www.guluwalk.com source of the nile
 
U.N. Under-Secretary General of Humanitarian Affairs Jan Egeland has called the situation in northern Uganda “the world’s most neglected humanitarian crisis” and “one of the biggest scandals of our generation.”

 


Background Information:
http://www.un.org/Pubs/chronicle/2004/issue2/0204p7.jpg
 The war in northern Uganda has been ravaging its people for nearly 20 years and has gone largely unnoticed. Over 20,000 children have been abducted by the rebel-led Lord's Resistance Army (LRA) to be used as soldiers and sex slaves, and over 90% of Acholi people have been displaced in camps that offer neither security nor basic provisions [1]. The Leader, Joseph Kony is terrorizing civilians, especially children. To date, the LRA has abducted over 12,000 children who comprise the vast majority of the LRA’s fighting forces, and displaced an estimated 1.3 million people. Internally Displaced People's (IDP) camps are vulnerable to attacks, and have been argued to be a gross violation of human rights. The Ugandan Army protects IDP camps, but has also been accused of violating human rights [1].

On top of the number of IDPs in Northern Uganda, there is also a large in Flux of Refugees from Neighboring countries, namely Rawanda, the Democratic Republic of the Congo, and Sudan. According to the CIA Worldfactbook, there are 184,731 Sudanese refugees,  18,000 Rwandan Refugees [3]. Of the tens of thousands of Sudanese refugees in Uganda, most of them settled in camps scattered around Adjumani district [2].

Vegetation is important because it helps to feed us, cloth us, give us building materials, and medications. When changes occur to the vegetation, health, economy, and environment may all be affected [4]. There is a severely limited lack of overall resources in both Refugee and IDP camps and officials have been forced to resort to rationing [2]. Problems related to the environment include a lack of food, lack of clean water affecting health and overall sanitation, wood for fuel, etc. This is a good reason to examine what is happening to the vegetation in these areas, as it is directly related to resident’s livelihood.

 

Sources:
[1] http://www.guluwalk.com/learn/
[2] http://www.worldpress.org/Africa/1851.cfm
[3] http://www.cia.gov/cia/publications/factbook/geos/ug.html
[4] http://earthobservatory.nasa.gov/Library/MeasuringVegetation/

Back To Top

Project Proposal:

Uganda is located in East Africa on Lake Victory. It borders the Democratic Republic of the Congo, Sudan, Kenya, Tanzania, and Rwanda. It is a beautiful country, with a welcoming culture, and has been referred to as the ‘Pearl of Africa’. Ecologically the region is located where African savannah meets jungle. This makes Uganda one of the most ecologically diverse places in the world. Uganda is home to over 2000 bird species, typical savannah wildlife, and many endangered primates including the Mountain Gorilla. Since 1986, the country has been developing quickly both economically, and in terms of social stability. But, the country is still recovering from rule by dictators like Amin and Obote. The population still suffers from poverty, and there is an unemployment rate of over 80%.  Many people live off of less than a dollar a day. There is still AIDS/HIV, malaria, poor infrastructure, and health care facilities. In the northern part of Uganda, the government continues to fight the Lords Resistance Army (LRA). The conflict is driving people from there homes, into Internally Displaced Peoples (IDP) camps, or onto protected lands where they ‘squat’. These areas, at times, are of great ecological importance. For the GEOG 376 project we would like to find out how displaced people and refugees affect biodiversity in Uganda. This might help to determine which locations should be the focus of efforts of the government, or local NGOs to help people to live sustainably in their new home.

 For this project we will need to get some maps:

  • outline, districts, roads, lakes/rivers of Uganda
  • IDP refugee camp locations (.pdf – WFP, UN)
  • Population density
  •  vegetation maps (probably will be of all of Africa)
    • could measure how refugee camps could affect: water, soil, animal populations, national parks
So to do this:
  • Take the vegetation map (or other) from 2 different years, find areas of vegetation loss.
  • Locate IDP/Refugee camps on the map, and see if it matches up with vegetation loss.
  • It would also be interesting to use a population density map, and do the same thing.
 This is what we hope to accomplish, but we understand data is difficult to obtain, and it will all depend on available information.
 

