Introduction

Since 1990, unprecedented loss of wetlands has occurred. To date, more than half of the world’s wetlands have been transformed into agricultural land or industrial and urban areas (Millennium Ecosystem Assessment 2005; Rebelo et al. 2009, 2010; Verhoeven and Setter 2010). A number of studies have reported immense pressure on urban wetlands due to competing demands for land to accommodate urban and industrial growth, infrastructural developments, and urban agriculture. Some examples include the East Kolkata Wetlands (EKW) in India, which are the world’s largest resource recycling ecosystem (Allen et al. 2016), the Champaign County in Illinois, USA (McCauley and Jenkins 2005) and urban expansion into wetlands in Beijing, China (He et al. 2011), in Dhaka, Bangladesh (Sultana et al. 2009) and in Uganda (Abebe 2013; Lukooya et al. 2013; Twesigye 2011; Vermeiren et al. 2012). In spite of the drive for wise use of wetland resources, which is defined as ‘the maintenance of their ecological character, achieved through the implementation of ecosystem approaches, within the context of sustainable development’ (Ramsar 2010), not much progress has been realised. Wetlands are well-known for their ability to store, purify, and gradually release water. In so doing, wetlands control floods and provide water for life. Often, the functioning of wetlands is dependent on the dominant vegetation (Kansiime and Nalubega 1999; Kansiime et al. 2007). Depending on the functions in question, which in the context of this study are flood control and wastewater treatment, emergent wetland vegetation, particularly papyrus has been reported to be a better candidate (Chapman et al. 2001; Kanyiginya et al. 2010). Concern is increasing that direct consumptive use of wetland resources is occurring at the expense of essential bio-physical and hydro-chemical processes (Apuyo 2006; Horwitz et al. 2012; Kaggwa et al. 2009). In the quest for wetland products, humans transform wetlands by draining the marsh and clearing the natural vegetation to maximise private benefits such as land for cultivation, settlement, industrial sites, and building materials. In the context of this study, encroachment on wetlands refers to human modifications which compromise the ability of wetlands to perform their ecological functions. While this definition may not be fully inclusive, it provides insight into the link between wetland use and conservation.

In Uganda, human communities using wetlands comprehend the products they get from wetlands, so much so that for many, wetlands are the sole source of livelihood (Kabumbuli and Kiwazi 2009; Nakangu and Bagyenda 2013). However, the link between wetland conservation and their ecosystem services are often not well understood or are simply taken for granted (Kansiime et al. 2007; Lukooya et al. 2013; Nakangu and Bagyenda 2013). Furthermore, some authorities perceive conservation of wetlands as hampering economic development, and subsequently afford it a lower priority relative to other issues (OECD 2006; Ostrovskaya et al. 2013). Encroachment on the Nakivubo wetland, which is the central wastewater drainage system for Uganda’s capital Kampala, is associated with significant public health and environmental risks (Fuhrimann et al. 2014, 2015). Prominent among these is the increased risk of flooding, vulnerability of human communities occupying wetland areas, and the pollution loads that end up in Lake Victoria, the city’s main source of water supply (Banadda et al. 2009; Fuhrimann et al. 2015). Notably, insufficient capacity in government to effectively ensure wise use of wetlands is among the key limitations (Ostrovskaya et al. 2013); specifically the lack of appropriate and up-to-date information for policy implementation at local levels (MWE 2012; WMD-MWE et al. 2009).

Draining and conversion of wetlands in Uganda was unchecked or even promoted for purposes of malaria control (Omolo-Okalebo 2011), cultivation, and animal grazing until 1986 when the National Resistance Movement (NRM) government through the Ministry of Environmental Protection banned further wetland conversion (Apuyo 2006; Nakangu and Bagyenda 2013). Although the government owned wetlands on behalf and for the good of all citizens, its control over their use was limited by inadequate legislation. Subsequently, the need to regulate wetland use led to formulation of the National Wetlands Policy, the National Environment Statute, and the National Guidelines for Wetland Resource Developers in 1995, and later the National Environment Regulations (for wetlands, river banks, and lake shore management) in 2000 (NEMA 2000). These documents provide guidelines for sustainable use of wetland resources. But there are still challenges in regulating small-scale human activities not subject to environmental impact assessment (EIA); activities which, collectively, have significant impacts on wetlands.

