Potential of GIS and remote sensing in mapping land degradation: catchment of the Manyame River, Zimbabwe

Natural and human activities significantly contribute to the deterioration of land across the globe and this has become a major concern as food production largely depends on sustainable land husbandry. The Manyame River Catchment, Zimbabwe, is undergoing enormous land degradation due to legal and illegal practices. Land degradation in the Manyame River Catchment is, therefore, negatively impacting its existing ecosystems’ productivity. The majority of the drier regions across the globe (41%) are important for over a third of the world's populace (Yu et al. 2017; Li et al. 2021). However, some human settlement activities are threatening healthy ecosystems with barren conditions (Barbosa et al. 2015; Berdimbetov et al. 2021). Undoubtedly, some of the existing terrestrial ecosystems in arid regions have been subjected to land degradation processes thereby resulting in reduced productivity (Masoudi et al. 2018). Unproductive land use practices have become a global challenge. This is affecting the population of Sub-Saharan Africa, where people live on unproductive land (Wessels et al. 2007; Gao & Liu 2010; Testa et al. 2014; Gu et al. 2022), as in the case of Zimbabwe.

The ravaging effects of climate change due to unsustainable human settlement activities are fuelling land degradation. Land degradation is a common problem in Zimbabwe, especially in the communal areas (Sharara et al. 2022), which includes the Manyame River Catchment community. The studies that have been carried out worldwide reveal the impacts of the current land use processes on the livelihoods of, mainly, rural communities whose livelihoods mainly depend on land resources (Wessels et al. 2007; Wardlow & Egbert 2010; Higginbottom & Symeonakis 2014; Wilson 2014; Abdel-Kader 2019).

The Manyame River Catchment Area, in North-Eastern Zimbabwe, is experiencing widespread land degradation (Kibena et al. 2014). The fragility of terrestrial ecosystems of the semi-arid areas including those of the Manyame River Catchment area helps in exposing Africa to the driving forces of climate variations (e.g., Huang & Kong 2016). Low and intermittent annual total rainfall variability, a characteristic feature of a part of the study area (Masocha et al. 2017), impacts the growth of the area's vegetation. The vegetation for this study area, especially the study area's semi-arid zone, requires large volumes of rainfall for a significant vegetative growth to take place. The interaction between climatic factors and human actions alters the status of terrestrial ecosystems, leading to land degradation (Abdel-Kader 2019; Teja et al. 2021).

Effective implementation of conservative, preventive, or remediation policies on environmental challenges should be the results of scientific research works on an environmental problem (AbdelRahman et al. 2018; Gbagir et al. 2022). Mapping the distribution of land degradation in an ecosystem can provide evidence that the problem really exists (Higginbottom & Symeonakis 2014; Hereher & El-Kenawy 2022). This implies that the mapped degraded zones become the target areas for monitoring and implementation of managerial strategies. Traditionally, environmental monitoring and assessment have been and are still being assessed by expensive field-based approaches.

Although field-based approaches are accurate, their demands are costly and they also lack a continuous spatial–temporal variation assessment of the environment (Katsurada et al. 2007; Droj 2013; Andreas 2020; Gedefaw et al. 2021). Thus, the invention of technological approaches such as the use of GIS and remote sensing have been useful in covering a larger areal extent as well as accessing very remote zone of the environment (Zhang & Kondragunta 2006; Griffith 2015; Abdel-Kader 2019). Current technological approaches (e.g., GIS and remote sensing) also make it easier for natural resource managers to assess, monitor, and then implement strategic approaches to the most critically disturbed zones within a very short period, and at less mobility cost. Knowing the degradation status and its possible causes through the use of the latest technologies is essential for the development of appropriate mitigation measures for the proper utilisation of land resources (Beardmore et al. 2019). Technology in GIS and remote sensing has been instrumental in mapping land degradation across the globe. Remote sensing provides a competitive advantage in monitoring land degradation and its various spatio-temporal resolutions from which researchers can maximise (AbdelRahman et al. 2018; Li, X. et al. 2022). Although several studies (e.g., Wessels et al. 2007; Huang & Kong 2016; Abdel-Kader 2019) have monitored land degradation by assessing the greenness of terrestrial ecosystems, the issues around climate-driven ecosystem degradation need attention as well. This is because in terrestrial ecosystems (e.g., savannah grassland, mopane, semi-arid, and arid), the growth of vegetation cover depends on rainfall, which is highly variable in arid lands (Borgwardt et al. 2019).

The normalised difference vegetation index (NDVI) is an index used to monitor vegetative vigour. The NDVI values range from −1 to +1, where in terms of vegetative healthiness, values less than 0.45 represent dry vegetation or degraded environment while values greater than 0.45 represent healthy vegetation (Valle et al. 2019). The NDVI trends in most dry regions across the world tend to vary in direction and magnitude. Therefore, for any meaningful mapping of permanent degradation, the contribution of precipitation to degradation has to be removed (Wessels et al. 2007; Dagnachew et al. 2020; Matarira et al. 2021). However, the separation of the two drivers (climate- and human-induced) of land degradation is regarded as important in the management of terrestrial landscapes, although it has been challenging (Abdel-Kader 2019). Recent studies (e.g., Ibrahim et al. 2015; Abdel-Kader 2019) on terrestrial ecosystem degradation have capitalised on the use of the residual trend analysis (RESTREND) method in distinguishing the influence of climate from human-induced land degradation. Although several studies have been carried out to identify land degradation in Zimbabwe, few studies (Kibena et al. 2014; Masocha et al. 2017) have focused on annual NDVI trends. RESTREND technique is worth exploring in Zimbabwe, especially the Manyame River Catchment. There are very few studies that had attempted mapping only human-induced land degradation in Zimbabwe. Therefore, this research aims to assess land degradation in the Manyame River Catchment using residual trend analysis. This was done by separating the contribution of rainfall to the primary production of the ecosystem so as to map only human-induced land degradation.

Weather stations that provide representative climatic data are limited in the Manyame River Catchment. All five hydrological subzones in the study area (Figure 1) do not have weather station data available for sharing in each hydrological zone. If these data were available, then they were not usable given that they have many missing data for most of the months for every year considered in this study. However, the nature of our study requires spatially distributed rainfall data (Ibrahim et al. 2015; Hou et al. 2022) like remotely sensed precipitation data. The rainfall dataset in raster format was obtained from the Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) data website. The advantage of using monthly gridded (raster) datasets like CHIRPS is that they provide a complete spatial representation of rainfall (Retalis et al. 2017). The 0.05° × 0.05° spatial resolution rainfall data were extracted from the CHIRPS dataset for the period 2000–2017. The annual rainfall data (resampled to 500 m × 500 m) were used in this study, and this was done by computing the mean monthly rainfall, which was obtained from the CHIRPS website.

Monthly remote sensed Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI = (near infrared band – red band/near infrared band + red band)] data for the period 2000–2017 were used in this study. The monthly NDVI data were extracted from the MODIS13A1 V6 product. MODIS data are available in Hierarchical Data Format (HDF) and was downloaded as tiles from the Earth Explorer website. The tiles were reprojected from sinusoidal projection to World Geodetic System (WGS 84) and finally to World Geodetic System/Universe Transverse Mercator Zone 36 (WGS 84/UTM Zone 36). The re-projection was conducted in QGIS. The MODIS NDVI values were rescaled to the usual NDVI value ranges (−1 to +1) and the NDVI for the Manyame River Catchment area was extracted using the study area shape file as the mask. Annual NDVI raster dataset was stacked while matching acquisition dates of the satellite imagery so as to come up with NDVI time series. Time series of the NDVI was calculated using 18 annual datasets of the NDVI.

Vegetation dynamics have been previously assessed using the NDVI, which can be derived from MODIS or Landsat bands, for example, or downloading the existing MODIS NDVI data (Wardlow & Egbert 2010; Testa et al. 2014; Adenle et al. 2022) from the Earth Explorer website. In this study, mean annual NDVI values were computed to represent the total green biomass production for each year. Thus, we used changes in vegetative greenness (NDVI trend maps) to explain areas experiencing low NDVI values as degrading (Epinat et al. 2001; Gu & Wylie 2015; Kuri et al. 2019).

The Linear Trend Analysis (LTA) method was employed to distinguish the degraded land from non-degraded areas in the Manyame River Catchment. Other scholars had applied LTA in assessing variations in NDVI values with changes in time (years) (Ibrahim et al. 2015; Radda et al. 2021). Change in NDVI values was assessed using linear regression analysis, following AbdelRahman et al. (2018). The annual NDVI values for the period 2000–2017 were correlated with time (year). An ordinary least squares regression model, see, for example, Huang & Kong (2016); Ghorbanian et al. (2022) is best in estimating slope reflecting a significant trend in vegetation greenness (NDVI) with time x (years). The resultant slope coefficient infers the rate and magnitude of change per year, following Ibrahim et al. (2015); Yadav et al. (2022).

Only maps which showed significant variations in greenness due to a combination of both climate changes and human activities were considered in this study; at a 95% significance level. The NDVI residual trend (severity of land degradation) categories were quantified in ArcMAP using slicing and zonal statistics functions thereby establishing percentages of areas covered by each trend category. This helped in the determination of the extent of the decline in the vegetation cover.

According to Ibrahim et al. (2015), trend analysis of the residual NDVI can explain the magnitude of degradation processes. The RESTREND method was used in this research to remove the role of rainfall in ecosystem productivity and detect only the influence of human activities. The significant residual NDVI trend, negative or positive, was used to identify regions experiencing degradation and those with improved conditions, respectively. These would be vegetation changes due to other factors different from rainfall variations (Wessels et al. 2007; Tolche et al. 2022). It has been argued that climate variability and other physical events do not cause a directional change in the residuals but human interference in the environment does (Wessels et al. 2007; Huang & Kong 2016). RESTREND analysis involved regression of the NDVI against rainfall, using the ordinary least squares model, following the study by Ibrahim et al. (2015). The annual NDVI and the annual precipitation were used. In this model, the NDVI was the dependent variable and rainfall was the independent variable. The correlation also produced slopes, intercepts, and R² values that were also useful in the analysis, where only significant regions at 95% were considered.

Analysis of results equated areas with a negative trend to be degraded, and those with a positive trend to have been improved (Ibrahim et al. 2015; Radda et al. 2021; Li, S. et al. 2022). Examining residual trends allowed the identification of areas with human-induced degradation as well as those with human-induced improvement of the vegetation cover. The study further conducted statistical analysis for the observed residual trends to identify areas with significant negative trends confirming significant degraded areas. Statistical significance of all the regression equations was tested using Mann–Kendall. To capture the dynamics of decreasing green biomass, pixels without significant slopes or with significant positive slopes were not included in further analysis. The areas without significant slopes are those areas where trends in vegetation greenness are associated with trends in rainfall dynamics.

This non-parametric statistical method was applied to examine the consistency of RESTREND in the Manyame River Catchment, Zimbabwe. It was first described by Mann in 1945 and has been widely used in environmental time series data analysis (Gao & Liu 2010; Matarira et al. 2021). Kendall's coefficient, τ, measures the extent to which a trend is monotonically increasing or decreasing. It ranges from −1 to +1, where −1 indicates a trend that is consistently decreasing and never increases and +1 indicates a trend that is consistently increasing and never decreases. A value of 0 indicates no trend.

The substantial decrease (severe degradation) is characterised mainly by the residential areas or towns covering parts of St. Cecilia, Chikafa, Dema, Norton, Ruwa, Chinhoyi, and Harare. The most degraded areas are spread throughout the Manyame River Catchment as indicated by the severe degradation class that represents a substantial decline in the vegetation cover. These areas are adjacent to the main settlement zones within the Manyame River Catchment. Significant severe degradation was also depicted during the same period in some parts of the area, usually coinciding with the boundaries of the semi-arid regions (Mushumbi Pools) and intense human activity influences such as land clearing (Raffingora). Moderately decreasing trends (moderate degradation) have been observed just in proximity to the settlement regions, whereas light degradation forms part of the mountainous regions and some wilderness in the study area.

Table 1 shows corresponding areas of significant degradation that were computed for the areal extent and percentages of residual NDVI trend values for each category in Figure 6. Areas that show a substantial decrease in vegetation constitute 4.08% of the area. The area with increasing vegetation covers 38.58% of the study region while 19.80% showed a moderate decrease and the remaining 37.54% represented those areas with non-significant residual NDVI trends for the period 2000–2017.

Table 2 shows a weak negative correlation between settlement (cities) and land degradation. This suggests that land degradation in the study area is not entirely due to human influence.

The Mann–Kendall coefficient τ was calculated to test whether the residual trend was monotonic or not from 2000 to 2017. The Kendall coefficient was applied to annual data (τ). The analysis shows that the annual trends of the RESTREND in the study area were weakly positive but highly significant (p < 0.05) for this RESTREND model in this study. This means that the trend of the adjusted NDVI residuals for both soil moisture and rainfall is increasing significantly in the area.

From the analysis of annual NDVI trends, vegetation cover decline was observed in Chikafa, Norton, Mushumbi Pools, and Harare of the Manyame River Catchment, during the period 2000–2017. Since significant negative trends were also exhibited during the same period (Figure 6) and considering high precipitation variability in the drier areas of the Manyame River Catchment and the influence of human settlement, which disturbs the vegetation cover in the catchment, it is suggested that the decrease could be a result of either both the human involvement and/or rainfall anomalies. Raffingora, Mushumbi Pools and Chikafa (dry regions), and some parts of Harare, Chinhoyi, and Norton (high settlement density) experienced wide coverage of negative NDVI trends. The dry regions of Zimbabwe lie in agro-ecological regions 3 (500–600 mm), 4 (300–400 mm), and 5 (<300 mm) averages per year, which are characterised by low and erratic rainfall (Marongwe & FAO 2012). Zimbabwe has experienced an increased frequency of drought associated with El Niño events. One to three droughts have been recorded every 10 years (Kuri et al. 2019). These droughts have been associated with the observed overall negative NDVI trends in most parts of Zimbabwe, including the Manyame River Catchment. Poor soils could also have accelerated land degradation in these semi-arid regions. Most communal areas that have suffered huge biomass losses have granite-derived sandy soils, that are highly erodible (Fischer et al. 2008). The communal areas of Mbire district are characterised by sodic and sandy soils (Chimweta et al. 2018) and have evidence of widespread degradation, presumably, because of these poor soils.

Some researchers had found that the NDVI and rainfall relationships can explain ecosystem productivity variations and deterioration in the land condition in terrestrial zones (e.g., Alcamo et al. 2003; Davies et al. 2016). The distribution of regression slopes of the NDVI against rainfall, in this study, agreed with those reported in other studies (Ibrahim et al. 2015; Dubovyk 2017). The dry areas of the Manyame River Catchment exhibited strong linear relationships as shown by high slope values in those semi-arid regions. The slopes indicate the amount of change in the vegetation cover per unit change in rainfall (Ahmad & Pandey 2018; Hui et al. 2022). Pixels in the semi-arid regions of Guruve and Mbire have higher regression slope values compared to the high-rainfall mountainous areas, which are covered by the evergreen Miombo forests. These findings indicate that vegetation in dry areas is highly responsive to high rainfall variability (Wessels et al. 2007; Gu et al. 2022). High-rainfall zones that include humid forest areas of Harare and Chinhoyi exhibited weak responses. This is because the amount of annual rainfall usually exceed a certain threshold, above which vegetation becomes non-responsive (Lu et al. 2015). Such high-altitude areas, characterised by deep loamy soils that have high water-holding capacity and are not easily eroded, have sustained vegetation growth even in low rainfall regimes (Mensah 2019).

According to Wessels et al. (2007), assessing vegetation changes without removing the rainfall impact has misleading implications for landscape management. This is because human beings alter the structure of landscapes, mainly through various land use practices. The study observed that approximately 4.08% of the study area exhibited negative residual trends. These trends were mainly concentrated in Greater Harare, Chinhoyi, Mushumbi Pools, and parts of Guruve. The study also revealed portions of severe degradation in the study area that coincided with almost the same areas of high susceptibility to intensive human involvement and high rainfall variability. The clearance of forests for agricultural land use, together with intensive fuelwood extraction is rapidly depleting communal areas of vegetation that results in soil erosion. The breakdown in sustainable natural resources management has led to the siltation of water reservoirs such as the Manyame River. The downstream effects of siltation have led to reductions in the river's capacity and depleted the original state of the aquatic ecosystems (Chamizo et al. 2012; Tamene et al. 2017). The excessive settlement pressure on communal lands seems to be the primary cause of degraded agricultural land. Clearly, most of the communal land areas have become unsuitable for agriculture due to the hostile terrains and impoverished soils. The unfavourable topographical and soil conditions pose restrictions on the sustainable delivery of ecological services (Wang & Cai 2010; Lu et al. 2015).

Based on the findings of this study, a residual trend analysis method was demonstrably useful in distinguishing between climate-induced and human-induced factors, as drivers of degradation in terrestrial landscapes. In this regard, this study has noted the role of climate in vegetation cover changes, particularly, in drier regions. The NDVI and rainfall relationships can explain ecosystem productivity variations. Vegetation in dry areas is responsive to the high rainfall variability, while high-rainfall zones exhibit weak responses. This is because the amount of annual rainfall usually exceed a certain threshold, above which vegetation becomes non-responsive. 4.08% of the land in the Manyame River Catchment was significantly degraded because of human activities. The degraded trend in vegetation cover was most severe in Harare, Norton, Chinhoyi, and Mushumbi Pools. However, this result has arrived from data showing the existence of patches of degraded land, interspersed between positive trends. The study recognizes that there are other factors that influence land degradation processes. We hypothesise that soil characteristics and/or geology (Faruqi 2021), vegetative water use efficiency, and topographic factors could also have an effect on our results in contributing to insignificant zones, which are represented in white coloured cells. As a result, possibly to improve the outcome and reduce insignificant cells observed in this study, we recommend further studies to investigate the influence of soil characteristics, vegetative water use efficiency, and topographic factors on vegetation cover in fragile ecosystems similar to that of the Manyame River Catchment in the context of land degradation. This is a worthwhile contribution to our understanding of the increasing need for informed integrated policy interventions for sustainable human settlements.

We are grateful to Dadirai Matarira for the useful comments that helped us improve this contribution.

No funding was received for this research work.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.