Case studies
Geospatial Resources for Freshwater Conservation > Freshwater guides, tutorials, and case studies > Case studies

The Rise and Fall of Water Hyacinth in Lake Victoria and the Kagera River Basin, 1989 - 2001

The content for this case study is from the Nile River Awareness Kit (RAK) Interactive CD-ROM which was produced in 2005. The case study was developed in 2003 by:

Mr. Tom Albright (SAIC, USGS/EROS Data Center, and Department of Zoology, University of Wisconsin, Madison, WI 53706)
Thomas G. Moorhouse and Thomas J. McNabb (Clean Lakes Inc.), and
Dr. Larry Tieszen (USGS).

Information about the project on which this case study is based can be accessed this USGS web page. This journal article can also be accessed here: Thomas P. Albright, T.G. Moorhouse and T.J. McNabb. 2004. The Rise and Fall of Water Hyacinth in Lake Victoria and the Kagera River Basin, 1989 - 2001. Journal of Aquatic Plant Management 42:73-84.

The Nile RAK CD is a collaborative partnership between the Nile Transboundary Environmental Action Project (NTEAP) of the Nile Basin Initiative and a project team led by Hatfield Consultants. The primary objective of the RAK is to use multimedia and interactive tools to support the sustainable management and use of the environment and resources within the Nile Basin. The Nile RAK project supports the objectives of NTEAP for addressing high-priority transboundary environmental issues through an improved understanding of the relationship between water resources and the environment.

The "Rise and Fall of Water Hyacinth in Lake Victoria and the Kagera River Basin, 1989-2001" was conducted by the United States Geological Survey (USGS) and Clean Lakes Inc. and illustrates the effective use of satellite imagery to monitor and aid in the management of an invasive aquatic species.

Water hyacinth (Eichhornia crassipes (Mart.) Solms) has been described as the world's worst aquatic weed. When this exotic plant is introduced or colonizes a previously uninfested area, it may explode into large infestations causing serious disruption to environments, economies, and societies. Aquatic weed species have been defined as, "an aquatic plant (or group of plants) which is not desired by the manager(s) of the water body where it occurs, either when growing in abundance or when interfering with the growth of crop plants or ornamentals". Unmanaged water hyacinth populations create serious impacts that ripple through infested areas. These impacts include: impeding transport of irrigation and drainage water in canals and ditches; hindering navigation; interfering with hydroelectric schemes, increasing sedimentation by trapping silt particles, decreasing human food production in aquatic habitats (fisheries, crops); decreasing the possibilities for washing and bathing; and adversely affecting recreation (swimming, water-skiing, angling) (Pieterse, 1990). Numerous additional impacts have been identified that interrupt and disturb both natural systems and human activities and infrastructure.

September 2000 Ikonos image of Lake Mihindi, with areas of water hyacinth and other aquatic weeds appearing on the lake in pink. Satellite image courtesy of GeoEye, 2005. Click here to enlarge the image.

Study Area:
Lake Victoria is shared by Kenya, Tanzania, and Uganda, who account for 6%, 49%, and 45% of the lakes surface water respectively. The lake is important for the regions inhabitants through the supply of drinking water, power generation, fisheries and food security, transportation, and ecological stability. The economy of the lake catchment has an estimated worth of US$ 3-4 billion annually, with the lake fishery benefiting the livelihood of at least 500,000 persons and having a potential sustainable fishery export value of $288 million (LVEMP, 1996). Social, environmental and economic benefits associated with power generation, tourism, clean drinking water, transportation, biological diversity, and other benefits add significantly to the value of the lake and Kagera River basin economies.

Water hyacinth is distributed throughout the near shore portions of Lake Victoria and up to the headwaters of the Kagera River in the highlands of northern Rwanda. For a full and detailed description of the infestation, please refer to the complete report.
Lake Victoria, its catchment (shown in light cyan), and major rivers. Click here to enlarge the image.

The primary remote sensing task in this study was to discriminate water hyacinth from other image constituents such as open water, land, waves, and other types of vegetation. To accomplish this, a variety of spaceborne and airborne sensors were employed: Landsat TM, Radarsat, Landsat 7 ETM+, JERS SAR, Ikonos. A baseline image mosaic of satellite imagery was created to provide a base for the co-registration of other images. A key element to the monitoring effort was to develop a high-resolution map of the open water areas of Lake Victoria and selected lakes in the Rwanda-Tanzania borderlands area. There are many features on land as well as permanent wetlands with reflectance and backscatter characteristics that resemble water hyacinth. By removing these areas from consideration with a water mask, the task is simplified greatly.