Back To Top

Purpose
Queen Elizabeth NP

The purpose of this project is to find out if Refugees and Internally Displaced People in Uganda affect the local environment where they live in high concentrations, relative to natural changes that occur due to normal population growth, and environmental fluctuations. We hoped to do this by using Normalized Difference Vegetation Index (NDVI). 


 

Back To Top

Data Collection and Preparation:

"I think you touch with your work the crucial question using GIS in refugee or IDP areas. Generally, refugees are settled in areas where no or little geographical information is available. First, you have to use GIS/RS/GPS to collect the information first. Second, you have to build up a GIS. Lastly you can use the GIS in a decision making process. " - Rainer Zaiss

To find information, we began searching the web and sending off e-mails requesting information on Ugandan human rights and ecological issues, as we knew this would be a challenging task. We sent out many e-mails to institutions and individuals involved with the issues we wished to investigate, and ended up with no pertinent replies except for one. Rainer Zaiss, a researcher who had done a previous GIS study in 2004 was kind enough to provide us with some data that covered refugee settlements in the North of Uganda. There were quite a few vector polygon files which included:

  •     Outline of Uganda [1]
  •     Political District Boundaries [1]
  •     Some selected refugee camp locations (specifically the Adjumani and Rhino Camps) [1]
  •     Refugee Areas, as defined by L'Institute de Recherche pour le Developpment (IRD) [1]

Unfortunately, the file format was not compatible with ArcGIS and had to be converted. So, we used the Feature Manipulation Engine (FME) to convert them from files that were compatible with another GIS to ArcGIS shapefiles.


Dave's Cousin Ben, who currently works for MacDonald Dettwiler in Burnaby was able to give us some great advice. He referred us to the   Global Land Cover Facility - Earth Science Data Interface (ESDI) website, where free satellite imagery can be downloaded. They have a search engine where you can find imagery from many different satellites to suit a variety of applications. After difficulties searching for data in Africa that had a fairly high resolution (1km. and up), we chose images from the The Moderate Resolution Imaging Spectroradiometer (MODIS) satellite with a 500 meter resolution. The 32-day composites available from the GLCF were derived from the MODIS level 3 surface reflectance product [2], and each of the 7 bands could be downloaded separetly.

The table below describes wavelengths represented by the 7 bands included in the MODIS imagery:
Band
 Wavelength (nm)
 Description
1

620-670

Red
2

841-876

Near-infrared
3

459-479

Blue
4

545-565

Green
5

1230-1250

Short wave infrared
6

1628-1652

Short wave infrared:(similar to Landsat band 5)
7

2105-2155

Short wave infrared:(similar to Landsat band 7)

To create an NDVI, we only needed to use bands 1 and 2; the red and near-infrared bands respectively. We dowloaded these bands from the following images:
  • 2001-08-13 to 2001-09-13,GLCF, Goode's, Africa, Online: 025-743 [2]
  • 2002-08-13 to 2002-09-13,GLCF, Goode's, Africa, Online: 026-863 [2]
  • 2003-08-13 to 2003-09-13 GLCF, Goode's, Africa,Online: 031-425 [2]
  • 2004-08-12 to 2004-09-12,GLCF, Goode's, Africa, Online: 071-990 [2]
  • 2005-08-13 to 2005-09-13,GLCF, Goode's, Africa, Online: 074-037 [2]

Each data file contained raster information for the entire African continent, which made its file size an imposing 579696015 bytes! This made data management very complicated in terms of storage purposes. Clearly, the images needed to be cropped to a more manageable size before proceeding. We cropped the raster files by zooming to the approximate region we were interested in, which was the country of Uganda. Using the 'Spatial Analyst' toolbar we selected options, the extent tab, and set the extent to the 'same as display' for the first band.  We then used the raster calculator, and multiplied the raster image by 1. To crop the remaining nine images we could simply multiply each one by the number 1 and obtain a manageable file for our specific storage and presentation needs.

