Using open-source software GRASS GIS for analysis of the environmental patterns in Lake Chad, Central Africa

Lake Chad is an endorheic shallow lake of Central Africa originating from the large paleolake Chad, a relict of a giant Quaternary Megalake (Drake et al., 2022). Major sediment types include the outcrops of Quaternary (Pleistocene) and Tertiary (Neogene) formations (Figure 2). Geologically, Lake Chad presents an intracratonic sag basin with basement comprising a suite of crystalline rocks related to the Pan-African orogeny (Schuster et al., 2009).

Figure 2.

Here, the geological units are formed during late Precambrian (pCm) and are mostly located within the Chad geological province (Figure 3). The surrounding provinces with connected watershed region include the Nubian Uplift in the north and Erdis Kufra basin in the northeast, which includes the prominent desert Mourdi Depression and is limited by the Paleozoic sandstones of the Ennedi Plateu (Figures 1 and 3). The Mourdi Depression presents a rocky valley and a grazing ground area located along a route from the Ennedi Plateau toward the northeastern region of the lake. The context of the geological setting is explained in the effects of mineral constituting soil structure and affecting its water permeability that, in turn, contributes to lake shrinkage. Besides, geologic settings are directly related to soil types, which, in turn, are connected to the vegetation types: soft soils near the lacustrine environment are dominated by aquatic plants, while rocky hard soils are typically found in the semi-desert landscapes with rare vegetation.

Figure 3.

The palaeoecological reconstructions identified relict Saharan river deltas, landforms, and paleoshoreline of Megalake Chad (Leblanc et al. 2006). Since the last glacial maximum, the transgressed lacustrine deposits in the depression of the Lake Chad Basin affected the aeolian geomorphology in the catchment area through flooded accumulation of dunes in the Bodélé Depression in northern Chad and dune islands in the surroundings (Bristow and Armitage, 2016). Over the past 50,000 years, the area of Chad Lake has varied significantly in size due to a variety of factors. These include the climatic fluctuations in Sahara since the Late Miocene (Durand, 1982; Servant, M., 1983; Pouyaud and Colombani, 1989), the monsoon-driven hydrodynamics (Bouchette et al. 2010), and the geological uplifts (Burke, 1976). As identified by the littoral sedimentary records including fine-grained deposits and diatoms, Chad Megalake was one of the giant lakes in the Sahara existing in Quaternary during the humid period (Mémin et al., 2020).

A southern border of the Lake Chad region is presented by the western and eastern segments of the Congo Precambrian Belt (Figure 3). These geological units forming the Late Proterozoic orogenic chain present blocks shaped by the crustal evolution of the region. Most of Chad’s granitoids originated from the end of Proterozoic (Kusnir and Moutaye, 1997). The Precambrian formations of Chad are dated by the Proterozoic Eon and include a series of sedimentary rocks originating from volcanic material with significant mineral potential. As a result of such complex geologic development, the lake has now immense catchment basins of the groundwaters, which exceeds an area of 500,000 km².

The association of geologic setting with land cover types is expressed through soil types. Thus, in the flooded plain of Lake Chad region, the main mineral type is smectite clay associated with kaolinite and illite (Jean Pierre et al. 2019), which supports groundwater hydrology and supplies water level in the lake during periods of droughts. In contrast to the rich groundwater supply, the surficial inflow of the southern segment of Lake Chad is supplied up to 90% from the Chari-Logone catchment (Black et al., 2021). The Chari and Logone rivers originate in southern highlands of Cameroon and the Central African Republic (Hutchinson et al., 1992). The northern segment depends on the minor inflows from the Komadugu Yobe River originating in Nigeria and the overflows from the southern basin (Fougou and Lemoalle, 2022). Early Cretaceous clastic sediments are found in the West African Rift Subsystem of Chad (Genik, 1992), which presents the extent of the East African Rift in the southeastern region. A part of the Great Rift Valley developed during Pan African orogeny. The East African Rift System has geologic implications on the tectonic and structural aspects of mineral resource distribution and groundwater level in Chad. The hydrology of Lake Chad plays a key role in the environmental sustainability of Africa through regulating energy and carbon cycles, participating in global climate balance, and as a key regional water source. Moreover, Lake Chad plays a key socioeconomic role in the region through support of agriculture activities and fishery and as a water source to over 30 M people in the surrounding countries: Chad, Cameroon, Niger, and Nigeria (Vaquero et al., 2021). As a transboundary water source, Lake Chad maintains the livelihoods for existence of local population, such as fishing and farming (Sarch and Birkett, 2000; Zieba et al., 2017; Baroin and Boutrais, 1997). Besides, it provides diverse ecosystem goods and services, and natural resources such as firewood or livestock rearing. This highlights the importance of measures on the environmental conservation of the lake and sustainable development of its resources. As a result, the agricultural activities of the Lake Chad region are concentrated around the lake and the two river systems which play a major role in the agricultural economics of the region.