Extraction of potential water hyacinth from water-masked imagery:
This section describes the process of spectrally determining the areas of potential water hyacinth in the images, the areas free of potential water hyacinth, and, when appropriate, the areas obscured by cloud cover and/or image noise. Clouds and noise that obscured observation were placed into the no data category since, as in areas with no data, it was not possible to discern the presence or absence of water hyacinth in these areas. This portion of the analysis varied according to the sensor employed, and is thus described separately. For the purposes of this discussion, potential water hyacinth was defined as areas having a spectral signature or high backscatter characteristic of aquatic vegetation that could include water hyacinth, but also other vegetation, and, in the case of radar data, waves, islands, ships and their wakes, and occasionally image noise.

(From left) a) The Murchison Bay portion of the color-coded RGB-clustered Radarsat ScanSAR wide B image. b) The 4 March 1998 potential water hyacinth component. c) The 26 July 1998 potential water hyacinth component. d) The 5 April 2001 potential water hyacinth component. Click here to enlarge the image.

Post-classification filtering and editing
While in some cases the results of the thresholding and classification were quite satisfactory, it was usually necessary to perform some spatial filtering and/or manual editing in order to improve the quality of the determination of areas covered by water hyacinth. Depending on scene characteristics, a 3 x 3 spatial majority filter was applied, or an elimination routine to generalize the data and eliminate extremely small specks that often correspond to noise. The motivation for these routines is to eliminate small, often noise-related, pixels spuriously misclassified. The primary benefit is in areas on the edge of the mask. Due to small variations in the co-registration of images and different pixel sizes used, there were frequently areas of shoreline on the initial classification that were misclassified as water hyacinth that were more likely land or permanent aquatic or wetland vegetation.

(From left) a) A portion of a 17 December 1999 band 5,4,3 ETM+ image in Winam Gulf, Kenya. b) After masking and unsupervised classification. c) After a 3 x 3 majority filter and after clumps of fewer than 3 pixels were eliminated. Note the disappearance of much of the shoreline fringe and some of the smaller speckled patches of water hyacinth. Click here to enlarge the image.

Final determination of water hyacinth was made with manual editing. For incidences where spectral means and filtering would not adequately identify water hyacinth, "false positives" were removed or, in rare cases, pixels were reclassified as water hyacinth where "false negatives" had occurred. Such manual edits were especially crucial for extracting water hyacinth information from lower resolution satellite imagery, which is subject to significant error caused by co-registration and resolution issues. Manual editing is facilitated by interactively overlaying co-registered images from multiple dates, allowing discrepancies and errors to be efficiently spotted.

The complete report features results from 6 geographic areas in Uganda, Tanzania, Kenya and the Rwanda-Tanzania border lakes. This summary only includes the results for Murchison Bay Uganda. For details of the full results of this study, please refer to the complete report.

With its location at the mouth of the Kagera River, its numerous protected bays and gulfs, and the proximity to major population centres, the Ugandan portion of Lake Victoria presents an environment that is both conducive to water hyacinth infestation and highly sensitive to its effects. In addition, prevailing winds tend to blow water hyacinth north into Uganda waters. The amount of water hyacinth identified in Uganda during the study is shown in the figure below. In Uganda, large amounts of water hyacinth (> 3000 ha) were present from the earliest countrywide image in February 1996 until a peak of 4732 ha on 4 March 1998. Following this was a sharp reduction to 2147 ha on 26 July 1998 and further reduction until a low of 53 ha was measured on 5 April 2001. Numerous bays and gulfs have experienced sizeable infestations and during some periods, large quantities of water hyacinth were found floating in open waters in Uganda. The Murchison Bay and greater Napoleon Gulf areas, however, distinguish themselves as among the most heavily infested and having generated the most concern among researchers, managers, government officials, and the general public.