The Projection Problem:
We ran into a problem when we tried to reference the satellite images with the polygons we had obtained through ArcMap. The projection systems were not the same. Furthermore, the projection of the MODIS rater imagery was 'Goode's Homolosine projection', and this specific projection coordinate system is not yet supported by ArcGIS.  So we had to look on the internet once more, for information on how to use the data effectively. The vector layers were in UTM coordinate system 84. We tried to project the raster maps to the same UTM coordinates in zone 36N (the Northern half of Uganda), but this didn't work. We were forced to learn more about the Goodes Homolosine projection, with the help of Jose!

We learned that in Goodes projection system there are 12 zones, and of course Uganda (being located on the equator) falls into 2 of them: Zones 4 in the northernmost part, and zone 7 in the south [4]. We found documents suggesting that we use the 'sinusoidal' projection that does exist in ArcGIS, but changed some of the numbers around accordingly.

The Following are the Goode's component projection files for Zones 4 and 7:

Zone 4:

input
projection GEOGRAPHIC
units DD
parameters
output
projection SINUSOIDAL
units METERS
parameters
6370997.0
30 00 00
3335846.22854
0.0
end

Zone 7:

input
projection GEOGRAPHIC
units DD
parameters
output
projection SINUSOIDAL
units METERS
parameters
6370997.0
20 00 00
2223897.48569
 0.0
end

We used the numbers that defined zone 4 only (30 00 00 was the  meridian, and  3335846.22854 was the false easting),  and it appeared to work.

We also found a GIF map that showed the location of the national parks in Uganda on a Safari Company's website.  In order to use the information from the map, we first had to georeference the image. Using the georeferencing toolbar in ArcMap, we georefernced this image according to the Outline map of Uganda which was in sinusodial projection. We chose distinct points along Uganda's boundary as control points to do this. There were no shared interior points to reference from. What we should have done is scanned a map from an atlas for better accuracy, but the GIF we chose was sufficient for our mandate.

Once the image was georeferenced we had to create a new polygon file that had only the outlines of each of the national parks on it. First we used ArcCatalog to create a new shapefile that would serve as the template for our national parks polygon image. We then imported the blank image into ArcMap and began to create the national parks. To do this, we roughly digitized the location of the national parks using the pencil icon in the editor toolbar.

We continued to e-mail many others, and scoured the web for more information. We found that Relief Web provided information about the numbers of refugees in certain locations, although we had been advised that most of their .pdf maps may contain many errors, and are very general because the data are aggregated. We understand that the information they provide us with are not useful for decision making process on local level. We downloaded the following .pdf files with the possibility of them coming in handy later in the project:

  • Uganda: Affected populations by district, refugees and internally displaced (2000), UN-OCHA [3]
  • Uganda: Affected populations by district, refugees and internally displaced (2001), UN-OCHA [3]
  • Uganda: Affected populations by district, refugees and internally displaced (2002), UN-OCHA [3]
  • Uganda: Affected populations by district, refugees and internally displaced (2003), UN-OCHA [3]
  • Uganda: Affected populations by district, refugees and internally displaced (2004), UN-OCHA [3]
  • Uganda: Affected populations by district, refugees and internally displaced (2005), UN-OCHA [3]

Before getting started on the analysis, we had to make sure that our raster files only contained information about Uganda, rather than all of Africa, or the horn of Africa. To do this we took our vector polygon image of Uganda and converted it to a raster image using the 'feature to raster' tool under conversion tools in the ArcToolbox.  We used the outline map of Uganda to create a Raster Mask with a cell value of 1 and pixel size of 500 m. This was then used to crop the MODIS images again. This time we used the Raster calculator. The equation we inputted was the MODIS image multiplied by the Uganda Mask we had just created. We did this for all 10 MODIS images. Once we had all of our Vector and Raster data collected, cropped, and projected to the same coordinate system, we were able to begin preparing the data for analysis!



