1995 Pilot Study: Cropland Use Intensity Interpreted from Satellite Imagery of the Mutarara/Sena Area, Mozambique

Tom Crawford, Hughes STX Corporation, EROS Data Center, U.S. Geological Survey
Ron Lietzow, Hughes STX Corporation, EROS Data Center, U.S. Geological Survey
Jim Verdin, EROS Data Center, U.S. Geological Survey

[This article is reprinted from the HSTX Center for Global Change newsletter, Vol. 3, December 1996, courtesy of the authors and Hughes STX Corporation. This file was created by scanning and optical character recognition; some minor edits were made for HTML compatibility.]

Background

In 1975, after 11 years of war, Mozambique became independent of Portugal under a single-party, Marxist-Leninist government. The new government began to collectivize large farms that had been owned mainly by Portuguese farmers. Following independence, however, economic conditions deteriorated, and in 1979 an armed, anti-communist rebellion began. Guerilla warfare engulfed Mozambique and continued until a general peace accord was signed by the government (Frente para a Libertacdo de Mocambique, FRELIMO) and the Resistenca Nacional de Mocambique (RENAMO) in October 1992. Following the signing of the Rome Peace Accord, and anticipating an end to a devastating regionwide drought, the first wave of Mozambican refugees returned from Malawi between October 1992 and January 1993. Some of the returnees planted, cultivated, and harvested crops during this first postwar agricultural season. In 1993, the U.S. Agency for International Development (USAID) began a program to rehabilitate transportation infrastructure through land mine removal and road and bridge repair. In addition, USAID continued a program to provide necessary inputs to agriculture, such as seeds and tools, as well as food and other urgent assistance needed by returnees and others affected by war and drought. Between 1992 and 1995, peace was restored, more than 1.8 million refugees and nearly 4 million internally displaced persons returned to their homes and farms, peaceful multiparty elections were held, and economic activity-- particularly agriculture related-- increased.

In 1995, USAID contacted the U.S. Geological Survey (USGS) to do a pilot study to evaluate the usefulness of interpreting satellite imagery to document the return of Mozambicans to the countryside around the towns of Mutarara and Sena near the Zambezi River. Cropland Use Intensity (CUI) can be interpreted from satellite imagery as an estimate of the percentage of land area used for cropping (Table 1). It has been used successfully in Africa for a number of applications. Estimates of CUI have been employed as a surrogate for rural population density by USAID's Famine Early Warning System (FEWS) project. In Mozambique, the FAO in 1985 prepared a Land Use Inventory of Mozambique, including CUI categories, that showed high correlations between CUI and population density. The USGS has used CUI in the Sahel to support interpretation of Normalized Difference Vegetation Index (NDVI) images derived from Advanced Very High Resolution Radiometry (AVHRR) satellite imagery and to spatially disaggregate agricultural statistics.


Table 1: Characteristics of Cropland Use Intensity (CUI)
-----------------------------------------------------------------------------
How is CUI expressed?           Class  Interval  Middle value
                                  1     70-100%      80% 
                                  2     50- 70%      60% 
                                  3     30- 50%      40%
                                  4      5- 30%      20% 
                                  5      0-  5%       1% 

Estimated Area Cultivated:      Product of polygon area and middle value. 

Ideal Date of Satellite Image:  Early dry season, after harvest and before burning. 

Key Indicator:                  Contrast between cut fields and surrounding vegetation. 

Modifiers:                      Alluvial, upland, and surface water. 

Scales:                         1:100,000 (1 cm = 1000 m, or 1 inch = 1.6 miles) 
                                1: 50,000 (1 cm =  500 m, or 1 inch = 0.8 miles)


Selection of Satellite Scenes

The area of the pilot study includes the towns of Mutarara and Sena in central Mozambique and portions of the Zambezi and Chire Rivers. It is roughly centered on the town of Sena (across the Zambezi River from the town of Mutarara) and encompasses agricultural areas of interest to the east of Sena, between the Zambezi and Chire Rivers and along the south side of the Zambezi River along the road to Caia.

Figure 1

A sequence of several steps was followed in carrying out the pilot study. First, cloud-free images were selected from archived satellite images. The ideal time of year for each of the images is early in the dry season, after harvest and before burning. In Mozambique, this period falls in April and May, when there is good contrast between the cut fields and still-standing adjacent, noncrop vegetation. Five time periods were targeted, corresponding to contrasting biophysical and socioeconomic conditions in the study area (Table 2).


Table 2: Target time periods
-----------------------------------------------------------------------------------------
April/May 1975   Cultivation patterns before independence on July 25, 1975. 

April/May 1985   Following departure of colonial companies, creation of state-owned 
                 farms, "villagization" program, and other postindependence agriculture 
                 policies; but before widespread war-related destruction and depopulation, 
                 which began in 1986. 

April/May 1992   Near-maximal population displacements due to civil war and drought, 
                 preceding the October 1992 Peace Accord. 

April/May 1994   Postwar land use patterns affected by cessation of hostilities, influx 
                 of refugees, greater rainfall, and removal of land mines from roads. 
                 Second harvest since Peace Accord and USAID rehabilitation activities. 

