Introduction

The LULC changes are key aspects of global ecological processes (Berihun, et al., 2019; Shiferaw, et al., 2019) and have significant implications on ecosystem services value (ESV) (MEA, 2005; Powers et al., 2020). Millennium Ecosystem Assessment (MEA) (2005) report highlighted that the socio-economic and biophysical environment of the earth is rapidly changing as the result of socio-economic and biophysical drivers (Crespin & Simonett, 2016; Feng et al., 2012) such as rapid population explosion, agriculture expansion (Kindu et al., 2016), expansion of socio-economy (Makwinja et al., 2019), infrastructure developments, and climate change. The dynamic changes in the hydrologic cycle triggered by LULC dynamics and climate change have forced several aquatic species to migrate to alternative water bodies forcing other indigenous species to live in danger of being endangered, threatened, or even extinct. According to Haregeweyn et al. (2015), LULC dynamics significantly contribute to global atmosphere interactions, land degradation, forest fragmentation, and loss of biodiversity. IPBES (2018) report predicted that by 2050, Africa is expected to have a 2.5 billion population which will likely put severe pressure on the continent’s biodiversity and ES. Estimates also show that about 13 million hectares of global forest lands are annually converted to cropland (Arowolo & Deng, 2018; Mendoza-González et al., 2012; Robertson & Swinton, 2005). At a global level, Costanza et al. (2014) estimated that about US$4.3 to US$20.2 trillion of ESVs are lost annually due to LULC changes. Sutton et al. (2016) also estimated that about US$ 6.3 trillion of ESV is lost in impairing ecosystem function instigated by LULC. In El Salvador, Crespin and Simonett (2016) estimated that LULC dynamics caused a 12% loss of natural biomes leading to a 2.6% decrease of ESVs. Many ES valuation studies using LULC changes in sub-Saharan Africa where Malawi is located have also pointed out that the region is under severe pressure of ES degradation with significant consequences for local populations (Arowolo & Deng, 2018; Kindu et al., 2016; Landell-Mills & Porras, 2002; Leh et al., 2013; Shiferaw et al., 2019).

In Malawi, freshwater ecosystems such as Lake Malombe provide multiple ES to the local population. However, LULC dynamics have negatively affected the lake ESs and promoted ecosystem disservices (EDs). Soil erosion and land use degradation triggered by the poor vegetative cover and continuous cultivation is cited as the most serious problem (Palamuleni et al., 2010) that has led to severe watershed deterioration and disruption of ES in many water bodies. In Lake Malawi catchment, Chavula et al. (2011) estimated that savana/shrub/woodland has significantly declined almost by 90% from 1982 to 1995 due to LULC conversion. With the increasing human population, coupled with worsening poverty, marginal areas such as hilly slopes, rocky areas, marshes, lake periphery, originally left for natural vegetation to flourish are being converted into farmland and settlements (Malawi Government, 2013). Lake Malombe catchment is currently facing serious ecosystem degradation caused by both human activities and environmental drivers, leading to a decline in the ESs. Currently, loss and modification of natural habitats instigated by land-use activities are one of the leading threats to the lake’s aquatic biodiversity (Hara & Njaya, 2016). For example, recent studies indicate that the annual soil loss in Lake Malombe catchment is estimated at 29t/ha/year above the national level (Mzuza et al., 2019), cultivated land has increased by 18% (Palamuleni et al., 2010). The lake area has dramatically decreased (Dulanya et al., 2013) and is currently experiencing accelerated eutrophication and catastrophic decline in fish stocks (Jamu et al., 2011; Njaya, 2007). Likoya (2019) further acknowledged that the areas around Lake Malombe catchment have been experiencing some worst ecological disasters through floods and droughts as well as periodic lake level fluctuations. Population explosion evidenced in 1966–1987, 1987–2000, and 2008–2018 in the area has further increased pressure on the land and causing further disruption of the lake ES (Government of Malawi, 2018). Currently, there is visibly severe land degradation in form of soil erosion and deforestation.

Although many LULC studies have been conducted in the Lake Malombe catchment (Dulanya et al., 2013; Palamuleni et al., 2010), they mostly focused on the dynamics of LULC changes and associated drivers, while little attention has been paid to the impact of LULC dynamics on ESVs. Involving various stakeholders to formulate strategies on how to manage the lake ecosystem through an ecosystem-based management approach has been perceived as the solution to Lake Malombe ES degradation. However, information on the extent to which ESV has changed in response to LULC dynamics is required. Unfortunately, no such studies have been conducted in the Lake Malombe catchment to provide a comprehensive understanding of how LULC dynamics have impacted diverse ESVs. Thus, the objectives of this study are as follows: to quantify the LULC dynamics in Lake Malombe catchment from 1989 to 2019, to assign the specific coefficient of ESV to each land use category using the established unit-value transfer method, and determine the impact of LULC change on ESV. The coefficient of sensitivity is then assessed to estimate the uncertainty in the value coefficient. The significance of this study is to depict the current trend and vulnerability of each ES due to LULC dynamics to enhance policymakers and various stakeholders to take action and address the current lake ecosystem degradation.