Several scholars emphasise the need to monitor and control human activities in wetlands for the sake of ecosystem services and values (Banadda et al. 2009; Kansiime and Nalubega 1999; Kansiime et al. 2007). However, the monitoring and control has not kept pace with the rate of encroachment. To this end, there is insufficient research explicitly quantifying the spatiotemporal extents of encroachment on wetlands. According to Huising (2002), human activities in wetlands, especially agro-forestry, have been traced as far back as the colonial times, long before wetland boundaries were drawn. A number of aspects complicate the control over wetlands, including the lack of data and unclear boundaries (WMD-MWE et al. 2009). The process of demarcating physical boundaries of wetland areas has dragged on for long, despite the growing pressure to encroach on the seemingly redundant chunks of land. The pressure arises from inadequacy of land to accommodate urban and industrial growth, rural–urban migration, growth of informal settlements, land tenure dynamics, protection of urban peasants’ livelihoods, and food security among others (Lwasa et al. 2012; Nyakaana et al. 2007; Waters 2013).

While several studies have reported on human activities in Kampala’s wetlands (Huising 2002; Kabumbuli and Kiwazi 2009; Kansiime and Nalubega 1999; Twesigye 2011; Vermeiren et al. 2012), explicit quantification of the spatiotemporal extents of human activities in the wetlands have not received much attention. In the few studies which have mapped land cover (Abebe 2013; Twesigye 2011; Vermeiren et al. 2012), the low resolution of the datasets used (i.e. Landsat imagery) did not permit detailed analysis of land cover at a local scale. Huising (2002) provides a historical examination of land cover changes in Nakivubo wetland from 1955 to 1999 and discusses the potential of using satellite imagery and aerial photos for change detection but does not spatially quantify any changes described in the study.

Explicit quantification of the spatial extents of changes in wetlands, showing human activities, is particularly vital because it provides concrete information at a local scale. A study on wastewater treatment by Nakivubo wetland (Kansiime and Nalubega 1999) provides 1996 estimations of the total area of the wetland covered by natural vegetation and cultivated area in the lower part of the wetland (bordered by the railway line to the north and the Murchison bay to the south). The authors compared their estimates with an earlier study (Taylor 1991) and reported that the total area of Nakivubo wetland covered by natural vegetation had reduced by approximately 14 % in 5 years (from 2.2 km2 in 1991 to 1.9 km2 in 1996), while cultivated area in the lower Nakivubo wetland had increased by more than 350 % (from <0.05 to 0.225 km2) in the same period. Their estimations of cultivated area, however, cover only part of the wetland and do not capture other forms of encroachment. These earlier studies, which now date almost 20 years back paint a clear picture of the increasing anthropogenic transformation of the wetland and certainly echo the need for up-to-date information to guide wetland conservation and risk reduction endeavours.

Given that the biggest threat to wetlands is posed by human activities (Kansiime et al. 2007), it is important to generate realistic and up-to-date information. In this paper, high resolution aerial photos and satellite imagery are used to classify and map recent land cover and provide a historical analysis of land cover change in Nakivubo wetland over the past 12 years, 2002–2014. We discuss the drivers and implications of these changes in the light of the accelerated loss of the functional wetland vegetation, and increasing risks of flooding and pollution.

Methods

Study area

Nakivubo is a gazetted wetland of prime importance located on the northern shores of Lake Victoria’s Murchison bay in Uganda (Fig. 1). The wetland is located between latitudes 00° 17′ and 00° 19′ N and longitudes 32°37′ and 32°39′ E, at an altitude of 1135 m above sea level, in close proximity to the central business district (CBD), the industrial area, and the lake. It discharges only 4 km away from the intake of the city’s water supply (Banadda et al. 2009). Though it is part of a network of wetlands that drain Kampala city, Nakivubo is the largest and receives most of the wastewater from the CBD, effluent from sewage treatment plant, stabilisation ponds, and the adjacent industrial area (Emerton et al. 1999; Kansiime and Nalubega 1999). According to the wetland boundaries from the Ugandan Ministry of Water and Environment (MWE), the Nakivubo wetland covers an area of about 5.29 km2 and serves a catchment area of slightly over 50 km2 of the urban neighbourhood drained by the Nakivubo channel (Kansiime and Nalubega 1999). It is accessible along the railway line that traverses the entire length of the wetland to Port Bell (Fig. 1).

Fig. 1
figure 1

Map of Uganda showing location of the study area in the Nakivubo wetland system

Full size image

The Nakivubo wetland lies within the equatorial belt with a moist sub-humid climate and has bi-annual rainy seasons; March to May and September to November. However, a number of studies have reported increases in seasonal variability (Buotte et al. 2016; Cooper and Wheeler 2015; Ide et al. 2014; Shuaib Lwasa 2010; Nsubuga et al. 2014; Tolo et al. 2014). In Uganda, given its dependence on rain-fed agriculture, critical climate-related changes are with regard to increased/reduced precipitation and increasing temperature (Ide et al. 2014; Nsubuga et al. 2014; Orlove et al. 2010; Tolo et al. 2014; UN-Habitat 2012). The rains are linked to the Intertropical Convergence Zone (ITCZ), altitude, local topography as well as the lake; with short-duration tropical thunderstorms being particularly common around Lake Victoria and Kampala area (Kansiime and Nalubega 1999). Given the extensive paving, compacted ground and roof area in urban neighbourhood, the thunderstorms are often followed by heavy runoff and flooding in low-lying areas. According to Kansiime et al. (2007), the dominant natural vegetation in the permanently inundated part of Nakivubo wetland is emergent comprising of Cyperus papyrus, grading to dry land through cattails (Typha sp.) and common reeds (Phragmites sp.) and a large area on the north east side covered by Miscanthidium violaceum. Although the Nakivubo wetland is categorised as a permanent wetland (Abebe 2013; Namakambo 2000), only part of the wetland is permanently inundated i.e. the papyrus dominated area buffering the lake. Most of the upper Nakivubo wetland is rather seasonally unaudited. The seasonal part of the wetland was dominated by grassland and shrubs which have been largely transformed into crop fields (Kansiime and Nalubega 1999).

Data types and sources

The data used include full-colour aerial photos taken in July 2010 (0.50 m resolution) from KCCA, and two high-resolution multi-spectral satellite images: a 0.62 m Quickbird image (captured on 3 April 2002) and a 0.50 m Pleiades image (captured on 5 December 2014). Both satellite images were acquired from the vendors (DigitalGlobe and Airbus Defence and Space, respectively), through the South African National Space Agency (SANSA). It is worth noting here that the image data used were not all captured in the same season. Except for the 2010 data, the 2002 and 2014 data were both in rainy seasons since in 2014 the rainy season continued through December. This could have affected the extent of certain land cover types such as water, since rains increase the lake water levels or create other pools of water. Also, although seasonality can affect vegetation, its effect in the Nakivubo wetland was not deemed significant since the lower part of wetland is permanently inundated. The image data was complemented by five key informant interviews (KIIs) with stakeholders and four focus group discussions (FGDs); of seven participants each. The KIIs constituted of two senior wetland officers from the Wetlands Department at the Ministry of Water and Environment (MWE), the Environment and Sanitation Specialist in the Directorate of Public Health and Environment at Kampala Capital City Authority (KCCA), the Chairperson—Nakivubo Farmers Association, and the Safety Manager for a non-governmental organization (NGO)—Hope for Children based in Namwongo, adjacent to the Nakivubo wetland. The FGDs were held with members of the Nakivubo wetland community, specifically landlords, tenants, male farmers, and female farmers.

Data processing and analysis

Data pre-processing

The multi-spectral satellite images were pan-sharpened with their respective panchromatic bands to improve resolution using PCI Geomatica 2014. All image data were terrain corrected and standard georeferenced to UTM zone 36 N and WGS 84. The analysis was limited to the extent of the Nakivubo wetland as identified by the Wetlands Department at the Ugandan Ministry of Water and Environment. Sample points for both training the classifier algorithm and for accuracy assessment were concurrently selected in ArcGIS to avoid inadvertent use of the same samples for both purposes. The sample points were verified against ground-truthed GPS points. The classification system decided on for this study consisted of the following seven classes: built-up, cultivated, grassland, wetland vegetation, trees and shrubs, bare, and water. These classes represented the main land cover in Nakivubo wetland and also were comparable to other studies on the impacts of urbanisation on wetlands (Cai and Wang 2013).

Object-based classification

Object-based classification is the process of grouping image pixels to form objects (segmentation), and subsequently assigning the objects to real-world classes (classification) (Blaschke et al. 2008, 2014; Campbell and Wynne 2011). Each of the image datasets was segmented in eCognition 9.0 to create unique objects corresponding to features in the images. Appropriate segmentation scale parameters of 50, 50, and 30 were used for the 2002, 2010, and 2014 datasets, respectively. Training objects were used to train a support vector machine (SVM) classifier. Classification was based on mean reflectance values of bands and Normalized Difference Vegetation Index (NDVI) values. No NDVI was computed for the aerial photos due to their lack of a Near Infrared (NIR) band. The classification was visually inspected for any misclassification of features, and where necessary, misclassifications were reclassified using rulesets based on feature information such as NDVI values and relational information e.g. distance. Extensive manual correction was performed based on expert knowledge in order to correct any further misclassifications, especially for the vegetated feature classes.

Accuracy assessment

Accuracy assessment was based on the spatial agreement between the known classes of the collected reference points and those of the classified raster datasets. For each classified raster, a confusion matrix was generated from which values for overall accuracy, Kappa, errors of omission and errors of commission could be calculated. All classifications yielded overall accuracies above 83 %, with Kappa statistics of 0.82, 0.80, and 0.89 for 2002, 2010, and 2014, respectively. The accuracy assessment details for each classification are summarised in Table 1.

Table 1 Summary of accuracy assessment
Full size table

Change detection

All classification exports were converted to polygon vector layers for spatial analysis in a GIS environment. Cross-tabulation based on land cover class was performed in ArcGIS 10.1 in order to determine total area changes in classes from 2002 to 2010, from 2010 to 2014, and from 2002 to 2014. This also allowed the generation of inter-class change statistics. In order to do spatially congruent (site-specific) analyses of inter-class changes, the vector layers for all three classifications were combined through a union operation. This allowed the generation of statistics and change maps that provide spatially referenced descriptors of changes between classes over time.

Handling of data from KIIs and FGDs

Upon transcription, qualitative findings from KIIs and FGDs were grouped into relevant themes and have been presented as narratives or direct quotes to elaborate on the patterns of encroachment on the wetland and the apparent driving forces.

Results

The results below provide spatiotemporal extents of land cover in Nakivubo wetland (for the years 2002, 2010, and 2014), spatially congruent land cover changes (for the periods 2002–2010, 2010–2014, and 2002–2014), and the drivers of increasing encroachment on the wetland.

Spatiotemporal extents of land cover

The spatiotemporal extents of land cover are shown in Fig. 2, while the corresponding total area and percentage covered by each of the land cover types in 2002, 2010, and 2014 are presented in Table 2.

Fig. 2
figure 2

Spatiotemporal land cover in Nakivubo wetland (2002, 2010, and 2014)

Full size image
Table 2 Total area (in m2) and percentage per land cover class for each classification date
Full size table

In 2002, the largest part of Nakivubo wetland was cultivated, grassland, and wetland vegetation. Analysis of land cover changes as a percentage of the total area showed that 62.7, 53.0, and 68.8 % of the area changed to different classes over the periods 2002–2010, 2010–2014, and 2002–2014, respectively. These results show that land cover in Nakivubo wetland is highly dynamic. Noticeably, built-up density increased along the peripheries of the wetland while wetland vegetation decreased significantly through the entire period. By 2014, most of the wetland vegetation that buffered the lake had been converted into cultivated, grassland and water, and a new road access to the lake via the wetland had been created.

Land cover changes within each class, including inter-class conversions were plotted as a percentage of their 2002 areas (Fig. 3). Between 2002 and 2014, built-up area, water, trees and shrubs, and grassland increased by about 403, 174, 66, and 29 %, respectively, while wetland vegetation, bare and cultivated areas decreased by about 62, 33, and 18 %, respectively.

Fig. 3
figure 3

Overall change in class areas as a percentage of 2002 over time, notice the sharp increase in built-up and water classes

Full size image

Spatially congruent land cover changes

In order to quantify the areas that changed over time, areas of inter-class conversions were calculated for the periods 2002–2010, 2010–2014, and 2002–2014 and are present in Tables 3, 4, and 5, respectively. In addition, we categorised ‘Grassland, Wetland vegetation, Trees and Shrubs, Water’ as ‘Natural classes’, and ‘Built-up, Cultivated, and Bare’ as ‘Artificial classes’. Aggregated conversions from natural classes to artificial classes between 2002–2010, 2010–2014, and 2002–2014 were 37 % of their 2002 cover, 38 % of their 2010 cover, and 35 % of their 2002 cover, respectively.

Table 3 Inter-class land cover changes (2002–2010), shaded diagonals indicate areas of no change
Full size table
Table 4 Inter-class land cover changes (2010–2014), shaded diagonals indicate areas of no change
Full size table
Table 5 Inter-class land cover changes (2002–2014), shaded diagonals indicate areas of no change
Full size table

Results in Table 3 show that by 2010, about 41.2 % of the original 2002 wetland vegetation was still intact while 1.4 % had been converted to built-up, 25.8 % to cultivated, 26.4 % to grassland, 1.3 % to trees and shrubs, 3.1 % to bare, and 0.9 % to water. Despite these conversions from wetland vegetation, some areas e.g. about 9.1 % of cultivated, 17.7 % of grassland, 9.9 % of trees and shrubs, 17.2 % of bare, and 34.6 % of water in 2002 had converted to wetland vegetation in 2010.

Results in Table 4 show that by 2014, about 24.6 % of the 2010 wetland vegetation was still intact while 1.1 % had been converted to built-up, 26.9 % to cultivated, 28.4 % to grassland, 1.5 % to trees and shrubs, 2.7 % to bare, and 14.8 % to water. Conversely about 6.7 % of cultivated, 7.7 % of grassland, 4.1 % of trees and shrubs, 3.0 % of bare, and 21.4 % of water had converted to wetland vegetation in between 2010 and 2014.

If we assess the overall changes from 2002 to 2014 (Table 5), we find that by 2014, only about 20 % of the original wetland vegetation cover in 2002 was still intact; about 3 % had been converted to built-up, 30 % to cultivated, 30 % to grassland, 2 % to trees and shrubs, 2 % to bare, and 14 % to water. Gains in wetland vegetation over this period came predominantly from grassland (8.5 %), bare (8.6 %), and water (9.4 %).

In order to show locations of the areas that changed as quantified in Tables 3, 4, and 5, spatially congruent maps were generated for each year. Spatiotemporal conversions from wetland vegetation to other classes for the periods 2002–2010, 2010–2014, and 2002–2014 are shown in Figs. 4, 5, and 6, respectively.

Fig. 4
figure 4

Conversions from wetland vegetation to other classes between 2002 and 2010

Full size image
Fig. 5
figure 5

Conversions from wetland vegetation to other classes between 2010 and 2014

Full size image
Fig. 6
figure 6

Conversions from wetland vegetation to other classes over the whole period (2002 to 2014). Note the dominance of the cultivated and grassland classes, especially towards the lake

Full size image

Rate of loss of wetland vegetation

Overall, the analysis of losses and gains shows a 62 % loss of wetland vegetation between 2002 and 2014 (Fig. 7). Site-specific trend line represents changes from wetland vegetation to other classes in areas that were originally covered by wetland vegetation at the reference year (2002). The overall trend line represents changes in the total area covered by wetland vegetation. This also includes areas that changed from other land cover classes TO wetland vegetation (thereby also encompassing ‘gains’). The differences between the overall and site-specific trend lines indicate the magnitude of the gains (i.e. areas that converted to wetland vegetation), expressed as a percentage of the 2002 wetland vegetation cover.

Fig. 7
figure 7

Loss of wetland vegetation as a percentage of 2002 area. The site specific curve describes the change of the original 2002 wetland vegetation areas, while the overall curve describes the total change in area of the wetland vegetation classes (including both gains and losses) over time

Full size image

Some of the drivers of increasing encroachment on the wetland

The contextual drivers of increasing encroachment on Nakivubo wetland that emerged prominent from FGDs and KIIs are presented below under three themes: land ownership, displacement of farmers, and the lack of coordination among stakeholders.

Land ownership

Land ownership in the wetland area was mentioned among the key barriers limiting the local authority’s control of land use. The 1995 Ugandan constitution recognises four land tenure systems i.e. customary, mailo, freehold, and leasehold. According to the Kampala Capital City Authority, these several land tenure systems complicate planning, especially where ownership is not by government. Some of the people who claim ownership of land in wetland areas also possess appropriate documentation to guarantee their security of tenure.

Quote:

It is difficult to control what happens where you do not own or access… we are engaging land owners who claim to have land titles obtained before 1995 constitutional land reforms, which within the provisions of the law are legal, while titles obtained after 1995 are illegal. The best we can do, when owners have legal titles is to engage them to only implement projects that are within regulated activities described in the wetlands, river banks and lake shore regulations (KI Supervisor Environmental Management KCCA).

Displacement of farmers

Farmers explained that they are compelled to cultivate further downstream into the wetland because they are displaced from the peripheries by other investors.

Quote:

…the space where the water would spread was given to an investor and he has already filled up about 35 acres with soil to displace the water; …government does not consider a poor person, it considers a rich person, even when a rich person destroys the wetland they (government) do not mind, but for us who are poor, when we plant our yams, they consider us very bad people who destroy the wetland (FGD Men farmers).

Lack of coordination

Lack of coordination among stakeholders was said to be a key institutional limitation hampering sustainable wetland management. Also, political interference was said to antagonise development control by the local authority, especially when wetland encroachers claim to have been permitted by higher authorities. However, according to KCCA, efforts are being made to actively engage lead-agencies and all the stakeholders.

Quote:

We are engaging lead-agencies, especially the National Environmental Management Authority (NEMA) to increase collaboration, coordination and decision-making with respect to wetland management. We are enforcing stoppage of further developments and denial of approval permits in wetlands; we do routine monitoring and inspection, and we engage parliament and cabinet who are the policy makers. This is important in controlling political interference (KI Supervisor Environmental Management KCCA).

The contextual drivers of encroachment on Nakivubo wetland presented above are however not exclusive of the underlying causes, which among others include poverty, population pressure, urbanisation, and capacity constraints that have been widely documented.

Discussion

In light of the conservation priorities for the Nakivubo wetland, which according to Namakambo (2000) are flood attenuation and wastewater treatment functions, insight into the dynamics of emergent vegetation was central in this study. Our results have shown that there was about 80 % loss and only 18 % recovery of wetland vegetation in 12 years (i.e. 2002–2014). The rate of encroachment on Nakivubo wetland, as measured by the loss of wetland vegetation, also accelerated between 2010 and 2014. As quantified in Table 4, large areas covered by wetland vegetation especially towards the lake (Fig. 5), were converted for instance to cultivated, grassland, and water. Another large form of conversion observed was from grassland to built-up and to cultivated areas. While transition towards cultivated areas is certainly human-induced, some of the inter-class transitions could be due to natural dynamics. For instance, the transitions towards grassland and water could be part of a natural process such that some patches which previously were cultivated, or where vegetation had been cleared gradually regrow into grassland. The dominance of grassland could also result from changes in soil conditions, for example, Miscanthidium grass tends to dominate where soils are acidic (Chapman et al. 2001; Whigham et al. 1993). Earlier studies had estimated about 14 % decrease in the total area covered by natural vegetation and a rapid increase of about 350 % in cultivated areas in the lower part of Nakivubo wetland between 1991 and 1996 (Kansiime and Nalubega 1999; Taylor 1991). Our findings not only agree with the high rate of loss of natural wetland vegetation reported by Kansiime and Nalubega (1999) and Taylor (1991) but also provide spatially congruent extents and site-specific conversions from wetland vegetation.

In our study, the explanations provided by FGDs and KIIs give insight into the dynamics of encroachment activities in the study context. The process seems to flow from clearing of the wetland vegetation and grassland, to draining for cultivation, and then where it is drier (especially the wetland peripheries), cultivated areas get gradually replaced by built-up areas and lawns. These areas then gain value faster due to their strategic location in the urban neighbourhood; settlements, commercial, and industrial establishments crop up. The farmers who are displaced from the peripheries and their counterparts seeking livelihoods from the wetland reclaim new areas, often further down into the wetland. Despite the slight decreasing trend in cultivated area (Fig. 2), much of the newly cultivated areas have replaced wetland vegetation, all the way down to the lake shore (Figs. 5 and 6).

Uganda’s regulations for wetlands, river banks, and lake shores require that a 200-m buffer zone of natural wetland vegetation be maintained for shore stability, pollution and flood control, fish breeding, and other ecosystem values (Nakangu and Bagyenda 2013; NEMA 2000). However, this is only one of many good environmental policies that barely get implemented due to competing uses which most often are short-term and consumptive. Agriculture, food security, livelihoods, and wetlands in Uganda are closely interlinked (Nakangu and Bagyenda 2013). Many of the crops that boost food security or generate income thrive best in moist soils. Analysis of the 2014 satellite imagery in this study shows development of a new road access to the lake via Nakivubo wetland, which will attract more human activities and further degradation. Our results support the notion that such short term, consumptive uses have taken precedence over the long-term benefits of conserving wetlands.

Additional to our findings, a visual inspection of Google Earth archive imagery from December 2013 to February 2015 (0.29° N, 32.64° E) clearly shows large portions of the wetland buffering the lake which are gradually drifting away into the lake. This is likely due to a loss of structural stability resulting from the increased cultivation. Calving away of wetland vegetation can occur naturally following sudden raise in water levels. Sudden raise in water levels can detach the roots of emergent vegetation from the substrate to form rafts of floating rhizomes. Much of the papyrus and Miscanthidium-dominated patches in the lower Nakivubo wetland are floating. During periods of rapid water level fluctuation and stormy weather, these rafts tend to break away from stable swamp together with fringe plants and form islands of floating vegetation (Whigham et al. 1993). The floating wetland vegetation on the lake-ward side of the Nakivubo wetland is frequently swayed by high speed-short duration south-east trade winds of up to 60 km/h for at most 2 min from May to July (Kansiime and Nalubega 1999). The diumal on and offshore winds lead to gradual displacement of surface water northwards and receding lake seiches drift the floating vegetation islands further into the lake. While these processes can occur naturally, cultivation in to the wetland vegetation buffering the lake weakens its ability to anchor.

In light of the proposed infrastructure developments to transform Nakivubo wetland into an urban park, in-land port, and lakefront (KCCA 2012, 2014), its future hangs in balance. The Ugandan Wetland Sector Strategic Plan 2001–2010 defines a critical wetland as one that is subject to ongoing degradation that jeopardises continuation of its attributes or existence (MWE 2001). Based on this definition, Nakivubo is a critical wetland that needs prompt monitoring, regulation of human activities so as to prevent further loss of the natural wetland vegetation and restoration of degraded areas.

To gain insight into what happened to the wetlands, why it happened, how the changes in the wetlands affect the people, and how the situation could be better managed or regulated in future, we adapt WHO’s ‘Driving force-Pressure-State-Exposure-Effect-Action’ (DPSEEA) framework (Briggs 1999; Hambling et al. 2011; Schirnding 2002). Based on this framework, we describe interlinkages between elements in the causal chain of effects and what the interventions will need to target. First the driving forces, such as population growth, urbanisation, and industrialization which result in pressures such as increased demand for environmental resources, food, space, and increased pollution streams. These pressures lead to the current state of the environment, which refers here to the transformation from the natural state such as the clearing of natural wetland vegetation, draining of wetlands, and altering of wetland attributes leading to loss of ecosystem services. The loss of ecosystem services such as flood attenuation and water purification leads to exposure to hazards, which in this context include floods and waterlogging, dampness, disease vectors, pathogens, and toxic substances (Isunju et al. 2015). Consequently, the people suffer the effects of the hazards which could range from damage to property, economic losses, high water treatment costs, ill health, and in extreme circumstances deaths. In order to manage the above situation, actions, or interventions should target each of the elements in the causal chain, including but not limited to wetland conservation and restoration, adaptation, hazard mitigation, and resilience building as well as policy interventions.

While the above measures have been recommended by earlier studies (Kansiime and Nalubega 1999; Kansiime et al. 2007; Lukooya et al. 2013), the big question of how to exploit opportunities as well as reduce risks still remains unanswered. It will require a multi-faceted approach to address aspects of equity, environmental integrity as well as economic development. Inadequate implementation capacity as reported by Ostrovskaya et al. (2013) calls for coordination of various stakeholders, and engagement of wetland communities as part of the solution (Kabumbuli and Kiwazi 2009). Implementation of wise use policies needs reflection local uniqueness and conservation priorities in each context. In the case of the Nakivubo wetland, it might require zoning out the wetland and actively engaging communities in wetland conservation and wise-use practices, as laid out in the wetlands, river banks, and lake shore regulations (NEMA 2000). A potential approach to consider here is a community conservation areas (CCA) approach, which is achieved through (1) raising awareness of the links between wetland biodiversity and livelihoods, (2) demonstrating and implementing wise-use practices, and (3) integrating community-based conservation models into policy and planning. A CCA approach has been piloted among rural wetland communities of the Lake Mburo-Nakivale and Lake Bisina-Opeta wetland systems in Uganda (Nakangu and Bagyenda 2013); however, its feasibility in an urban context such as Nakivubo needs to be studied.

Conclusion

Overall, our analysis showed a 62 % loss of wetland vegetation between 2002 and 2014, which is mostly attributed to crop cultivation. Cultivation in the buffer wetland vegetation makes it unstable to anchor, implying that it will likely be calved away by receding lake waves as evidenced by the 2014 data. With barely no wetland vegetation buffer around the lake, the heavily polluted wastewater streams will likely further deteriorate the quality of lake water. Furthermore, with increased human activities in the wetland, exposure to flooding and pollution will likely have more impact on the health and livelihoods of vulnerable communities. The findings also indicate that implementation of wise use policies is not only constrained by limited capacity but also by the fact that wetlands support livelihoods of many wetland-dependent communities. This calls for a multi-faceted approach, coordination of the various stakeholders and engagement of wetland-dependent communities as part of the solution. The wetland will need to be clearly zoned out and certain activities restricted to specific zones; for example, restricting agricultural activities to wetland peripheries and restoring the emergent wetland vegetation to control floods and pollution.