Identifying Zambian Regions of Staple Food Net Consumption and Supply
Geo 425: GIS Project
April 17, 2009
Kim Borland, Almaz Naizghi and Steve Longabaugh

Information About the Problem

This project will identify regions in Zambia that are net producers of staple food and those that are net consumers of the same staples. By mapping the areas of supply and demand it is hoped to give insight into the flows of staple food (maize and cassava) from production to consumption areas.

We will look at how the pattern of supply and consumption varies between 2 survey years: 2001 and 2004. This type of analysis would be most useful if it took into consideration the supply and consumption of neighboring countries, as there are significant trade flows between neighbors, but this project will be limited in scope to only the country of Zambia.

Read the original project proposal.

Background top

Zambia is a country in southern Africa. Here are some sources of information on Zambia:

 

Zamiba in Africa
image courtesy of: http://www.state.gov/cms_images/zambia_map.jpg

 

The two main staple foods in Zambia are maize and cassava. Maize and corn are the same crop so should be well known to most readers of this site. Cassava, on the other hand is a crop that might not be known to many even though it is an important staple crop in much of the world. To find out more about cassava take a look at these sites:

Cassaava
image courtesy of: http://home.nau.edu/envsci/env499/farming.asp

Discussion of Methods, Data Sources top

Net staple purchases and sales estimates. This data was made available to us by the Food Security Group in the Department of Agriculture, Food and Resource Economics at Michigan State University. This project, in conjunction with in-country partners, conducted household surveys in rural and urban areas of Zamiba. From this data, we obtain mean per capita net purchase and net sales of maize and cassava in the survey years of 2001 and 2004 in kilograms. For more information about this project and it's research in Zambia and other countries and regions in Sub Saharn Africa, see their project web site at: http://www.aec.msu.edu/fs2/index.htm

Population Density. In order to generate the total amount of net sales and net purchases for an urban or rural area, we use population density models. Fortunately, there are 2 quality sources of this information. As part of this project, we would like to briefly compare these two models. The first population model is available from the Global Rural-Urban Mapping Project (GRUMP) produced by the Center for International Earth Science Information Network (CIESIN) of the Earth Institute at Columbia University. For more information about this project, view their web site: http://sedac.ciesin.columbia.edu/gpw/index.jsp. ( Figure 1 shows how the GRUMP model allocates population in Zambia.) The second population density model that we will be evaluating is available from the LandScanTM Dataset which has been developed as part of the Oak Ridge National Laboratory (ORNL) Global Population Project. For more information about this dataset, view their web site at: http://www.ornl.gov/sci/landscan/. ( Figure 2 shows how the LandScan model allocates population in Zambia.)

Raster and Vector Files. Since the data that we have are for urban and rural areas, we need to have spatial files that are capable of matching this level of data detail. To help us in this reguards we used an Urban Extents grid (Figure 3) available from the Center for International Earth Science Information Network (CIESIN). Similialry, a vector file of Zambian districts (Figure 4) was made available by the Food Security Group. Through a series of processing steps, we generated a vector file of the entire country of Zambia, defining thosed areas that are urban and those that are rural (Figure 5).

Software. The spatial component of this project was carried out in ArcGIS: ArcMap, ArcScene and ArcCatalog. The net staple purchases and sales estimates were produced with STATA. STAT TRANSFER was used to convert the STATA data files to dbf files that could be joined to the vector files.

Flow Chart (see full size image) top

Flow Chart

Results top

Models of Zambian Population Density, people per square km

Figure 1: GRUMP Model (see full size image)

GRUMP Population Density

Figure 2: LandScan Model (see full size image)

LandScan Populaiton Density

Comments on Figure 1. The relative uniformity of population density across large swaths of Zambia shows that this model of population density aggregates over very large areas. No attempt is made to allocate rural population to areas smaller than districts.

Comments on Figure 2. The Landscan model allocates rural population to areas smaller than districts. The spider web of coloration follows the road network. One assumption used by the Landscan model to allocate population is that people generally live along roads.

Zambia Extents

Figure 3: Rural and Urban Extents (see full size image)

Urban Extents

Figure 4. Districts (see full size image)

Zambia Districts

Figure 5: Rural and Urban Extents Within Districts (see full size image)

Rural and Urban Extents

Comment on Figure 3. This shows the areas of Zambia that were defined as urban by CIESIN. These were used to generate a vector file of Zambian urban areas.

