Potential of aquatic weeds to improve water quality in natural waterways of the Zambezi catchment

Study area description

The Zambezi is Africa’s fourth largest river and the most important draining into the Indian Ocean. We focused our analysis on four sub-catchments of the Zambezi River in southern Zambia where invasions of Water Hyacinth have been reported in literature and/or which we observed from satellite imagery. These rivers are: Kafue, Chongwe, Maramba and Little Chongwe (Fig. 1). The catchments differ in size and dominant human land-use.

The Kafue River is a major tributary of the Zambezi and is characterized by two major hydropower dams bracketing the large Kafue Flats floodplain: Itezhi-Tezhi Dam upstream and Kafue Gorge Upper Power Station (which we subsequently refer to as simply “Kafue Gorge Dam”) downstream. Operation of these dams has greatly altered the hydrology of the lower reaches of the Kafue River leading to changes in the seasonal biochemical functioning of the Kafue Flats^16,17 compared to the Barotse Plains, a reference floodplain unmodified by dams¹⁸. The lower end of the Kafue Flats has expansive coverage of sugar cane cultivation and the downstream river reach above the reservoir formed by the Kafue Gorge Dam (“Kafue Gorge Reservoir”) receives urban and industrial wastewaters from Kafue. Local reports implicate these nutrient sources as drivers of the recurrence of Water Hyacinth blooms in this area including the Kafue Gorge Reservoir. The Kafue River was subject to nutrient loading studies and intensive weed control campaigns from 1998 to 2000¹⁹. Confusingly, control efforts are described as “ineffective” in the literature¹⁹, and yet the weed problem abated for a decade from 2001 to 2011. Water Hyacinth is back in recent years, following a seasonal pattern of coverage on the Kafue Gorge Reservoir surface. Mechanical control efforts have also been resumed (personal observations of the authors, February 2019).

The Chongwe River drains one of the more densely populated catchments in Zambia, including parts of the capital, Lusaka, and several nearby townships. After tumbling down the same geographic escarpment as the Kafue, it meets the Zambezi in the Lower Zambezi National Park, approximately 55 km downstream from the confluence of the Kafue with the Zambezi River.

The Maramba River meets the Zambezi just upstream of Victoria Falls and drains a small, but highly urbanized catchment containing most of Livingstone, a popular tourist destination for its convenient access to the falls. According to reports, overflow discharge from Livingstone’s wastewater treatment pools enters the Maramba just upstream of the confluence with the Zambezi¹⁹. During the latter stages of the dry season, low flows in the Maramba allow the river to build up a dense coverage of Water Hyacinth. These mats are seasonally flushed into the Zambezi when the rains at the end of the dry season restore the flow of the Maramba^20,21.

The Little Chongwe River meets the Kafue just 8 km upstream of the Kafue’s confluence with the Zambezi. Its small catchment drains a small portion of a large array of pivot irrigation agriculture.

Assessment of floating vegetation cover

To assess levels of floating vegetation, we used a combination of field surveys and analysis of satellite imagery. Floating vegetation on the Kafue Gorge Reservoir can clearly be detected with Landsat imagery. We used Google Earth Engine to group all available images in the Landsat archive in two-month intervals and extracted floating vegetation cover from 1990 to 2019²².

To assess floating vegetation cover on the Maramba, we visually inspected all available historical high-resolution imagery in Google Earth Pro (35 images from 2005 to 2019). We hand-digitized floating vegetation cover over the lower reaches of the river for which channel morphology was clearly and consistently visible and resolvable (2.3 km of linear stream reach). Since there were multiple cases of complete coverage on the Maramba River surface we interpreted complete coverage as a conservative estimate for annual Water Hyacinth biomass export in the catchment.

Relatively few high-resolution satellite images are available in Google Earth Pro for the lower reaches of the Chongwe River and the Little Chongwe (8 each) and they do not provide complete seasonal coverage, making it impossible to apply the same process as for the Maramba. Instead we simply hand-digitized the area of visible floating vegetation from the image with the most cover (16 June, 2016) and added to this an estimate of the area of fringing floating vegetation based on four ground-based inspections of the coverage in 2018 and 2019. We estimated that fringing vegetation has a coverage of 0.5 m wide running for 1.2 km upstream of the confluence with the Zambezi. Hippo presence prevented us from surveying further upstream and we assumed no floating vegetation presence beyond this point, making our estimates conservative.

To convert areal coverage to biomass and nutrient content, we used a synthetic mean of biomass per area from 15 studies and synthetic mean nitrogen and phosphorus content from 14 and 17 studies respectively to estimate the total pool macronutrients bound to Water Hyacinth (Supplemental Tables S1,S2). The nutrient content of Water Hyacinth varies substantially between and within studies, presumably in response to differences in nutrient availability and limitation is experiences as it grows and ages²³. The Water Hyacinth in Zambian waters spends different parts of its life cycle in very different nutrient settings, ranging from urban wastewater effluent to hyperoligotrophic natural rivers, and we should therefore expect that its nutrient content should also vary substantially in space and time. The synthetic mean nutrient content from varied nutrient settings probably represent a reasonable estimate the nutrient content of Water Hyacinth in Zambia, which grows under a similarly wide range of nutrient conditions. We report uncertainty surrounding mean nutrient content using the standard error of this mean.