Back To Top

Analysis
Creating NDVI's:

        To determine the density of green on a patch of land, researchers must observe the distinct colors (wavelengths) of visible and near-infrared sunlight reflected by the plants. When sunlight strikes objects, certain wavelengths of this spectrum are absorbed and other wavelengths are reflected. The pigment in plant leaves, chlorophyll, strongly absorbs visible light for use in photosynthesis. The cell structure of the leaves, on the other hand, strongly reflects near-infrared light. The more leaves a plant has, the more these wavelengths of light are affected, respectively [1]. An algorithm called a "Vegetation Index" is used to quantify the concentrations of green leaf vegetation [1].


The equation to input is: NDVI = (Near-Infrared - Red)/ (Near-Infrared + Red)
for us this meant: NDVI = (band 2 - band 1)/ (band 2 + band1)

To make an NDVI we had to use the Raster calculator from the Spatial Analyst Toolbar. Unfortunately this was not providing us with the intended results. The Equation had to be done in a 3 step proces with some fine adjustments:
Three Raster Calculations:
  1. We created the numerator: numerator =  float ([band 2] - [band 1])
  2. Create the denominator: denominator = float ([band 2] + [band1])
  3. Create the NDVI: numerator / denominator

An example of the results (for the horn of Africa rather than Uganda) is on the left. We did this for each of the years between 2001 and 2005 resulting in 5 NDVI Images.

An NDVI is a ratio between the measure of reflectivity in the red and near Infra-red portions of the electromagnetic radiation spectrum. Because it shows the absorption of chlorophyll and the density of green vegetation, the contrast between vegetation and soil is at a maximum. The index value is sensitive to the presence of vegetation on earth's land surface. Very low values of NDVI correspond to barren areas of rock, sand, or snow. Moderate values represent shrub and grassland, while high values indicate temperate and tropical rainforests [1]. As you can see, Uganda has a relatively high vegetation index.

We wanted to track the change in vegetation over the years so we simply subtracted the images from each other using the raster calculator. The following Imanges show the change in vegetation index from year to year.

   

Base Information to try and Isolate Human Impact:

Because vegetation changes naturally over the years, we needed to find a way to adjust the data in order to isolate the impact caused by humans. A major factor in the variation in vegetation is related to the amount of rainfall in a given year. Unfortunately we were not able to find detailed yearly rainfall data for Uganda that could be used in our study. We then decided to use National Parks as our control, all the while making the assumption that protected areas within Uganda are not impacted by anthropogenic sources.


We did this by first retrieving the vegetation change data for each of the national park polygons. We were able to do this by using ' zonal statistics' in the spatial analyst toolbar menu. The output table of this function showed among other values, the mean value of vegetation loss or gain for each of the parks. Since Murchison Falls National Park was located closest to our area of interest we used the mean values from this park for each of the years. The following images show the changes in Vegetation in Murchison Falls National Park:





Year
 Mean Vegetation Change in Murchison Falls
2001-2002
 -0.014185
2002-2003
 0.008979
2003-2004
 -0.007473
2004-2005
 -0.011532

 

Two Different Spatial Scales

To isolate the vegetation change in the refugee camp locations, we were able to use the 'extract by mask' function under spatial analyst tools in the ArcToolbox. The images then showed the vegetation change in the camps over the year range. These maps showed all of the vegetation change, both natural and anthropocentric. So we took each of the images and subtracted the Mean Vegetation change in Murchison Falls National Park using the Raster Calculator for each of the years. We followed the same steps to isolate the vegetation change in the IRD refugee camp areas.


Incomplete Analysis using Population:

We also attempted to find a correlation between the population density of refugees and IDPs with the amount of vegetation change. We could do this by using the District map of Uganda, and the figures from the Relief web maps we had obtained.