Comment on Figure 4. This shows Zambia divided into its administrative districts

Comments on Figure 5. This shows the result of overlaying the urban extents over the districts. This produces a vector file that we can use to map the net sales and net purchase of staples of urban and rural extents.

Table 1: Maize and Cassava Net Position by Rural and Urban Areas, Per Capita, Kilogram (view full table)

Variable Descriptions

  • Gridcode: desigination of a polygon as a rural or urban area
    • 0: Rural extent
    • 2: Urban extent
  • Maize04: Maize mean per capita net sales in 2004, kg
  • Maize01: Maize mean per capita net sales in 2001, kg
  • Cassava 04: Cassava mean per capita net sales in 2004, kg
  • Cassava 01: Cassava mean per capita net sales in 2001, kg
  • Mboth: the maize net position status over both survey periods
    • 1: always a net buyer
    • 2: always a net seller
    • 3: changed
  • Cboth: the cassava net position status over both survey periods
    • 1: always a net buyer
    • 2: always a net seller
    • 3: changed

Note: A negative number (for variables: Maize04, Maize01, cassava04, cassava01) indicate net purchases, a positive means net sales.

GridCode DISTRICT PROV Maize04 Maize01 cassava04 cassava01 Mboth Cboth
0 Kaputa 6 25.7151 -13.5795 12.8699 39.0741 3 2
0 Mpulungu 6 48.3134 6.34935 19.8604 28.9091 2 2
0 Chiengi 4 16.9275 -3.03128 28.2724 14.9325 3 2
0 Mbala 6 35.7253 35.8931 13.8283 12.3456 2 2
0 Nchelenge 4 -2.0616 0.910452 27.1508 18.296 3 2
0 Nakonde 6 67.1055 12.2906 72.9976 30.9177 2 2
0 Kawambwa 4 23.2705 7.40712 12.7267 4.73871 2 2
0 Mporokoso 6 34.2285 -13.7321 0.877585 12.5878 3 2
0 Mungwi 6 38.2518 15.9243 22.839 64.4029 2 2
2 Mufulira 2 -116.1 -116.1 -2.8 -2.8 1 1
2 Chingola 2 -116.1 -116.1 -2.8 -2.8 1 1
2 Kalulushi 2 -116.1 -116.1 -2.8 -2.8 1 1
2 Lufwanyama 2 -116.1 -116.1 -2.8 -2.8 1 1
2 Kalulushi 2 -116.1 -116.1 -2.8 -2.8 1 1
2 Kafue 5 -116.1 -116.1 -2.8 -2.8 1 1
2 Kafue 5 -116.1 -116.1 -2.8 -2.8 1 1
2 Mazabuka 8 -116.1 -116.1 -2.8 -2.8 1 1
2 Mazabuka 8 -116.1 -116.1 -2.8 -2.8 1 1

 

Table 2. Summary of Rural and Urban Net Position Status Over Both Surveys

 
Maize
Cassava
Status over both surveys Number % Number %
Always net buyers 58 50% 47 40%
Always net sellers 37 32% 44 38%
Changed 21 18% 25 22%
Polygons in the country 116 100% 116 100%

Comment on Table 2. The majority of the rural and urban extents did not change their status over the two surveys. 18% of the areas changed their maize status and 22% changed their cassava status.

Rural and Urban Extents Net Position: Maize

Figure 6: 2001 (see full size image)

Maize 2001

Figure 7: 2004 (see full size image)

Maize 2004

Figure 8: 2001 and 2004 (see full size image)

Maize 2001 and 2004

Rural and Urban Extents Net Position: Cassava

Figure 9: 2001 (see full size image)

Cassava 2001

Figure 10: 2004 (see full size image)

Cassava 2004

Figure 11: 2001 and 2004 (see full size image)

Cassava 2001 and 2004

 

Total Positition in 2004, kg per square km

Figure 12: Maize 2004 (see full size image)

Total Maize 2004

Figure 13: Maize 2001 (see full size image)

Total Maize 2001

Figure 14: Cassava 2004 (see full size image)

Total Cassava 2004

Figure 15: Cassava 2001(see full size image)

Total Cassava 2001

Comments on Figures 12-15.