Nutrient sampling

In order to assess the potential for floating vegetation to sequester nutrients from river systems we modelled river nutrient loading for four Zambezi tributaries infested with invasive floating vegetation and estimated the amount of nutrients bound within floating vegetation biomass. We collected surface water samples from these rivers: the Maramba, the Chongwe, the Little Chongwe and the Kafue near the town of Kafue, once every three months for a year starting in March 2018. To evaluate the nutrient environment in the backwaters where we expect Water Hyacinth to be originating, we also sampled two additional sites in the larger catchments in November 2019. In the Kafue Catchment, we sampled two drainage canals conveying industrial wastewater from Kafue. In the Chongwe catchment, we sampled two points on the Gwerere River, an urban stream draining densely populated portions of Lusaka. We sampled additional points on the Kafue and Zambezi Rivers to provide reference nutrient conditions for the region’s major rivers. These sites include the Kafue River: near Hook Bridge, below the Itezhi-Tezhi Dam, near Mazabuka, near Chirundu; and the Zambezi: near Livingstone, below Kariba Dam, near Chirundu, just above Lower Zambezi National Park (for coordinates and exact sampling dates, see Supplemental Table S3).

We passed nitrate and phosphate samples through pre-combusted, pre-weighed glass fiber filters and collected unfiltered samples for analysis of bulk nitrogen and phosphorus content. We collected all samples in triplicate and kept them cool during handling and until analysis in the laboratories at Eawag, Switzerland. We digested unfiltered water samples via autoclave with an alkaline potassium peroxidisulfate solution. We analyzed filtered water samples and digests colorimetrically using a Skalar (Breda, Netherlands) SAN++ automated flow injection analyzer following²⁴ and standard procedure ISO 13395:1996. We note that acid digestion of unfiltered surface water is susceptible to underestimation of total phosphorus because of settling of clay particles during sample storage and sub-sampling leading to a bias against potentially phosphorus-rich particulate matter²⁵. We assume that such underestimation in our samples is relatively minor because of low stream velocities we observed at our sampling sites, dominance of sandy rather than clay-rich soils in the region, and because we observed low C:N ratios in particulate matter (Supplemental Table S4), indicating a low-density microbial (rather than mineral) composition. Nevertheless, we refer to our P data as “digestible” rather than “total,” to allow for this potential underestimation.

Discharge calculations, nutrient load estimation and relative importance

We calculated average monthly discharges based on hydrographs collected by the Zambia Electricity Supply Corporation (ZESCO) between 1977 and 2017 for the Kafue Gorge Dam. For the Maramba, Chongwe and Little Chongwe Rivers, we estimated discharge by generating a catchment area: discharge curve using nearby stations from the Global Runoff Data Centre. We estimated rainfall for the Maramba catchment through monthly means from the Climate Hazards Group InfraRed Precipitation with Stations (CHIRPS) dataset²⁶.

In order to estimate annual river nutrient loading, we multiplied mean annual discharge by mean digestible phosphorus and mean digestible nitrogen concentrations from our surface water sampling. Our reliance on using the mean of four seasonal concentration values may not fully capture the seasonality of load and is susceptible to effects from outliers. In order to check for evidence of bias from outliers, we also calculated loading based on median concentrations and found that differences translated into a maximum decreased relative importance of plant-bound nutrients of approximately 2%. This is a minor difference compared to other sources of error, such as those stemming from uncertainty in plant biomass per area and nutrient content of plant biomass.

We estimated the relative contributions of plant-bound versus bulk surface water nutrients to riverine nutrient export, by simply calculating the percentage of each relative to their sum. We used the product of standard errors of mean plant biomass and nutrient content values from literature to generate uncertainty envelops around our estimations for the importance of plant-bound nutrients. Here we assumed that the peak plant biomass we detected represents an annual export of intact plant material from each watershed. This assumption provides a conservative underestimate of the importance of plants because the choke points where the plants accumulate seasonally are unlikely to be 100% efficient as traps. Some plant biomass may be exported without being accounted for, but at least in the case of the Maramba and Kafue where hydrologic conditions favor clear trapping and flushing seasons, this is the most reliable approach available to estimate annual plant-bound nutrient export.

Landcover analysis

The river catchments were extracted from the HydroSheds global database²⁷. For the Kafue River, we considered only the catchment below the Itezhi-Tezhi Dam, as we expected the dam to interfere with the nutrient transport from further upstream²⁸.

To calculate the area of urban and agricultural land within each catchment we used fractional land cover maps for 2015 from the Copernicus Global Land Service²⁹. Using the raster package³⁰ in R, we extracted pixel values within each catchment polygon and then calculated the area for all pixels with a fraction of more than 50% for the respective class.

All analyses and figures were completed using R version 3.6.1³¹.

Source: Ecology - nature.com