Material and methods

Study area

The study was conducted in Lake Malombe catchment—geographical location between latitude 14° 21′ to 14° 45′ south and longitudes 35° 10′ to 35° 20′ east in the southern district of Mangochi (Fig. 1) (Makwinja et al., 2021). The catchment covers approximately 3387 km2 stretching from Liwonde National Park, in the southeast to the upper Shire River (FISH, 2015). It lies entirely within the Great African Rift Valley system and is characterized by a series of major and minor faults (FISH, 2015) and calcimorphic soils, which occur along the rift valley floor (Malawi Government, 2013). Typical calcimorphic soils include mopanosols, which are dark grey, sandy-clay soils with a low permeability; and alluvial soils which are grey to brown. The eastern part of the catchment has a low risk of soil erosion as it is partially protected by the forest reserve and Liwonde National park (Kapute, 2018). The western part is deforested and is prone to soil erosion (Malawi Government, 2013). Lake Malombe is fed by water from Lake Malawi via a 19 km stretch of the Upper Shire River and shares the same aquatic ecology with Lake Malawi, including a high level of fish biodiversity, genetic plasticity, and endemism (Kapute, 2018). Hydrologically, Lake Malombe is a shallow, turbid, and nutrient-rich lake with shelving vegetated shores as compared to Lake Malawi (Dulanya et al., 2014). Fish diversity, though, is relatively lower than in Lake Malawi, the biomass and productivity of the lake are higher probably due to nutrient-enriched water from inflowing streams from highly populated catchment areas and by recycling of nutrients in the sediments attributed to its shallow depth (Gondwe, 2009). The swampy shoreline of the lake is covered by Phragmites australis, Typha domingensis, Cyperus papyrus, Vossia cuspidate, Pennisetum purpureum, interspersed Eichhornia crassipes, scrubs, woodland, thicket, and scattered patches of mopane (FISH, 2015). The common crops grown in the surrounding flood plain include maize, groundnuts, rice, cotton, and tubers. The lake has approximately 47 species of fish belonging to families such as Anguillidae, Bagridae, Cichlidae, Claridae, Characidae, Cyprinidae, Mastacembelidae, Mochokidae, and Mormyridae which support the livelihood of the majority of the local population (Weyl et al., 2010). The lake has further diverse habitat types ranging from bottom substrate characterized by muddy, shallowest areas especially in the western and southeastern part of the shoreline, and dense beds characterized by oxygen weeds in the northeastern, and southwestern shoreline of the lake (FISH, 2015).

Fig. 1
figure 1

Lake Malombe catchment

Full size image

Satellite images pre-processing and land use classification

In this study, dry season Landsat images of LULC of 1989, 1999, 2009, and 2019 were produced from spectral Landsat imagery with a spatial resolution of 30 m retrieved from United States Geological Survey Website (USGS, http://glovis.usgs.gov/) using Worldwide Reference System Path 29 and Row 32. Before interpretation, atmospheric correction and geometrical rectification were performed including the resembling of a 1989 satellite image to match the pixel resolution. The criteria used to select Landsat imagery for the development of a land cover database for the Lake Malombe catchment was governed by existing multi-temporal images, vegetation phenology, and image class (cloudiness, haze), and the catchment boundary. All images were captured in August and September during the dry season in Malawi to have clear distinctive phenology and diverse land cover characteristics. The individual catchment differentiation shapefile was used to sub-set the individual Landsat images. The area of the shapefile was revised from the location base map of the study area.

Data processing

Land cover mapping and subsequent quantitative change detection require geometric registration between image scenes and radiometric rectification to adjust for differences in atmospheric conditions, viewing geometry, and responses (Lillesand et al., 2007), and sensor noise. Landsat images were processed using ENVI 5.1 geospatial software. A multiband combination of Landsat images was done in preparation for research spectral characteristics of various types of LULC. Geometric correction of remote sensing images was done using Malawi DeM, Universal Transverse Mercator Projection, Arc 1960, and UTM zone 36S, based on 1:50,000 topographical map scale to fit the Landsat images (Pettorelli et al., 2005) and eliminate errors of Landsat images which could probably be produced due to position of the sun and angle sensor.

Image enhancement

To check for applicability of the LULC classified map obtained after running the land classification tool in Arc Map 10.2, the location points of the current known LULC were compared with the classified map. Coordinates were captured for 30 different training points using a Garmin Etrex Venture HC, GPS. Observations were made on the current LULC for each point. With a false-color composite, band combination of 4, 3, and 2 for Landsat 5, 7, and 8, various features in the imagery were identified. In this standard false composite, the vegetation appears shades of green, water blue, soil, and other dark grey. This TM band combination gives results similar to traditional color infrared aerial photography and highlights vegetation in green color thereby making them easy to visualize.

Normalized difference vegetation index

The normalized difference vegetation index (NDVI) was used to assess the presence of green vegetation and was computed as follows:

$$NDVI=\left(\frac{NIR-RED}{NIR+RED}\right)$$
(1)
$$NIR=Near-infrared,\;RED=Red\; band$$

NDVI values ranged from −1 to 1 means the higher the NDVI value, the higher the fraction of green vegetation present in the area. Landsat band 4 (0.8–9.94 mm) measured the reflectance in the NIR region and band 3 (0.63–0.69 mm) measured the reflectance in the red region. However, for Landsat 8, the NIR and red regions had different wavelength ranges. Therefore, they were computed using bands 4 and 5 for red and NIR respectively which resulted in values ranging from 0 to 200 and fit within an 8-bit structure. The green color showed the presence of vegetation and other colors show the absence of green vegetation. These attributes were useful in classifying the images.

Land use/land cover change image classification

The land in the study area was classified based on physiographical knowledge of the study area, visual interpretation using historical Google Earth images, supporting supplementary data, and the researcher’s prior local knowledge of the study area. The study area was classified into six characteristics (Table 1) using the spectral signature. Stratified random sampling techniques were employed to collect 320 points for accuracy assessment. Land classes were verified by field observation in situ. Reference data was retrieved from Google Earth. Kappa coefficient and user accuracy derived from the error matrix were used in determining the accuracy assessment of the classification. The 2019 LULC imagery was subjected to a minimum of 80% overall accuracy and was used as a reference to classify historical images. The 2019 imagery was captured to compare phenological data during the study period and historical images (1989, 1999, 2009, and 2019) were further visually interpreted taking into account image tone, texture, shape, and class patterns. The LULC change analysis was further determined using the post-classification comparison (PCC) technique and was reported in a cross-tabulation matrix. Arc GIS software version 10.7.1 was used to compute LULC change-traditional matrix using overly procedure to quantify the area converted from a particular LULC class to another LULC category during the study period. The annual rate of change was determined as shown in the following equation:

$$\boldsymbol{r}=\left(\frac{1}{{\boldsymbol{t}}^{2-}{\boldsymbol{t}}^{1}}\right)\mathrm{*}\;\mathbf{ln}\left(\frac{{\boldsymbol{S}}_{2}}{{\boldsymbol{s}}_{1}}\right)$$
(2)

where r is the annual rate of change for each class, \({S}_{1}\) and \({S}_{2}\) are areas of each LULC class at a time \({t}^{1}\) and \({t}^{2}\), respectively.

Table 1 Description of land use/land cover categories
Full size table

Calculation of Lake Malombe ESV as a consequence of LULC dynamics

The benefit transfer approach (BTA) was used to estimate the values of ES of different LULC types and their dynamics (Niquisse et al., 2017). Since literature focuses on Lake Malombe’s ESV is scarcely available, existing values and data from the study sites of similar locations were used (Kindu et al., 2016). Based on the global ESV coefficients proposed by Costanza et al. (1997), the equivalent value per unit area of ES for Lake Malombe was extracted. Using Costanza et al.'s indexes as a foundation, ESV per unit area for each land use attribute was assigned based on the equivalent coefficient value of the ES (Table 2). The ESV of Lake Malombe catchment for each of the periods (1989, 1999, 2009, and 2019) was determined using the models described by Costanza et al. and recommended by Li et al. (2010) and Kreuter et al. (2001).

$${ESV}_{k}=\sum _{f}{A}_{k}{VC}_{kf}$$
(3)
$${ESV}_{f}=\sum _{k}{A}_{k}{VC}_{kf}$$
(4)
$$ESV=\sum _{k}\sum _{f}{A}_{k}{VC}_{kf}$$
(5)

where \({ESV}_{k}\),\({ESV}_{f}\), and \(ESV\) represent the ESV of land use type k, service function f, and the total ESVs, respectively (in the unit of US$). The \({A}_{k}\) is the area of land use type k (ha) and \({VC}_{kf}\) is the value coefficient for land use type k with ecological service function type f. The sensitivity analysis (CS) was also performed to verify the changes in the ESV for a given change in the coefficient. Using the standard economic theory of elasticity as an indicator of how well an economic variable responds to another economic variable, the CS was calculated as:

$$CS=\frac{{ESV}_{j}-{ESV}_{i}}{{VC}_{jk}-{VC}_{ik}/{VC}_{ik}}$$
(6)

where ESV refers to the estimated ESVs, VC refers to the value coefficient, i and j represent the initial and adjusted values, and k is the land type.

Table 2 Global Ecosystem service values (ESVf) per unit area for different land use categories (US$/ha/year (Costanza et al., 1997)
Full size table

Exploratory survey and assessment of perceived drivers

To identify the perceived drivers of LULC dynamics, eight focus group discussions (FDG), comprising an average of fifteen members per group were conducted. The FDG was coordinated by a single facilitator to maintain consistency in question delivery, timing emphasis, and reactivity (Maxwell, 2013). The FGDs consisted of young and elderly of local communities in the Traditional Authority Chimwala, Chowe and Mponda, group village headman Changamile, Changali, Mpinganjira, Mtambo, Chowe, Nalikolo, Kadewere, Dimu, Chapola, with a balanced composition in terms of gender. Questions focusing on LULC dynamics that have occurred over the past years and reasons why such dynamics occurred were presented to each group. The drivers identified by the communities were ranked based on their degree of importance using a Likert scale from 1 to 5 (1 = least important and 5 means most important). Six key informant interviews targeting chiefs, government officials, NGOs, leaders of environmental conservation groups, village beach committees, village elders, and fish traders were also conducted. The key informant interviews were conducted using a phenomenological research approach (Patton, 1990) to triangulate different data sources to improve validity. The open-ended questions were used to allow the local communities to discuss and elaborate more on their experiences (Maxwell, 2013). Data collection procedures followed high ethical principles recommended by Levy and Lemeshow (1999). The FGDs and the key informant interviews were conducted between October and November 2020.

Statistical analysis

Geospatial LULC Landsat satellite data were analyzed using GIS and remote sensing approaches. ArcGIS version 10.7.1 and ENVil 5.4 software were used. Some qualitative and quantitative data were analyzed using descriptive and inferential statistics. STATA version 15 was used in the data analysis.

Results

Table 3 presents the results of the confusion matrix for accuracy assessment of LULC classification of the years 1989, 1999, 2009, and 2019. With Kappa coefficiency of 87% in both 1989 and 2019, the confusion matrix produced overall very good accuracies for the six classification points and the defined LULC classes. Both class-specific user accuracies and producer’s accuracies were at least 80%, suggesting that a considerable percentage of pixels were correctly classified. The LULC dynamics from 1989 to 2019 derived from image classification are presented in Fig. 2. The changes in the size of six land use categories are presented in Table 4. Comparing the land use of 1989, 1999, 2009, and 2019, there was an obvious change in LULC. Built-up, shrubs/grass/bare land, and cultivated land progressively increased over the study period. The built-up area increased by 11.5% in the period (1989–1999), 27.6% in the period (1999–2009), and 30.7% in the period (2009–2019). The overall increment of the built-up area in the study period was 149%. Cultivated land increased from 52,932 ha to 62,852 ha in the period (1989–1999) translating to an 18.7% increase, slightly increased by 6.96% in the period (1999–2009) and then 17.48% in the period (2009–2019). A similar trend is observed in shrubs/grass/bare land with a percentage increase of 11.5% in the period (1989–1999) and a sharp increase of 27.6% in the period (1999–2009) and then a very slight increase of 3.07% in the period (2009–2019). Overall, the increment of the cultivated area was 49.2% and shrubs/grass/bare land 46.72% during the study period. The total area of forest land was reduced from 56235 to 8585 ha during the study period translating to an 87.3% decrease and the value was the highest of its negative dynamic index among all land use classes suggesting an increased rate of deforestation in the lake catchment. Other land-use classes such as water bodies and marshes showed mixed results. Waterbody decreased by 5.99% period (1989–1999), slightly increased by 1.93% period (1999–2009), and then dropped by 4.47% period (2009–2019). Overall, the water body showed a negative trend of 8.46% throughout the study period. Marshes increased by 42.8% in the period (1989–1999), significantly decreased by 48.27% in the period (1999–2009), and then fluctuated positively by 36.96% in the period (2009–2019). Overall, the area covered by marshes slightly increased by 1.18% throughout the study period. The major perceived drivers for LULC dynamics were rapid population growth, climate change impact, poverty, government policy conflicts, poor governance, poor education, food insecurity, and limited alternative livelihood (Fig. 3). Limited alternative livelihood was ranked the highest followed by poverty, population growth, food insecurity then poor education. Climate change, poor governance, and other associated drivers were ranked the lowest. The fluctuation of water bodies was strongly linked to the impact of climate change (Table 5) defined within the context of high temperatures and declining rainfall patterns in the catchment. The decrease in forestry land was linked to increased rapid population growth, government policy conflicts, and poverty. An increase in the built-up area was also highly linked to an increase in human population suggesting that as the population increases more land for settlements is required at the expense of forest land.

Table 3 Accuracy assessment of classified image
Full size table
Fig. 2
figure 2

Land use/land cover classes

Full size image
Table 4 Absolute area coverage
Full size table
Fig. 3
figure 3

Drivers for land use/land cover changes

Full size image
Table 5 Lake Malombe LULC dynamic drivers’ correlation matrix
Full size table

Lake Malombe ecosystem services

Considering ESV changes across the different time intervals (1989–1999, 1999–2009, 2009–2019), ESV for forest land displayed a negative trend whereas the ESV for shrubs/grass/bare land increased (Table 6). For marshes and water bodies, the ESVs were mixed up. The ESV for marshes increased in 1999–2009 interval, decreased in 1999–2009, and then increased in 2009–2019. Waterbody’s ESV on the other hand increased in 1999–2009 and then decreased between the 2009–2019 interval. For cultivated land, shrubs, bare land, and grassland, the ESVs have been increasing across different intervals. The net ESVs between 1989 and 1999 increased by US$20.77 million and decreased by US$ 79.83million between 1999–2009 and slightly increased by US$13.48 million. The increase in net value between 1999–2009 time intervals was due to an increase in marshes as the lake level declined. On the other hand, the dramatic reduction in ESV observed between 1999–2009, was due to the high rate of conversion of marshland into cultivated as well as a fishing ground. It was difficult to estimate the ESV value for build-up areas, though the rural houses played an integral part of the lake landscape matrix and had an impact on ES provisioning. Table 7 presents the estimates of sixteen ES categorized as provisioning, regulating, supporting, recreation, and culture. In terms of the estimated ESV for provisioning services, water supply registered a significant net loss of US$28.5 million between 1999–2009 and a slight gain of ESV values of US$14.23 million and US$8.75 million between 1989–1999 and 2009–2019. Overall, there is a general net loss of water supply ESV of about US$ 5.5 million during the study period. The ESV net gain for food production slightly increased by US$0.16 million throughout the study period. The ESVs for raw materials and genetic resources have been declining throughout the study period with a total net loss of US$6.56 million and US$ 0.76million. The regulatory services such as gas regulation and biological control experienced a slight net gain of US$0.12million and US$0.85million. Climate regulation had a net loss of US$6.72 million, disturbance regulation (US$0.3 million), hydrological regulation (US$15.46 million), erosion control (US$4.18 million), and waste treatment (US$4.37 million). Supporting services such as soil formalization had a net gain of US$0.03 million, habitat (US$ 1.31 million), and pollination (US$0.71 million). The ESV values for recreation and culture services were reduced with a net loss of US$3.73 million for recreation services and US$4.26 million for culture. Table 8 shows the coefficient of total ESVs and the coefficient of sensitivity (CS). The CS is defined as the ratio of the percentage change in the estimated total ESVs and the percentage change in the adjusted valuation coefficient. In this study, the CS in all analyzed LULC classes were less than one with the highest (0.708) reported in the waterbody class after adjusting the value coefficient for these land use categories by 50%. The CS for the water body was the highest because of the highest ESVs coefficient. The CS greater than 1 means that ESV is elastic to the coefficient while less than 1 means that ESV is inelastic. The greater the proportional change in the ESVs relative to the proportional change in the valuation coefficient, the more critical it becomes to use the ESV coefficient. The results from this study indicate that the total ESVs estimated in the study area were relatively inelastic to the ecosystem value coefficients, suggesting that the ESVs used in this study were practical and robust.

Table 6 Change in Lake Malombe ESV (million US$)
Full size table
Table 7 Values of Lake Malombe ecosystem service functions (million US$)
Full size table
Table 8 Percentage changes in estimated ESV and CS resulting from a 50% adjustment in VC
Full size table

Discussion

Analysis of LULC dynamics

The LULC dynamics for the Lake Malombe catchment from 1989 to 2019 were computed. The analysis showed a massive expansion of built-up areas and cultivated land. The forest cover decreased during the period (1989–2019). This is in line with many studies in Malawi (Chavula et al., 2011; Dulanya et al., 2013; Gondwe et al., 2019; Haak et al., 2015; Jamu et al., 2003; Mawenda et al., 2020; Munthali & Murayama, 2015; Pullanikkatil et al., 2016) and other countries around the global (Bone et al., 2017; Duraisamy et al., 2018; Hailu et al., 2020; Kabanda, 2019; Käyhkö et al., 2011; Kilic et al., 2006; Kindu et al., 2016; Landell-Mills & Porras, 2002; Mtui et al., 2017; Mucova et al., 2018; Phiri et al., 2019; Ritse et al., 2020; Teucher et al., 2020). The major perceived drivers for LULC dynamics were rapid population growth, climate change impact, poverty, government policy conflicts, poor governance, poor education, food insecurity, and limited alternative livelihood. Kindu et al. (2015) also identified similar drivers as the main cause of LULC dynamics in the Munessa-Shashemene landscape of the south-central highlands of Ethiopia. Limited alternative livelihood option was ranked the highest because most of the vulnerable local population in Lake Malombe catchment live at the margins of the cash economy and have very limited choices for generating an income (Bulirani, 2003). Without access to land and marketing, ES generally offers the greatest option for sustenance (Stringer et al., 2020). Consequently, these vulnerable individuals who live close to the lake ecosystems such as fishers, hunters, and small-scale farmers are increasingly pushing forest land and the lake biodiversity to the verge of extinction as the result of deforestation, agriculture expansion, and overexploitation (Woodruff, 2010). Behind direct threats to the lake ES, complex factors such as high rates of illiteracy, poor governance, social-economic, and rapid population growth interact (Hall et al., 2017).

Mangochi district (where Lake Malombe is located) is ranked as one of the districts in Malawi with high illiteracy rated at 65.9% (NSO, 2012). Increased unregulated sand mining, illegal charcoal production, brick molding, and kilning that lead to massive deforestation, reduction in aesthetic beauty, and high rate of erosion are adopted by the poor with a low level of education (Government of Malawi, 2014). The main challenge of the high illiteracy rate is that the local communities are held back in terms of accessing knowledge on the value of ES (FISH, 2015). For example, continuous over-exploitation of biodiversity in the Lake Malombe catchment is attributed to the fact that local communities lack knowledge on how to embrace an effective governance system to achieve sustainable ES management (Makwinja et al., 2019). There is further limited knowledge on the development of sustainable adaptation strategies as well as alternative livelihood options. This consequently leads to unsustainable exploitation of the scarce ES. The unsustainable exploitation of the ES leads to ecosystem disservices. Ecosystem disservices often breed conflict over the limited resources. Conflict leads to resource scarcity through habitat destruction, over-exploitation of natural resources, pollution, and breakdown of environmental programs (Nagoli et al., 2016). Lake Malombe provides the best example. The onshore and offshore aquatic vegetation has dramatically declined due to the high level of deforestation instigated by the wide expansion of agriculture, high demand for firewood and charcoal production. Fishery resources have been overexploited due to human pressure putting human-wellbeing at risk (FISH, 2015; Malawi Government, 2010). There is a general piece of evidence that large four endemic Oreochromis species-Oreochromis shiranus, Oreochromis karongae, Oreochromis squamipinnis, and Oreochromis lidoles have almost vanished from the lake due to among other reasons being recruitment overfishing, destruction of habitats, increasing silt-load, and disappearance of weed areas in the Lake (FISH, 2015; Hara & Njaya, 2016). This is also a case in many African Great Rift Valley lakes (Ogutu-Ohwayo et al., 1997) such as Lake Victoria (Kolding et al., 2008; Matsuishi et al., 2006), Lake Malawi (Bootsma & Hecky, 1993), Lake Chilwa (Jamu et al., 2003), Chia lagoon (Makwinja et al., 2019), and other lakes in Africa (Kafumbata et al., 2014).

Poverty is one of the major drivers for the unsustainable exploitation of ES. Malawi is a small but densely populated country in sub-Saharan Africa with 62% of the population live on less than US$ 1.25 a day and 89% lives below the US$ 2 a day threshold (FISH, 2015; NSO, 2005). The majority of the poor are heavily reliant on the low-input production of maize on small landholdings where the soil is degrading (Sanchez, 2002). Agriculture is the major source of income for rural households (63.7%) but land productivity is generally low. Most farmers rely on rain-fed farming or limited forms of irrigation (Nagoli et al., 2017; Pullanikkatil et al., 2015). In Lake Malombe catchment, increased the rate of poverty is evidenced by an increased percentage of households experiencing food shortages annually (Fig. 4a) which is instigated by increased drought cases, crop damages due to pest and diseases, lack of inputs such as fertilizers, excessive rain (Fig. 4b), low level of education, high rate of youth unemployment, limited livelihood options, land scarcity for farming and social conflicts (FISH, 2015; Hartzell & Hoddie, 2003). Lake Malombe catchment is surrounded by hills in the west bank, 3–7 km away from the lake, bordered by the Mangochi hills on the eastern part and the Liwonde National park in the southeast banks trapping the fishing communities within the narrow strips of land along the lake at a high population density of 500/km2 (Bell & Donda, 1993). Charcoal burning and fishing are the main sources of livelihood while farming is done in the steep slopes, rocky areas, flood plain, wetlands, and around the lake periphery. Poverty is forcing the fishing communities to remove trees for cultivation and income generation (Government of Malawi, 2014). It further forces them to seek different unsustainable coping strategies that have serious implications for ES (Fig. 4c). The endangered tree species such as Percopsis angolensis, Terminalia sericea, Colophospermum, Pterocarpus angolensis in the forest reserve are being cleared out for charcoal production (plate 1). The decline in marshes in the period of 1999–2009 is linked to the shifting of riparian communities’ coping strategies from fishing to farming systems to avert the shock (FISH, 2015; Salas et al., 2010). Many fishing households diversified their livelihood through cultivating upland areas and steep slopes a few meters away from the lake. Flood plain cultivation and small-scale irrigation are also common during the lake recession leading to a high rate of conversion of marshland into cultivated.` The situation reported in Lake Malombe also applies to other Malawian inland lakes such as Lake Chilwa, Lake Chiuta, and Chia lagoon (Jamu et al., 2003; Makwinja et al., 2019; Mzembe, 1990; Nagoli et al., 2016; Palamuleni et al., 2010).

Fig. 4
figure 4

The rate of food shortages, the reasons, and coping strategies

Full size image

The government policy conflict is also one of the major drivers for LULC dynamics. For example, the National Forest Policy of 1996, considers the forest as life-sustaining as it provides watershed protection and enhancing water resources (Malawi government, 1996). However, this contradicts the food security policy of 2006, the land policy of 2002, and agriculture policy which promote a market-oriented approach as a response to the national economic development agenda. The later policies do not explicitly explain how economic development and ES should be balanced in the face of unprecedented population growth and economic activities. Agriculture policy priority also contradicts other related policies such as fisheries, land policy, environmental policy, and water policy as it focused much on transformative utilization of land for the farming system to increase agricultural production (Government of Malawi, 2016). The increasing demand for agricultural areas affects a wide range of ES, including forest, water quality, soil retention, carbon sequestration, biodiversity conservation, and nutrient cycling (Dale & Polasky, 2007). Agriculture policy also does not explicitly explain how to balance between these two. Malawi's National Water Policy of 2005 prohibits river bank cultivation and encourage buffers along river banks as one way of protecting marsh/mangroves, while agriculture and irrigation policies on the other hand are framed to promote small scale irrigation, river bank cultivation, thereby increasing the conversion of marshland into cultivated—a scenario which is depicted in 1999–2009 where 48.27% of the marshland in Lake Malombe catchment were lost. In developing countries such as Malawi, irrigation is seen principally as a tool for social-economic development, poverty alleviation, and enhancing food security (Sanmuganathan, 2000). Malawi’s irrigation policy of 2016 emphasizes the need to effectively use the freshwater ecosystem for irrigation development to serve human needs such as food, nutrition, and income (Government of Malawi, 2016). Despite the significant contribution of irrigated agriculture to increasing food production and overall socio-economic development, irrigation has come under increasing criticism over the past decade for concerns over increasing environmental problems such as siltation, reduction in vegetative land, decrease in water bodies, and water quality degradation (Alauddin & Quiggin, 2008). In Pakistan, evidence shows that increased irrigated agriculture is associated with increased salinity and depletion of water resources in lakes, rivers, and reservoirs (Qureshi et al., 2008). Malawi’s irrigation policy does not explain how these problems should be addressed to achieve a sustainable freshwater ecosystem. The Malawi customary land laws also offer legal rights on the land if the household converts forest land or buffer zone into farmland. This encourages farmers to legally or illegally covert forest land and buffer zone into cultivated (Government of Malawi, 2002). This policy weakness has also been identified by many researchers in various tropical countries (Green et al., 2013; Muleta et al., 2020; Shiferaw et al., 2019; Solomon et al., 2019; Tendaupenyu et al., 2017; Tolessa et al., 2017).

Climate change is also ranked as a major driver for LULC dynamics in the Lake Malombe catchment. Malawi has experienced severe droughts for the past years notable among these occurred in 1948/49, 1991/92, 1996/97, and 2001/02 seasons (Chavula, 1999). In Malawi, the observed climate changes are evidenced by the changes in the rainfall season, pattern and temperature, and changes in the frequency of droughts and floods (Mkanda, 1999) as well as significant variations in the lake and river levels (Calder, 1995). The historical data (Fig. 5) shows that there have been increasing in maximum temperatures over the past years while total annual rainfall has been declining with negative consequences on Lake Malombe water levels (Likoya, 2019; Ngongondo et al., 2020). Waterbody fluctuated negatively in 1989–1999 and then slightly fluctuated positively between 1999 and 2009 and later declined between 2009 and 2019 (plate 2). This fluctuation is linked to climatic forces such as El Niño/Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) as previously confirmed by Dulanya et al. (2013). Malawi government report states that the Shire River Basin where Lake Malombe is located has been facing significant extreme weather events with severe drought conditions in 1991–1992 (the same period which corresponds to the decline in water bodies) (Malawi Government, 2013). The report suggests that prolonged droughts that have been experienced in Lake Malombe catchment over the past years are responsible for a decline in water levels, complete drying up of rivers (plate 2) and marshes, decline in fish production, land degradation, loss of soil fertility, loss of forest through the fire, losses in biodiversity, and migration of mammals and other animal species (Malawi Government, 2013). Lake Malombe catchment is further experienced extreme flash floods in the 2000–2001—the period which corresponds to the increase in water bodies. The extreme flash floods led to the loss of agricultural land, low crop production with serious implications on ES and local communities. The marshland also increased in the subsequent years. The increase in marshland was attributed to increased lake siltation and low lake levels that exposed lake beds to sunlight and encouraged the growth of herbaceous grass (Keim et al., 2012).

Fig. 5
figure 5

Rainfall and temperatures of Lake Malombe catchment

Full size image

Lake Malombe ecosystem services

The sensitivity analysis undertaken to test the robustness of our estimation showed that our ESV was reliable which confirms the other previous findings such as those of Tolessa et al. (2017), Leh et al. (2013), Li et al. (2007), Hao et al. (2012), Kindu et al. (2016), and Gashaw et al. (2018). The LULC dynamics were described as the most important driver of change in Lake Malombe ES. The overall ES of the study area experienced a net loss of US$45.58 million during the study period. The value ES loss due to LULC dynamics is greater than the estimated total annual ES benefit generated from Lake Chilwa which is estimated at US$21 million (Malawi Government, 2010), Lake Chuta estimated at US$17 million (Zuze, 2013), and less than the economic benefit generated from Lake Chad (US$750 million) (Kafumbata et al., 2014) and Lake Naivasha (US$613–640 million) (Mekonnen & Hoekstra, 2010). On the other hand, Msofe et al. (2020) reported a total ESVs net loss of US$811.5 million in Kilombero Valley Floodplain, Southeastern Tanzania, over 26-year period—a value which is higher than the one reported in this study. On the other hand, Kindu et al. (2015) reported slightly lower net loss ranging from S$ 19.3 million to US$ 45.9 million in Munessa–Shashemene landscape of Ethiopia than the current net loss value reported in Lake Malombe. The highest ES net loss value of US$79.832 million was recorded from 1999 to 2019 due to increased loss of forest and marsh areas, which is in line with other research findings in other countries around the globe (Badamfiroo & Mousazadeh, 2019; Hasan et al., 2020; Hu et al., 2008; Li et al., 2007; Muleta et al., 2020; Tolessa et al., 2017; Wang et al., 2006; Zhang et al., 2011). When disaggregated based on ES functions, the provisioning services had a net gain of US$13.16 million in the period of 1989 to 1999 and dramatically experienced ES net loss of US$33.74million and then a slight net gain of US$7.92 million in the period of 2009–2019. Cumulatively, the ES provisioning services from 1989 to 2019 had a net loss of US$ 12.68 million suggesting that the lake ecosystem is degrading and is losing its economic value. The regulatory ecosystem services have been fluctuating since 1989 with the highest net loss of US$36.32 million recorded between 1999 and 2009. The cumulative net loss of ESV for regulations from 1989 to 2019 was US$29.73 million. In terms of supporting services, Lake Malombe catchment had a cumulative net gain of US$1.01million though with a net loss of US$ 1.51 million in the period of 1999 to 2009. Culture and recreational value also cumulatively declined by US$7.99 million during the study period and this could be attributed to the loss of biodiversity and degradation of water quality in the lake. Our results are in line with other research findings in the tropical region (Cawthorn & Hoffman, 2015; Kindu et al., 2016; Li et al., 2010; MEA, 2005; Wangai et al., 2016). Figure 3 b on the other hand shows that the total fish biomass has also been dramatically declining from 13,000 metric tons registered in the early 1980s to approximately 4575 metric tons as of 2019 (Fig. 6a) with the corresponding decline in economic benefit (Fig. 6b) from US$14.32 million recorded in 1982 to about US$5.69 million as of 2019 translating to 60.3% net loss. The value of net economic loss from Lake Malombe fishery is less than the total annual economic benefit generated from Lake Chilwa fishery estimated at US$ 18 million (Malawi Government, 2010) but higher than (US$4.7million) reported in Lake Chiuta (Zuze, 2013).

Fig. 6
figure 6

Total fish biomass and total revenue within the period from 1976 to 2020

Full size image

Conclusion

The ESV estimation based on LULC analysis is very vital within the landscape and ES. It demonstrates how much ES changed in responses to anthropological activities and ecological stress on spatial and temporal scales. Such estimation is of paramount importance in influencing decision-making processes through modifying national accounting systems to reflect the true value of ES that will ultimately be used as a basis for substantial and sustainable developmental activities within the catchment. We found that forest land was significantly reduced to other land uses by 84.73% during the period while other land classes such as built-up areas and cultivated increased. The overall ESV decreased by US$45.58 million during the study period. The highest ES net loss value of US$79.832 million was recorded from 1999 to 2019 due to increased loss of forest and marsh areas. Within the same period, the lake fishery also recorded a net loss of US$8.63million. The CS for water body was the highest among all land uses throughout the study period indicating a higher ESV coefficient assigned to the water body. This study suggests that the only way to avoid ecosystem disservices in the catchments of tropical shallow lakes such as Lake Malombe is by reducing the impact of anthropogenic and ecological stressors within acceptable levels to maintain ecosystem functions. This can be achieved by firstly understating the strategies on how the local population around the lake have survived over the years from the shocks and stress triggered by the loss of ESVs and develop an incentive economic model for the lake ES based on the data generated from the study of this nature to support local population livelihoods which are currently at the greatest risk due to increased LULC dynamics and other associated stressors such as poverty and environmental threats. The economic incentive models should include the introduction of integrated agriculture–aquaculture in the upland areas to reduce pressure on the capture fishery, improve the income generation through income diversification and agroforestry, introduction of cage culture, fish restocking program, fish habitat, and sanctuaries to improve income generation, promoting village savings, climate-smart agriculture, and freshwater ecosystem marketing approach to reduce poverty, improve food security and livelihood of the local population. However, the success of this approach depends on an effective multi-sectoral collaborative effort at a grass root level from various partners to convince the local population who are predominately fishers and small-scale farmers the economic incentives of adopting these approaches as compared to capture fishery and the traditional farming systems. Joint efforts should also be made to address the conflicts that exist among the various government policies. There is an urgent need to harmonize various multi-sectoral conflicting policies and projects to ensure that there are synergies among them. There is also a need to enforce the buffer zones of about a 10 m radius around the lakeshore and riverine corridors as stipulated in the water resources act through a co-management approach to provide greater protection of the mangroves as well as marshes. Capacity-building among local governance structures especially on the integrated ecosystem-based management approach to ensure that they understand the value of the lake ES and a need to conserve for the future generation is also required. The combined initiatives from both local and national could promote conservation of forest as well as marshland. The current study findings show that Lake Malombe catchment is at a high rate of ES loss due to several underlying and proximate causes of LULC dynamics. Of the LULC within the Lake Malombe landscape, our results indicate that marshland, forest, and water bodies provide a higher level of ES than others due to the relatively larger coefficient assigned to them as well as larger area in case of forest and hence a reduction in these land classes significantly hamper the balanced flow of ES.

Limitations

We noted that coefficient ESV developed by Costanza et al. (1997), however, had some limitations. For example, the unit values derived from one area of a specific good were based on a generalization of average unit value in all areas with an assumption that ESV within the entire LULC categories was homogeneity (Kindu et al., 2016) which may not be true in reality. Despite this limitation, the Costanza et al. coefficients have been widely used by several environmental scientists, ecosystem managers, and policy analysts for timely assessment of various ES at multiple geographical scales (local, national, regional, and global) when primary data is scarcely available (Wong et al., 2015). Another limitation was that studies of ESV are very scarce in Malawi and our context was based on ESV global indexes after perfectly and accurately adjusted them to reflect local ecosystem conditions (Song & Deng, 2017). Given the location of Lake Malombe which falls within East African Great Rift Valley systems, the ESV estimation using LULC and established coefficient is extremely important where ground data collection is expensive, land degradation is pervasive and the historical land uses data for rural areas is scarce. It provides the best alternative and robust data for decision-making processes at the local level and a similar study can be applied in other parts of the country’s freshwater ecosystem.