To do this we had to add the numbers from the map to the Attribute Table of the District Map. We did this by Opening the attribute table, selecting 'options', 'add field' and entered the name of the field. We entered fields for the population in each year. We started an editing session and entered the numbers according to the corresponding districts on the map.

Since we wanted refugee and IDP population density, we first had to calculate the area of each district. We did this by adding a field for the area, right clicking on the new field, and selecting 'calculate'. We used the advanced options and the VBL code for area from the help menu to calculate the area of each district.

Once we had the area, we were able to add fields for the population density for each year. We right clicked on each field and selected 'calculate'. This time we chose to take the field with the population from the corresponding year and divide it by the area we had calculated.   

We were then able to use the zonal statistic function again, under the spatial analyst toolbar's menu and isolate the data for each affected population density, relative to the mean vegetation change in the corresponding district.

Sources:
[1] http://earthobservatory.nasa.gov/Library/MeasuringVegetation/

Back To Top

Results:

With the 2 different spatial scales and the 4 diffrent vegetation change maps for each, getting results was amoung one of the toughest parts of the project.  Nonetheless,  the following Images Show the results of our analysis:

The Refugee Camp Locations:

It appears as though the resulting vegetation change over the years in the Refugee camp locations is random.

The IRD Refugee Camp Areas:

In these images the resulting vegetation change also appears to be random. Taking a closer look you can see a region west of the refugee camp locations in red. In 2002, there was a major increase in vegetation in the area, but in 2003 there had been an overall decrease. In 2004 and 2005 the decrease in vegetation appears to be intensifying.

Back To Top

Conclusion:
We attempted to find a correlation between vegetation change, and the location and numbers of Refugees and IDPs in Uganda, and found that there was probably none.  Although our results at a larger scale show the possibility of a correlation between Refugee camp locations and vegetation loss, we are not ready to conclude that there is a correlation for the following reasons.

Although this may be the case, it is overwhelming clear that there are serious gaps and flaws in the available data.
  1. lack of rainfall and data
  2. lack of local population data
  3. lack of data and accuracy/reliability about IDPs and Refugee locations and numbers
  4. lack of free high resolution remote sensed data over an extended period of time
It is very difficult, if not impossible, to obtain reliable detailed information on the physical environmental conditions, and the socio-demographic characteristics of the population and its localisation, the economic activity of the region, its infrastructures,  and so on from those implementing refugee programs [1].

"The almost complete absence of information on this type of question – which notwithstanding represents the environmental and social context in which the refugees have to live – does not appear particularly to worry the humanitarian organisations the most directly involved in the concrete immediate action or those involved in studies and analysis [1]."

Furthermore, spatial resolution can impact the way data is interpreted. As with all raster files, one runs into the mixed pixel problem. The resolution of 500 square meters is fairly coarse to infer any subtle changes in the refugee camp locations that may take several years to propagate into a problem. Making the assumption that theoretically, normal vegetation cover in Uganda will show up as 100% green. In order for any changes in green cover to show up, there may have to be at least a 50% decrease in any given pixel before the satellite sensors will detect change. This means that vegetation depletion may already be under way, we are just not aware of it yet.

Geographical units also affect the way results are interpreted. Relationships tend to grow stronger when looking at larger geographical units, and it appears as though this is the case with our results. On the small scale of the refugee camp locations, there was no correlation between the locations of the camps and vegetation change. When looking at the refugee camp areas, you can see a slight possibility of a correlation. Further analysis that attempt to correct the problems with resolution, and the scale of geographical units is needed to gain a clearer picture of what is happening.


 We would like to extend a special thanks to those who played a major role in the completion (or close to completion) of this project.

 Rainer Ziass, Ben Foster, Jose Aparicio, Kaoru Tachiiri, Jeff Phillips, Brian Klinkenberg, and Andy Knox: We cannot thank you enought  for their advice, support, and help over the term.


[1] http://www.bondy.ird.fr/carto/refugies/rapuga/1_intro.pdf

Back To Top

Related Links :

 

Other Notes:

Possible sources for GIS information about Uganda: