clear that greenness maps are not fully reliable since not all vegetation can be detected. In this example, ground teams in central Mauritania checked suitable habitats for Desert Locust but were not detected by the greenness maps (green circle indicates where green vegetation was seen by the survey; orange circle indicates drying vegetation). It is hoped that work can continue in improving the satellite sensors and image processing associated with greenness maps so they become more accurate and reliable.
Although significant gains have been made in using MODIS imagery with the development and introduction of greenness maps, further work remains to be undertaken. FAO DLIS and national locust centres use MODIS at its technological limit for detecting the presence of very sparse, yet important, vegetation in the desert in order to reduce the vast areas that require regular monitoring. It is clear that greenness maps are not fully reliable since not all vegetation can be detected. In this example, ground teams in central Mauritania checked suitable habitats for Desert Locust but were not detected by the greenness maps (green circle indicates where green vegetation was seen by the survey; orange circle indicates drying vegetation). It is hoped that work can continue in improving the satellite sensors and image processing associated with greenness maps so they become more accurate and reliable.
See also
eLocust2: saisie et transmission de données de terrain
(en anglais)
Appareil mobile qui permet aux spécialistes antiacridiens de consigner les résultats des opérations de prospection et de lutte sur le terrain et de les transmettre par satellite à leur centre national antiacridien.
eLocust3: saisie et transmission de données de terrain actualisées
(en anglais)
eLocust3 est un appareil mobile que les agents de terrain nationaux peuvent employer pendant les opérations de prospection et de lutte antiacridienne dans les pays touchés par le criquet pèlerin.
Alerte rapide
(en anglais)
Composantes fondamentales requises pour garantir l’efficacité et la
fiabilité des systèmes d’alerte rapide.
Google Earth Engine
(en anglais)
Technologies novatrices et nouveaux outils gratuits de Google: la Division des forêts de la FAO utilise déjà Google Earth Pro, My Maps, Fusion Tables et Google Earth Engine, et leur utilisation est en train d’être étendue à la lutte contre le criquet pèlerin.
Service d’information sur le criquet pèlerin
(en anglais)
Service d’information sur le criquet pèlerin (DLIS) au siège de la FAO à Rome (Italie).