April/May 1995   Most recent harvest following end of major return migrations of refugees 
                 and displaced persons. USAID-sponsored land mine removal and 
                 rehabilitation support continue. 


Catalogs of archived, satellite imagery acquired by Multispectral Scanner (MSS) and Thematic Mapper (TM) sensors on Landsat satellites, and by Multispectral (SX) sensors on the Systeme Probatoire d'Observation de la Terre (SPOT) satellites, were searched for cloud-free scenes. Images that were cloud free or nearly cloud free were located for 1973, 1985, 1992, 1994, and 1995 for the area of interest (see Figure 1) for the dates shown in Table 3. It was not possible to obtain a cloud-free scene for the April/May period for 1973 and 1992.

Table 3: Available Satellite Imagery

Date          Sensor  Res.   Extent, km   Conditions 
--------------------------------------------------------------
15 Aug 1973   MSS     80 m   180 x 180    Preindependence 
11 May 1985   MSS     80 m   180 x 180    Postindependence 
24 Feb 1992   TM      30 m   180 X 180    Peak Displacement 
28 May 1994   TM      30 m   180 X 180    Returnees 
20 Apr 1995   Xs      20 m    60 x  60    Most Recent 

MSS = Landsat Multispectral Scanner
TM  = Landsat Thematic Mapper
XS  = SPOT Multispectral 


The digital data for these five scenes were geocorrected and then printed at scales of 1:100,000 for MSS and TM, and 1:50,000 for SPOT on 40" x 40" (102 cm x 102 cm) paper. False color (red) was used to indicate response in near infrared wavelengths, which is especially useful for discriminating different vegetative cover types.


Cropland Use Intensity (CUI) Estimation

An experienced satellite image interpreter delineated zones of CUI with ink on mylar overlays for the satellite image prints. Cues including color, texture, and context were used to define the zones of relatively homogeneous level of cultivation. The landscape was also divided into major regions-- alluvial, upland, and surface water. The mylar manuscripts were then digitized to create the information "layers" used in a geographic information system.


Geographic Information System (GIS) Analysis

The analyst and GIS Data Management Specialist created data layers to indicate the location of towns, roads, and hydrographic features in the area of interest. An information layer for the roadway network was used to provide context to the CUI findings, and information of December 1994 was used to indicate road rehabilitation activities that have taken place. Techniques of GIS overlay analysis were employed to quantify the differences in CUI over the five dates chosen for the pilot study.


Figure 2

Results: Study Area CUI -- Four Maps

The study area includes 1,078,031 hectares, or 10,780 sq km (Figure 2). The Mozambique Cropland Use Intensity map of August 15, 1973, indicates 30-50 percent CUI in most of the area in an approximately triangular zone defined by Sena, Tembe-Tembe, and Caia, and in an area south of the Zambezi River between Sena and Caia. With few exceptions on that date, CUI is 30 percent or less in the remaining portions of the study area. Overall, about 16 percent of the study area was cultivated on August 15, 1973, a time during which there were still large colonial agricultural activities operational but during which there may have been some guerilla activity in the area.

In May 1985, the amount of cropped area is estimated to have been 43 percent of the area cropped in 1973. Ten years after independence, collective farms had been established and guerilla warfare was being conducted by RENAMO in the study zone. Cropland Use Intensity was generally reduced to between 5 and 30 percent in areas where CUI had been in the range of 30-50 percent before independence (August 15, 1973). On May 11, 1985, large areas that in 1973 were cultivated between 5 and 30 percent were apparently uncultivated. It is notable that there was a substantial amount of open water along the Chire River just south of Malawi on May 11, 1985, in contrast to the smaller lakes in the same area on February 24, 1992, and even less standing water on the study dates in 1973, 1994, and 1995. Although there was apparently abundant water during the 19841985 cropping season, the civil war had continued since 1979, and it is possible that both postindependence agricultural practices and conflict in the area contributed to the reduced area cropped as of May 11, 1985.

By February 24, 1992, the Mozambican civil war had been waged for about 15 years. There had been major displacement and isolation of the population of the study area due to armed conflict, mining of roads, and sabotage of other infrastructure within the study area. The area was also hard hit by the 1991-1992 drought, which drove many people from their homes in search of food and water and seriously reduced crop production even for those who remained. The CUI interpreted from the TM scene of that date indicates only between 4 and 5 percent of the study area being cultivated, with very few, small areas 30 percent or more cultivated.

Following the signing of the Peace Accord in October 1992, refugees and displaced persons moved back into the study area, removal of mines began, inputs to agriculture (e.g., seeds and tools) began to be supplied to local people, and the normal agricultural rains arrived. The CUI interpretation of the May 28, 1994, TM scene indicates that the percentage of the study area under cultivation increased to about 10 percent. Most of the areas cropped between 30 and 50 percent on May 28, 1994, were west of the Zambezi River.

Analysis of the Mozambique Main Study Area shows a trend of decreasing crop production there from 1973 through 1992, the year in which the civil war ended. In comparison to CUI in the study area on February 24, 1992, 9 months before the Peace Accord was signed, CUI interpretation indicates that crop production had doubled 19 months after the Peace Accord was signed. Key factors in the increase in cropland use intensity are that the drought had ended, farmers had reentered the country, land mines were being removed, and inputs to agriculture were being contributed by donors.


Figure 3

Results: Substudy Area CUI -- Five Maps

Figure 3 focuses on a portion of the larger study area, the "substudy area," as defined by the smaller SPOT scene. The substudy area encompasses 253,756 hectares (about 50 km x 50 km, or one quarter of the 1,078,031-hectare study area). Twenty-two percent of the substudy area was cultivated in 1973, whereas only about 9 percent of the substudy area was estimated to have been cultivated as of May 11, 1985. During the 1991-1992 drought, and during one of the last months of the civil war, only 4 percent of the substudy area was estimated to have been cultivated. CUI interpretation of the February 24, 1992, subset of the TM scene covering the same area as the SPOT scene reveals that the conditions of warfare, mined roads, and drought resulted in very little crop production. As of May 28, 1994, hostilities had ceased, many of the roads in the substudy area had been cleared of land mines, refugees and displaced persons had returned, and development assistance and food aid were being provided by USAID and other donors. Agricultural production had increased to include 10 percent of the substudy area.

The CUI interpretation of the SPOT scene of April 20, 1995, shows about 17 percent of the substudy area under cultivation. The trend is clearly one of increase in crop production following the signing of the Peace Accord of October 1992. In April 1995, relatively intense cultivation (50-70 percent) is evident in some portions of the substudy area, south of the road between Sena and Tembe-Tembe.

Figure 4

Figure 4 depicts a decrease in cropped area within the study area subset, which includes an area approximately 50 km x 50 km, followed by an increase. Using the CUI interpretations of the five satellite images makes it possible to determine a general pattern of decreasing crop production, followed by increasing crop production. It is clear that crop production was severely decreased as a result of drought and effects of war, such as migration out of the area. Following the signing of the Peace Accord, it appears that crop production in the study area subset is increasing rapidly. This should result in improved nutrition of local people and increased demands for inputs to agriculture and transportation infrastructure (roads and vehicles) to ship agricultural products to markets outside the area.


Figure 5

Results: Change in Agricultural Intensity -- Two Maps

Between August 15, 1973, and February 24, 1992, crop production decreased in most of the study area subset, but there were increases in cropland use intensity in some areas west of the Zambezi River. Between February 24, 1992, and April 20, 1995, increases in cropland use intensity are evident along the Zambezi River and in alluvial soils associated with watercourses in the area. Decreases in cropland use intensity between 1992 and 1995 are seen in few, small, and scattered areas throughout the study area subset.

The changes in the intensity of cropping between February 24, 1992, and April 20, 1995, are, for most of the study area subset, increases. There are during that period, however, some portions of the study area subset for which cropland use intensity decreased. The reasons for decreases in CUI in some parts of the study area subset, in which overall there is a net increase in CUI, are not clear. It could be that marginal lands that had been cultivated were abandoned upon the cessation of hostilities in favor of more productive lands, or that as the peace holds over time, production areas shift from more isolated to more market-accessible locations.

The detailed map of the study area subset showing change in agricultural intensity between August 15, 1973, and February 24, 1992, also shows road status as of December 1994. It is assumed that all roads in the study area, which were indicated as Mined, De-Mining in progress, or De-Mining completed, were mined on February 24, 1992, 8 months before the Peace Accord was signed. Most of the area, and particularly the portions adjacent to mined roads, shows a substantial reduction in the use of the land for agriculture. There are notable exceptions in the westernmost parts of the scene. These areas, which historically did not contain any medium or large population centers, may have been more secure from land mines and isolated from hostilities than other parts of the study area.

In contrast, the scene representing change in agricultural intensity in the study area subset between February 24, 1992, and April 20, 1995, shows increases almost everywhere in the substudy area. Relatively large increases in cropland use intensity are evident in the area bounded by Charre, Sena, Tembe-Tembe, and the confluence of the Zambezi and Chire Rivers. The roads cleared of land mines as of December 1994 are also indicated, suggesting that removal of mines has been a major factor in enabling the inhabitants of the area to recommence crop production.


Conclusion

The use of cropland use intensity interpretation of satellite imagery of a portion of the Zambezi River Valley at different times over a period of 22 years shows dramatic differences in utilization of the land for agriculture. The devastating effects of war in virtually eliminating crop production in the study area are evident in maps derived from the interpretation of satellite imagery. On a brighter note, results of the October 1992 Peace Accord and subsequent assistance by USAID and other donors are evident in increased cultivation of land, the basis for economic activity in the study area. Periodic analysis of satellite images in the future can yield information about location of agricultural production and, with appropriate ancillary information, location of population.


Back to the main Mozambique article...


Bookmark www.usgs.gov/Earthshots for Earthshots, 8th ed., 12 January 2001, from the EROS Data Center of the U.S. Geological Survey, a bureau of the U.S. Department of the Interior.