    • The values in these grids were calculated by multiplying the per capita mean net sales position for each crop and year combination (Figures 6, 7, 9 and 10) by the Landscan population density grid (Figure 2).
    • Urban polygons are outlined on these maps to point out how all of the urban areas were black or grey, areas of net purchasing as one would expect, and also to portray the spatial location of the net selling areas (dark green areas).  High selling areas generally surround urban areas that need to purchase because they can't produce their own food, or they are near the borders with other countries suggesting they may be exporting their crop (perhaps to their urban centers).

     

Table 3. Zambia National Net Sales Position by crop and year, kg
Net Purchase Position in Zambia by Crop and Survey year, kg
Maize 2004
Maize 2001
Cassava 2004
Cassava 2001
-207,748,166
-216,805,838
68,130,733
97,285,285

Comment on Table 3: The figures in this table were derived by summing the net positions of each crop and year combination over the entire country. In both survey years more maize was purchased than sold (indicating the likelihood of imports from neighboring countries) and more cassava was sold then purchased (indicating exports of cassava). Regional staple food trading is a significant component to Zambian food security.

 

Discussion of Results top

Findings

Population density model evaluation. The population mapping of Zambia by the GRUMP (Figure 1) and Landscan (Figure 2) models show the differences in how they allocate population spatially within a country. Overall, we find that the Landscan population density model approximates more closely what we would expect to see, variations of population over rural areas as well as population diffences between urban and rural areas.

Net Sales Position. The status of the net sales position across the country varied less than we might have imagined. Overall, just 18% of the rural and urban extents changed maize net postion status and 22% changed cassava net position status (Table 2).

National Net Sales Position. There was also uniformity over the survey years of the net maize and cassava positions. In both years, more maize was purchased than sold while more cassava was sold than purchased (Table 3). This indicates imports of maize and exports of cassava in both years.

What this tells us about emergency response. The survey years 2001 and 2004 were NOT considered drought years. Given that maize needed to be imported even in these non drought years, imports would certainly be needed to meet staple needs in drought and other emergency years. But the needed import might be less than expected. This has to do with the coping strategies that rural households adopt in drought years. In such years, more cassava tends to be harvested due to it's drought tolerance and ability to remain in the ground and continue to grow over multiple crop years. In effect, it is a food security bank that can be accessed in difficult times (see Cassava – underground food reserve cushions against catastrophic food loss). It has been shown that Zambia sells more cassava then it purchases even in non drought years. One would expect cassava sales to increase in drought years especially as maize becomes more expensive and consumers substitute cassava for maize in their diets.

Problems

Display of results. It was out intention to produce 3d maps to portray the spatial distribution of the net position across Zambia. Conceptually, it was simple to think of multiplying the per capita net position by the population density (per square km). We were able to carry out those calculations (multiplying a raster by a raster to obtain a raster), but displaying results in a 3d map was problematic. With some assistance from Dr. Shortridge we understood the problem to be due to the courseness of resolution in the mapping program. Given more time, we might have been able to figure out how to mitigate this problem. Here are samples of the initial 3d mapping efforts:

problem 3d map #1 Problem 3d map #2 Click on these image to see a full size map of this initial approach.

We tried a 2nd approach by multiplying the population of each rural and urban extent by the per capita net position. This was done within a shapefile, multiplying variables to obtain a new variable. We were not satisified with that result either.

maize national net position in 2004Click on this image to see a full size map of this 2nd approach.

In the end, we decided the best visualization was the set of maps Figures 12, 13, 14 and 15. Even though they are not 3d, they do the best job of displaying the spatial distribution of the net sales position by urban and rural extents.

Areas for future research

Good and bad year responses. It would be instructive to evaluate how net sales position patterns are impacted during good and bad growing seasons.

Reasons behind changes in net position status. While only 18% of areas changed maize status and 22% changed cassava status, understanding the dynamics behind these changes would be very useful. Perhaps the changes were due to local conditions that negatively impacted production or perhaps they were due to a different type of change that impacted people's abililty to access markets or to earn income for purchases.

Regional trade. Zambia is a landlocked country and has significant staple food trade with its neighbors as well as the global market. To really understand the complexity of the Zambian food security situation, an analysis of these staple food trade flows would need to be undertaken.