The human-induced triggers contributing to the shrinking of Lake Chad include unsustainable agriculture with groundwaters intensively used for irrigation (Lemoalle and Magrin, 2014), increasing demand for freshwater from the lake by the population, and drilling of boreholes as hydrogeologic activities for constructing large-scale irrigation systems (Birkett, 2000). The intensive modern irrigation and farming methods in agriculture practicing include fertilizers and farm machinery for maize and millet production, which led to land degradation (Nilsson et al., 2016). Further, the analysis of the vegetation patterns and surface water levels in Lake Chad demonstrated the links of land cover types with rainfall fluctuations during wet and dry periods (Balme et al. 2006), as well as the effects from seasonally contrasted climate (Gbetkom et al., 2023).

Annotating of the land cover classes around Lake Chad based on the recognized image features aimed at accurate object recognition during classification of the satellite images, which required preliminary examination of the images and analysis of major landscape types according to FAO classification. Hence, the following 12 classes were selected and identified, according to FAO: Class 1 – water bodies; Class 2 – wetlands and floodplain with aquatic plants; Class 3 – swamp forest; Class 4 – woodland; Class 5 – rocky land; Class 6 – bare soil; Class 7 – sandy areas; Class 8 – steppe; Class 9 – agriculture and irrigated areas; Class 10 – settlements and urban areas; Class 11 – savannah (Sahelian short grass savannah and herbaceous savannah); and Class 12 – gallery forest and riparian forests. Image features and associate land cover classes were recognized in the images, identified, and described as patches of land cover classes constituting the landscapes of Lake Chad. Global warming plays a key role in the meteorological setting of Chad Lake through increase of temperatures and modification of precipitation patterns. Thus, annual rainfall trends significantly decreased from 2018 to 2022 at all locations in Chad Lake. This is especially notable for northern regions located close to the Sahara Desert, which receive less precipitation in the form of rare and irregular rainfalls. Regional difference in precipitation level was reported by Pattnayak et al. (2019) as follows: rainfall (cm/month): 0.30 for Faya, 3.26 for Ati, 4.47 for Ndjamena, 9.09 for Mondou, which shows the increasing trend for cities located southward and few precipitations for the northern regions close to Sahara. Such variations in precipitation level affect the drought patterns in the northern regions of Chad and the surrounding regions. Apart from precipitation, the temperature differs in various regions of Lake Chad according to the distance from Sahara as follows: the mean temperature ranges between 27.10 °C (Moundou) and 30 °C (Mongo), while the variability ranges between 0.5 °C (Ndjamena) and 1.03 °C (Abeche). Finally, besides climatic factors, the demographic pressure is also a serious factor in shrinkage of the lake surface and increase in droughts. Thus, the number of people living around the lake increased from 13 M to ca. 40 M from 1960 to 2022, which contributed to the changes in land cover types, especially increase in land cover classes such as 8) bare soil and 7) settlements and urban areas, which replaced the natural land cover types dominating earlier.

Pixels or objects with similar spectral characteristics on the images were allocated to the designated classes using the algorithm of the maximum-likelihood discriminant analysis classifier embedded in GRASS GIS. Using this approach, the “i.maxlik” module enabled to classify the three images and indicate specific land cover classes around Lake Chad. Annotations of the objects as land cover classes included recognition of the distribution of their extent, changes from 2013 to 2022, and detecting shrinkage in water areas and wetlands around Lake Chad and partially overlapped classes in the savannah regions. The polygon structure on the landscape mosaic on the classified maps demonstrates the interspersed patches of the landscape segments, which are identified on the Landsat images from 2013 to 2022 accordingly. Likewise, changes in the annotated land cover classes on the images demonstrate the climate effects which resulted in variations of the land cover classes used for identification of the environmental changes around Lake Chad.

The comparison of the land cover associations over years detected on the multispectral images and their changes with time on the original images and classified maps enabled evaluating the results. Although some land cover classes experienced shrinkage or changes in contours, their inner structure and topology remained identical with regard to neighboring classes, for instance, the distribution of savannah, sandy areas in the Sahara Desert and wetlands around Lake Chad. Since images were obtained with 4–5 years of time gap, such features were matched to find all possible cases of land cover changes to complete the map of the land cover classes in the region of Lake Chad.

As all the Landsat images were collected in late November–early December period, the phenological cycles of vegetation represented in the images did not affect identification of the vegetation patterns since extracting the information on land cover classes was possible in a comparative manner with regard to the period of plant growth in the Chad region (late November–early December). Using the procedure described above, the land cover classes were created with the aim of identifying the target patches of each landscape type to detect changes from 2013 to 2022. This mosaic of land cover types represents the extent of the diverse landscape types around Lake Chad to accurately detect landscape changes caused by environmental and climate effects. The changes were detected between 2013, 2018, and 2022 and used for evaluating variations in land cover types around the Lake Chad region. The repetitive vegetation patterns and objects identified as land cover classes were recognized and annotated on the three resulting maps showing the landscape changes in the land cover classes around Lake Chad. The annotated objects for patches of land cover classes identified on the images forming the land cover classes around Lake Chad are presented as three respective maps in Figure 5, with the statistical evaluation shown in Figure 6.

Global warming plays a key role in the meteorological setting of Chad Lake through increase in temperatures and modification of precipitation patterns. Thus, annual rainfall trends significantly decreased from 2018 to 2022 at all locations in Chad Lake. This is especially notable for northern regions located close to the Sahara Desert which receive less precipitation in the form of rare and irregular rainfalls. Regional difference in precipitation level was reported by Pattnayak et al. (2019) as follows: rainfall (cm/month): 0.30 for Faya, 3.26 for Ati, 4.47 for Ndjamena, 9.09 for Mondou, which shows the increasing trend for cities located southward and few precipitations for the northern regions close to Sahara. Such variations in precipitation level affect the drought patterns in the northern regions of Chad and the surrounding regions.

Apart from precipitation, the temperature differs in various regions of Lake Chad according to the distance from Sahara as follows: the mean temperature ranges between 27.10 °C (Moundou) and 30 °C (Mongo), while the variability ranges between 0.5 °C (Ndjamena) and 1.03 °C (Abeche). Finally, besides climatic factors, the demographic pressure is also a serious factor in shrinkage of the lake surface and increase in droughts. Thus, the number of people living around the lake increased from 13 M to ca. 40 M from 1960 to 2022, which contributed to the changes in land cover types, especially increase in land cover classes such as 8) bare soil and 7) settlements and urban areas, which replaced the natural land cover types dominating earlier.

The analysis of the classified images (Figure 5) shows an overall trend in dryness and desertification of Chad Lake, which includes expansion of the desert zone southward as well as disappearance of forests replaced by the savannah shrubs and grasslands. At the same time, since Lake Chad is a part of the complex region of the sub-Saharan region, there are also reported fluctuations of the dry and wet phases in this region reconstructed for Quaternary period (Rognon, 1976). These are supported by the modeled climate changes that existed in Central Africa as proved by the sedimentological and geomorphological analysis indicating the location of alluvial flood plains, terraces, and lacustrine deltas in the past. Therefore, the additional triggers causing variations in the water level of the lake experience the effects from the long-term climatic oscillations.

The presented series of maps shows a fluctuating behavior of Chad Lake in Central Africa on the border with Sahara through the investigated periods as a result of the seasonal effects from temperature acting as triggers to land cover changes. Several segments of land cover types were compared for the lake and the surroundings. Moreover, the correlations between the extent of Chad Lake and the effects from temperature are related to the influence of climate on the environment of southern Sahara. In view of the advantage of the GRASS GIS-based approach to analyze the environmental properties of the landscapes in Chad, this paper presents a short period times series analysis of Chad Lake along with the applications of the image processing methods. The main idea of this study was to apply in a single workflow framework using open-source data Landsat OLI/TIRS and software GRASS GIS.

The study evaluated both the multi-year images and vector maps to reveal the information on Chad Lake behavior with varying data with regard to a sequence of different years. In such a way, the Landsat images were analyzed as multi-temporal time series as variation analysis. Changes in land cover types and water extent of the depressions of Chad Lake were evaluated using a series of Landsat 8–9 OLI/TIRS images processed with GRASS GIS using advanced scripting approach.

Visualizing spatiotemporal changes in the borders of Lake Chad is essential for environmental monitoring. The basin of Lake Chad is subject to the changes caused by anthropogenic impact and variability in climate patterns. Therefore, mapping changes using satellite images on different dates enables evaluating these changes in the lacustrine landscapes. The anthropogenic effects may be diminished to the adaptive ecological monitoring measures, as well as sustainable agriculture and farming in the Chad Lake region. Advanced approach to visualization of the land cover types in Sahelian Africa is through cartographic processing of the remote sensing data (Buma, and Lee, 2020; Lemenkova and Debeir, 2022a). Using remote sensing data in mapping enables applying advanced methods for monitoring environmental changes (Bardinet, 1981; Zhu et al., 2017; Li et al., 2021; Lemenkova and Debeir, 2022b).

The comparison of several images covering Lake Chad classified for various years enabled visualizing the changes in the land cover types and increased shrinkage of land in the cold period of autumn/winter months. Thus, it highlighted the relationship between climate that contributed to the shrinkage of lake via evaporation and raised temperatures and the hydrological response of lacustrine environment around Chad that reflected in the changes of lacustrine landscapes and depleted nearby vegetation. The difference in behavior of various land cover types and Lake Chad suggests that landscape dynamics along with seasonal variations of temperature and precipitation around Chad as quantified by the clustering and classification of the images are primarily determined by the properties of the soil in the land cover types of the landscapes surrounding Chad Lake due to the different content of minerals in soil.

The present study extended the existing research focused on variations of Lake Chad through investigation of the land cover changes in its surroundings. To evaluate land cover dynamics in the Chad Lake area, the GRASS GIS algorithms of modules were applied for image processing. Three satellite images Landsat 8–9 OLI/TIRS were classified, compared, and analyzed. The results showed the visualized changes in the land cover types of Chad Lake. For comparison of changes, the images were collected in the same period of different years. The objective approach of the machine-based classification of the images strengthens the discrimination of the images. The images are represented as triplets of the multispectral Landsat bands processed by the maximum likelihood classification. The theoretical basis of image analysis lies within the frame of remote sensing theory on image properties and variations in spectral reflectance.

The cumulative effects from several factors – geographic location, closeness to the area of Sahara desert, access to groundwater inflow to Chad Lake and local climate setting – control the water level in the lake. Moreover, during different seasons of the calendar year, depending on the global weather and integrated heat from the Sahara Desert and Sahel region, Chad Lake demonstrated fluctuations in shrinkage. Furthermore, the performed time series analysis indicates gain in urban areas and transitional land cover types from forest to savannah, as well as an increase in bare soil on the analyzed images covering autumn and winter months. Such information can support the selection of the optimized methods of agricultural lands adopted for such seasonal climate–environmental fluctuations, allowing identification of the best methods of supporting agriculture and plantation activities in the surrounding areas of Chad Lake and enabling to achieve success in land management in Central Africa.

Apart from the detected general trends in drying of Chad Lake, visible on the satellite images, there is a regional difference identified between the southern and northern segments of Chad Lake. Compared to the southern segment, the northern pool has a larger extent of variability due to higher vulnerability to droughts, as also proved in previous studies (Lemoalle et al., 2012). The reasons for such spatial heterogeneity in the desiccation of Lake Chad reflect local sensitivity to climate effects triggered by regional differences in the rainfall patterns and vegetation patterns in the landscapes. As visible on the classified images of several years (Figure 5), the remaining area of the permanent open water areas in the southern segment of the basin is surrounded by more densely vegetated swamps and small pools. This helps to maintain a lacustrine environment and prevents the southern segment from desiccation, as also reported in earlier works (Lauwaet et al., 2012).

The proposed framework benefits from the GRASS GIS functionality and coding approach and exhibits the advantages of the Landsat OLI/TIRS scenes that balance between the accuracy of visualization and complexity of the land cover types while retaining sufficient discrimination threshold for image classification. Practical results are presented as three maps based on the classified images. Finally, the workflow of image processing and plotting is effectively optimized using the alternative methods of scripts. Such methods are implemented in this work using both GMT and GRASS GIS and evaluated for mapping, annotation, and classification tasks. The GRASS GIS method is particularly efficient in terms of mapping production costs since the toolset is free and open for use. However, although the range of technical functionality of GRASS GIS is wide, the use of scripts in cartography has certain limitations. In particular, it requires a familiarity with the programming paradigm, which supports scripts through the embedded syntax.

In fact, the availability of the open-source advanced cartographic programs, such as GRASS GIS, scales with the learning approach which is required to master these methods. In addition, the import of datasets is often performed via the integration with other libraries and Python. This article presented a case of such approaches showing a GRASS GIS script-based method for satellite image processing and cartographic visualization of land cover changes in Lake Chad. Unlike previous traditional GIS-based methods of mapping which operate in the spatial domain, we performed mapping through the GRASS GIS codes, which ensured comprehensive workflow for the cartographic specification of the maps. The GIS-based methods relying on graphical user interface (GUI) tend to use the predefined designed tools and graphical methods of representation. In contrast, the proposed GRASS GIS method outputs maps made using individual adjustments in each line of the programming code using expressions and its syntax. The maps were compared pairwise for various years to analyze the underlying reasons for land cover changes with regard to the climate–environmental setting in Sahelian Africa, Chad.

Remote sensing data contain essential information regarding the landscapes of the Earth. Accurate classification of time series of the satellite images enables detecting changes in landscapes using this information derived from the satellite images. Such a technique supports environmental modeling of remotely located areas where direct observations are difficult to perform, such as Chad lacustrine landscapes located in Central Africa. To this end, this paper presents a case of the analysis of changes in areas of land cover types in Chad as a representative case of African regions. The dataset included the series of satellite Landsat 8–9 OLI/TIRS obtained from USGS. The detection of the land cover types was performed using the algorithms of image classification. The advanced solution of GRASS GIS for geospatial data processing is demonstrated by a programming approach that utilizes scripts for data automation. The objects that belong to the same category of segments were defined and characterized as representing land cover classes using the difference in spectral reflectance retrieved from the satellite images.

The current study contributed to the environmental monitoring of Lake Chad, which has a long tradition, going back to many years of hydrological surveying. Existing reports enabled detection of changes in the actual level of the water horizons in the lake during the continuous research conducted in recent decades (Tilho, 1928; L’Huillier, 1957; Cabot, 1970; Riou, 1972; Tubiana, 1989). Evaluation of changes in Lake Chad relied on the estimation of the climatic and hydrological data that assess the volume balance in the lake based on the field observations (Bader et al., 2011). The present study continued these works by using remote sensing data processed by a scripting approach as a novel cartographic method to environmental analysis aimed at detecting changes in Chad Lake. As demonstrated, GRASS GIS provides a logical language with a straightforward syntax for a console-based cartographic mapping using shell scripts for image processing.

The algorithm approach of scripts by GRASS GIS is largely based on using its syntax, which is very similar to the programming languages. The lines of codes perform operations controlling the appearance of the cartographic elements on the maps and form a script for presented maps plotted in JPG (or other bitmap raster formats), which can be converted to other formats. Apart from the traditional JPG, GRASS GIS recognizes a variety of miscellaneous grid formats and enables data capture via GDAL (r.in.gdal). Using GRASS GIS for satellite image processing and mapping leads to a higher quality of the produced maps, and lower time is required for workflow and scripting, compared to the cumbersome and time-consuming data processing in traditional menu-based GIS. Thus, in contrast to the existing software, GRASS GIS presents a free set of the command-line modules and options for processing geographic data in both raster and vector formats, and in particular, for satellite image processing.

The maps were plotted by the GRASS GIS scripts using the image analysis and classification of the Landsat scenes covering the region of Chad. The demonstrated methods included an automated scripting approach that ensures high print quality and accuracy of maps, prepared using the high-resolution data. Rather than using handmade cartographic workflow as in similar GIS-based studies, GRASS GIS scripting presents improved methodology that emphasizes the importance of programming as a new trend in contemporary cartography and remote sensing. A useful extension of the presented study would be to automatically generate a series of maps covering particular periods of droughts in semi-arid climate of Africa for a complete environmental risk assessment. Adding more datasets, for example, climate data, could extend the variability of factors for risk assessment.

As a machine-based method, it is a fast and effective technique compared to the conventional mapping workflow. The application of GRASS GIS showed the advantages of scripting languages as part of the new directions in modern cartographic development and remote sensing data processing. Setting the parameters of the GRASS GIS codes enables producing the highest possible quality of maps, including matching the projection and color palette tables (CPT). As demonstrated in this paper, processing earth observation data by GIS enables to better understand changes in wetland areas in the Sahelian Africa, thus supporting monitoring of the terrestrial water cycle and land cover changes in drying lakes (Jones et al., 2009). This study revealed the shrinkage of Chad Lake and visualized changes in associated land cover types caused by the cumulated effects from climate factors (droughts, increase in temperature, and decrease in precipitation) and human-related activities.