The figure also shows the evolution of water hyacinth distribution and coverage for Murchison Bay. The graph includes data from Schouten, van Leeuwen, and others (1999), which, unlike most of the other preexisting reports, are high-confidence estimates derived from remote sensing. The available imagery and reports indicate that a rapid increase in water hyacinth occurred during 1994, followed by a peak of 1974 ha (8.6% of bay) on 19 January 1995, and a period of abundant water hyacinth, ranging from 1140 ha to 1522 ha on dates observed between 1996 and 1997. During these periods of abundance, giant mats covering 100s of hectares could be found in inner Murchison Bay, Wazimenya Bay, and Gobero Bay. In 1997, a steady decline occurred until, in 1999, there were only 15 and 1 ha detected in March and July, respectively. In 2001, there were reports of increased water hyacinth in Murchison Bay. Indeed, a slight increase to 35 ha was apparent in the data in January 2001. These quantities are the best estimates of water hyacinth present at the time the image was acquired. Due to strong winds and the highly mobile nature of water hyacinth (including even the largest of mats), estimates could differ greatly between morning and evening, and between seasons, due to changes in wind direction. The first four data points and the graph, for instance, were from two different times of day on two different dates and reveal how daily wind cycles can affect measured water hyacinth amounts. The significant "reduction" that occurred between 1995 and 1996 was thus most likely caused by a wind induced migration of water hyacinth out of Murchison Bay into other parts of the lake.

In late 1995, two species of Neochetina weevils were released into the Uganda portion of Lake Victoria. However it was not until February 1997 that weevil feeding activity became visible on plants in Murchison Bay. Weevils multiplied rapidly, attaining an average number of 13.8 weevils/plant in 1998, and 24.7 weevils/plant in 1999 on Lake Victoria in Uganda. By late 2001, weevil numbers had declined to an average of 8.8 weevils/plant (Uganda National Agriculture Research Organization, 2002). Weevil monitoring exercises carried out by Clean Lakes Inc. within inner Murchison Bay indicate that weevil numbers had declined to 1.2 weevils/plant for stationary water hyacinth growing along the shoreline and to 2.3 weevils/plant for floating mats of water hyacinth by January 2002.

The increasing weevil populations can be partly attributed to the dramaticdecline in water hyacinth. Between 1997 and 1998, however, another event took place that is difficult to quantify as far as its impacts on water hyacinth. East Africa was hit by an El Niño weather phenomenon during the last quarter of 1997 that continued well into the first half 1998. In mid October 1997, prior to the beginning of the rains, the lake level was at near all time lows of 11.26 meters (datum level of 1121.65 a.s.l.). These lows had only been experienced at two other points in time (both low periods in 1994) during the previous 40 years. The lake level then climbed to 12.96 meters by mid May 1998 - a change of 1.70 meters in a period of seven months. This rapid rate of rise over the period was matched only once by an event that occurred in 1962/1963 as evidenced by a lake level monitoring table that exists showing data starting in 1899. Already in a weakened state due to insect attack, the water hyacinth also experienced heavy weather conditions that created wave action, which mechanically damaged large quantities of water hyacinth.

Evolution and distribution of water hyacinth coverage in Murchison Bay, Uganda. Note that the first four data points on the graph are estimated derived from Synoptics BV. Click here to enlarge the image.

Analysis of satellite imagery collected between 1994 and 2001 confirms the severity of water hyacinth infestation in Lake Victoria and the Rwanda-Tanzania borderlands lakes. The northern portions of the lake in Uganda and Kenya were most severely infested, with Winam Gulf experiencing the most water hyacinth detected in the study. In most locations, the infestation reached a maximum in 1997 or 1998, with a lakewide maximum of approximately 20,000 ha in November 1998. By 2001, however, the severity of the water hyacinth infestation in Lake Victoria was dramatically reduced relative to 1998.

The degree to which each of the control measures and environmental factors are responsible for the decline in water hyacinth cannot be determined from this study alone. It does appear that biocontrol with weevils were effective in reducing the abundance of water hyacinth, in at least some populations. However, other E. crassipes populations experienced declines prior to, or concurrent with, the introduction of weevils. Heavy winds, rain, and waves associated with a stong 1997-1998 El Niño event may have dislodged and battered mats of macrophyes contributing to their demise. The timing of the decline in many parts of the lake is consistent with this explanation. Refer to full article for a more in-depth discussion on the decline in water hyacinth.

Finally, several rapid increases in different parts of the lake have been documented, which should serve as a reminder of the rapidity with which water hyacinth is capable of expansion. We therefore conclude that the region must continue with aggressive and active monitoring and management strategies in order to reduce the possibility of resurgence to extremely high levels.

Contact information
For more information about the NTEAP please contact:
Nile Transboundary Environmental Action Project Gedion Asfaw, Regional Project Manager P.O. Box 2891 Khartoum, Sudan Web site:


Grant S. Bruce, Vice-President and Partner, Hatfield Consultants Partnership, Email: