Using geospatial data to identify land grabbing. Detecting spatial reconfigurations during the implementation of the Nacala Development Corridor in Mozambique with remote sensing and land conflicts databases

ABSTRACT

The contemporary food system pushes agriculture to a globalized value-chain, affecting landscapes, resource access, and institutional arrangements. Institutions operating in Africa adopt development corridors to integrate multisector investments and induce export-driven primary sector, leading to massive land deals, also known as land-grabbing. Organizations struggle to monitor land deals accurately, lacking spatial precision and contextual information for affected communities. This research examines Mozambique's Nacala Corridor, using geospatial data as a tool to detect spatial (re)configurations due to exported-oriented policies and infrastructure. Data from land conflicts databases (Land Matrix and Environmental Justice) were analyzed with remote sensing Landsat and MODIS imagery using multiple indexes, an EVI time series, and the application of the LandTrendr algorithm. The results show that the temporal and spatial analysis of remote-sensing data is in line with the major political and economic dynamics of the region. Hotspots of land cover changes were detected in the same areas where land grabbing were reported; however, reported and detected land areas did not coincide. Temporal analysis showed that institutional changes played a greater role in triggering land use changes than infrastructure implementation. We conclude that land cover modifications, conflicts, and spatial development initiatives follows policies and institutional arrangements targeting international investments.

KEYWORDS:

• Land grabbing
• remote sensing
• development corridor
• development policies
• Africa

Introduction

The phenomenon of massive land acquisition by foreign investors is not recent nor homogeneous across the world and has deep effects on physical and institutional configurations (Sassen, 2013). Although much research has been done, there is no common definition of the phenomenon. Since the first decade of the 2000s, a bigger concern has emerged and many reports were made tracing the scale of land deals, the actors involved, and the impacts. These reports attempted to characterize the phenomenon of land grabbing showing, above all, its complexity and multi-faced nature (Cotula & Vermeulen, 2009; Edelman et al., 2013; Grain & UNAC, 2015; Hall, 2013).

Although international acquisition of agricultural land is often considered as typical land grabbing cases, critical authors advocate for a more nuanced and comprehensive analysis (Borras et al., 2013; Oskarsson et al., 2019). Also, Edelman(2013) presents a critical perspective about weaknesses of databases of global land grabbing arguing for a more “reflexivity” from researchers when drawing trends and conclusions.

In addition to open databases, the present research, thus, utilizes remote sensing (RS) as a supporting tool to enhance land systems’ analysis on regional scale in areas where political and infrastructural changes were introduced, and which led to social conflicts. This study provides insight by combining spatially explicit data and descriptive cases.

The use of RS in land grabbing analysis is still underexplored, with recent studies on land class changes such as mining sites (Hausermann et al., 2018; Oskarsson et al., 2019), building infrastructure (Rienow et al., 2022) and forest cover loss (Neef, 2020). In the context of Mozambique, Bey and Meyfroidt (2021) analysed tendencies of land cover changes related to industrial forest plantation, concluding, among other findings, unconformities between official documents and the actual detected changes.

Remote sensing, however, is also politically explored by land grabbers as a narrative tool to determine “marginal land” and thus qualify certain areas as “open to exploitation” (Nalepa & Bauer, 2012). Consistent to the latter, one of the main drivers of international land deals is the expansion of agricultural land. In this sense, the contemporary food system plays a role by shifting the production and distribution to a more globalized value chain (McMichael, 2012), thus impacting landscapes, access to resources, and the peoples in its frontiers. Much of these frontiers are located in developing countries that embrace an extroverted economic pattern of development (Mosca & Dadá, 2021). In the first decade of the 2000s, more than 70% of reported cases of massive land deals occurred in Africa; Ethiopia, Sudan, and Mozambique being the main targeted countries (Grain & UNAC, 2015).

There is no unanimity on the amount of land that has been under transnational trade, the definition, nor the methodologies that could be used to identify such activities (Borras et al., 2013). Despite the efforts of organizations to monitor and track land deals with local and global networks producing reports (Land Matrix¹ and Environmental Justice Atlas,² to name the most prominent), there is a lack of spatial and temporal precision. Due to difficulties in acquiring reliable data, either by the lack of transparency from companies or governments or lack of systematic data collection in more remote regions, reports are normally based on specific cases and utilize nonspatially explicit methods, or rely on secondary sources of information (Calengo, 2016; Grain & UNAC, 2015).

The case of the Nacala Corridor in Mozambique is emblematic of this, as it is a strategic infrastructural project praised as an alternative development strategy in which massive investments in agriculture would take place. The region is, however, one of the most vulnerable and food-insecure in the country and there was strong resistance from local communities (Mosca, 2021).

This research focuses on the time frame of the implementation of the Nacala Corridor using geospatial data as a tool to detect changes in the spatial configurations that could be outcomes of political and economic policies favourable to international land deals and export-oriented infrastructure.

We aim to address the interdependent and co-evolving nature of land systems (Lambin et al., 2001, 2003), by analysing landscape modifications. The hypothesis is that the implementation of the corridor materializes the reconfiguration of the institutional environment, focusing on primary sector export-oriented infrastructure and policies. This facilitates the processes of land grabbing and by mapping massive land cover modifications, land grabbing hotspots can be detected.

The research addresses the following questions:

• Does the implementation of the Nacala Corridor lead to massive land use/cover (LUC) changes that could be considered land grabbing?

• To what extent can RS support the identification and monitoring of land grabbing cases on a regional level?

• How effective is the combination of spatially implicit and explicit data from multiple sources in detecting land grabbing in data-poor regions?

This paper is divided into seven sections, the second section describes the operationalization of the main concepts, e.g. spatial development initiatives and land grabbing. The third section contextualize the region presenting the Nacala development corridor and the ProSavana program. The fourth section describes the methods utilized in the research and the fifth presents the results. The sixth section is the discussion, and the final section is a brief conclusion of the main findings of this paper.

Operational definition of SDI and land grabbing

Spatial development initiatives and development corridors

Spatial development initiatives (SDI) are associated with public sector efforts to dynamize regions by integrating them into global chains. As government expenditure on infrastructure, they are optimized and planned areas intended to attract international investments and disclose local potentialities (Lambin et al., 2001). SDIs are linked to strategies used to magnify the agglomeration capacity and economies of scale. Nevertheless, it is acknowledged that such initiatives tend to create socio-spatial inequalities as “scale economies generate an uneven pattern of specialization and trade (of both intermediate and final outputs), and market dominance, ultimately creating an irregular mosaic of economic development across regions and countries” (Nogales, 2014, p. 15).

The concept of spatial fix (Harvey, 2002, 2011; Zajontz, 2020) instrumentalizes the analysis of international capital flow as a strategy of establishing conditions to solve the inner and outer dialects of capital accumulation by releasing surpluses of capital, establishing aligned institutional environments and expand fixed infrastructure. Thus, SDIs can be understood as spatial fixes once it creates institutional arrangements and infrastructure that enable the flow of commodities and transnational capital.

In recent decades, there has been a paradigm shift in SDIs as a strategy utilized for international development organizations. What was previously focused on urban-centred, or sector-oriented strategies such as growth poles, industrial districts, or clusters, has shifted to a spatial integrated approach, particularly development corridors (Albie & Cox, 2015). Mozambique is among many African countries that are implementing corridors as the main SDI, generally designed to “(i) enhance physical connectivity; (ii) ensure food security; (iii) support regional trade integration; (iv) nurturing agricultural growth; and (v) absorbing the expansion of large urban areas and cleverly connecting various spots of urban growth, including their hinterlands” (Nogales, 2014, p. 26).

The goals of spatial development initiatives are strongly connected to the need of states and private interests in expanding their territorial influence to ensure food security, energy security (thought biofuel production), non-food commodities supply, vertical integration of value chains, and externalization of commodity production costs, which, according to Cotula and Vermeulen (2009), are some of the main drivers of land dispossession or land grabbing.

Land grabbing

Different strategies have been used to enclose lands, speculate on their values, interfere with the market, control production, and free the labour force into the formal capitalist market. Waves of geopolitical expansion, according to Harvey (2003), are processes inherent to global capitalism that exploit the “uneven geographical condition” for capital accumulation to take advantage of spatial asymmetries.

In the report “Land Grab in Africa” (Kachika, 2013), a differentiation is made between land acquisition and land grabbing. According to the author, land acquisition is the acquirement of “ownership rights” and “use rights” – through lease or concession for a period of time. On the other hand, land grabbing is related to the mean of control, the type of use, and the relative size of the land. Land grabbing is the process of “taking possession or controlling a scale of land for commercial and industrial-agricultural production that is disproportionate to size in comparison to the average land holding in the region” (Kachika, 2013, p. 15).

The open database The Land Matrix Project (2022) collects information about land deals across the world using different parameters to define land grabs. As presented by Sassen (2013) the pertinent types that are included have the following characteristics:

• “Transfer of rights to use, control or own the land thought sale, lease or concession;

• Land-use conversion – from smallholders or environmental functions to large scale commercial use;

• The size is 200 ha or larger, and were not concluded before the year 2000 when the FAO food price index was stable” (p.30)

Borras et al., (2020), and 2012, cited in Hall, 2013). propose a political definition, instead of legal or technical, on which land grabbing is related to the control over land and the resources associated. Their definition does not necessarily imply peasants’ dispossession or its international or national orientation but rather a fulfilment of capital demand. On the other hand, another line of thought (Levien 2012, cited in Hall, 2013) defines land acquisitions as land grabs only when operated through extra-economic means.

Among civil society organizations (CSOs), there is also no consensus about the definition of land grabbing. According to campaigns or specific interests, certain definitions gain more weight. The International Land Coalition (ILC, 2011) defines land grabbing as related to human rights, access to information, and transparency. It also includes the need for proper assessment of massive land operations regarding the socio-economic and environmental impacts.

Land grabbing can also be considered in cases in which it occurs without the expulsion of peasants (Geisler & Feldman, 2011), therefore not concentrating land and not necessarily changing the type of crops, but, for example, changing commercialization methods – e.g. outgrowing schemes. These are especially challenging cases to identify as sometimes the arrangement is not explicitly formalized and there are no visible changes in the land cover.

From the discussion above, land grabbing can be defined by different legal, technical, or human rights approaches; however, they are usually combined depending on the case and the objective of the analysis. As a starting point, authors define land-grabbing based on either the means by which the dispossession occurs, the amount of capital invested, the relative or absolute size of the land, or the actors involved in the process (Hall, 2013). Additionally, for a comprehensive understanding of the phenomenon, the drivers (economic or not), violation of rights and changes in local regulations, and the access to information and transparency of the deal are considered.

The present research operationalizes the concept based on its material (visually detectable) aspects regarding three features:

• the size of the land deals, e.g. concentration of detected land cover modification that sum up to over 200 ha;

• the type of crops produced, once, from reported cases, most crops implemented after land deals are international tradable commodities using capital-intensive technologies, irrigation, and external inputs; and,

• the observable land cover modification into a more intensive production system, characterized by the abrupt increase of specific reflectance parameters.

Contextualization of the Nacala Corridor and the ProSavana program

After the end of the civil war (1992) and the establishment of the new constitution (1994), Mozambique has increasingly opened its economy and privatized former state-owned companies. Known as the second wave of privatization (Byiers et al., 2020) in the early 2000s, ports and railways were either privatized or set to long-term concessions with public-private arrangements. The process of privatization was combined with the focus on attracting foreign direct investments (FDIs) for local and international partnerships. These projects were mainly linked to the extractivist and mining sectors. Especially in the mid−2000s, discoveries of one of the largest coal reserves in the world along with significant deposits of other valuable minerals in the Tete province attracted international capital and high rates of economic growth (Byiers et al., 2020). Although these mega-projects led to high economic growth and a surge in FDI, the poverty rate and food insecurity did not improve either nationally or regionally (DiMatteo & Schoneveld, 2016).

Economic decisions leading to liberalization and export-oriented policies have both direct and indirect impacts on land conversions, land degradation, and farmers’ displacement (Barbier, 2000). Mosca and Dadá (2021) present a comprehensive analysis demonstrating the economic tendency to the adoption of an extraverted model in Mozambique. There is continuous economic growth together with an increase in international debt and trade balance deficit. This condition is characterized by the nature of goods internationally traded, increasing imports of food and consumer goods, and the export of natural resources and agricultural commodities with low added value. In the last 20 years, new commodities such as coal, gas, tobacco, and aluminium started to be exported, showing the increased proportion of the primary sector for economic growth. Also, another important feature is the steep increase in national imports of machinery from 2011 to 2014 (World Bank, 2022), which represents the intensification of resource exploitation in the period.

The process of opening the economy to transnational actors led by the implementation of the Nacala Integrated Logistic Corridor (CLIN) had the beginning of operations of the Brazilian mining company Vale in 2004 as a starting point. The Nacala Corridor was also one of the priority development strategies for the government’s strategic plan for the agrarian sector from 2011 to 2020. The same plan allowed the government to establish a propitious environment for the private sector to invest, produce, and commercialize, enabling the transformation of subsistence agriculture to competitive agriculture (Kato, 2019; MASA, 2015; National Development Strategy, 2014).

The main objectives of the corridor are the connection of the mining region of Tete to the Nacala Port, encouraging commercial agriculture in the already highly productive hinterlands of Nampula, and fostering the industrialization of the province’s capital. The institutional arrangement includes privatizations, partial or total concessions, and partnerships among national and international investors (Kato, 2019).

Coal dominates the Nacala rail line in trade volumes reflecting current political attention to extractive resources at the national and local levels (Byiers et al., 2020). Nevertheless, following Nogales (2014) conceptualization of development corridors, the Nacala corridor is not foreseen to be a transport or logistics corridor but to be a comprehensive growth corridor.

This can be understood once several initiatives are being implemented alongside the corridor benefitting from the institutional and infrastructural setup (SDIs). Government projects such as PROIRRI, PROMER, SUSTENTA, and several projects from the World Bank, African Union, and transnational institutions built a “complex web of projects and initiatives” more or less connected with local actors (Byiers et al., 2020).³

Not least of all, the most visible and important initiative beyond the mining sector is the Mozambique-Brazil-Japan cooperation program ProSavana. The program was first signed in 2009, but was only disclosed to the public in 2013 after pressure from social organizations (Grain & UNAC, 2015). The program encompasses 14 million hectares in 19 districts in the provinces of Nampula, Zambezia, and Niassa. The aim is to shift from traditional family farming to competitive production, utilizing techniques and systems that increase productivity. For that, the program has envisioned the facilitation of access to agriculture inputs, development of productive infrastructure, support for farmers’ organizations, research, and financial schemes (MASA, 2015).

In 2015, the government developed a master plan targeting the agricultural sector within the Nacala corridor. The master plan has the ProSavana program at its core and further develops some of its visions. It presents a strong export-oriented vision and states that to be fully realized, the increase in productivity should be linked to improved income for farmers, thus improving access to markets that can provide better payment for its production. Its vision states that “a prosper, competitive and sustainable agrarian sector” offers solutions to “food security and nutritional challenges and is able to reach global markets” (MASA, 2015, p. 1). The master plan acknowledges the importance of smallholder farmers. The plan is focused on emergent farmers – property areas up to 10 ha – aiming at their conversion into the logic of commercial exploration and entrepreneurship (Garcia & Kato, 2019; Kato, 2019). Figure 1 presents the districts included in the ProSavana program and the location of the corridor’s railway, contextualizing the districts included in the present research.

Figure 1. Location of analysed districts, districts included in the ProSavana program, the renewed railway, and its stations. Source of base map: Earthstar Geographics.

The promise of the implementation of the project generated anxiety and fears of land dispossession and conflicts (Gonçalves, 2020). An international alliance to resist the project was organized and fostered by Brazilian and Mozambican social movements (Grain & UNAC, 2015). The master plan was reviewed, moreover, due to changes in the international context – reorientation of Brazilian external policies, fluctuation of commodity prices, economic shock due to the pandemic – reports are that the ProSavana program was suspended, the pace of implementation of the Corridor has decreased, and some projects in Nacala reoriented (da Silva, 2020; Economico, 2020; Garcia & Kato, 2019).

Materials and methods

The conflicts reported in the Nacala corridor have a strong connection with the concentration of land (changes in land management), increase in capital and external inputs, and change in crop types (Calengo, 2016; Mosca & Bruna, 2015; The Land Matrix, 2022). The reason for this choice in operationalization of land grabbing is that this research focuses on the capacity of satellite imagery in supporting the monitoring of massive land deals. Therefore, little can be affirmed by solely using satellite imagery to assess land management changes for what would be defined as land grabbing or human rights conflicts. However, the combination of data from land conflicts with LUC intensification offers an important approximation of possible massive land deals that could be characterized as land grabbing.

Figure 2 presents the main concepts of this research, their characterization, and the interlinkage with its spatial impact, thus feasibly analysed by RS data.

Figure 2. Summary of the operational description of SDI and land grabbing adopted in this research.

This research combines the analysis of secondary data sources of land conflicts and remote-sensing imagery. Regarding the latter, two different methods were used, both methods are focused on spatial-temporal changes through time-series analysis, a method that “allows prediction of the timing of changes, opening new avenues to better link macro-economic transformations to land-use changes” (Veldkamp & Lambin, 2001, p. 3). Methodologically, this research addresses the less explored perspective of land cover modification and intensification, instead of LUC changes or LUC classes conversions (Lambin et al., 2003; Ramankutty et al., 2006). Land-use intensification is directly connected with socioeconomic processes and sociometabolic flows, thus “natural science based approaches are not sufficient to conceptualizing, quantifying and understanding land-use intensity and intensification processes” (Erb, 2012, p. 10) as most processes are not detectable by RS.

The methodology considers the detection of changes regarding the introduction of new production methods, concentration of land, increment of inputs, change of crop types, and/or the introduction of mechanization and irrigation rather than of land-classes changes, since it is considered that the land deals being dealt imply fewer changes in classes (any class to agriculture) but changes in intensities.

The RS analysis comprises of two different methods that are later used for spatial comparison and, together with the analysis of land deals, identifying major regional hotspots of changes. The first is a 30-year period of Landsat imagery using the Normalized Difference Moisture Index (NDMI) from images with 30-m resolution; and the second, a 20-year time-series of Moderate Resolution Imaging Spectroradiometer (MODIS) imagery using the Enhanced Vegetation Index (EVI) band with 250-m resolution. The described workflow is presented in Figure 3 and in the following subsections.

Figure 3. Workflow indicating the main steps of this research, indicating by each method the data preparation, the spatial and temporal analysis, and the where the results were integrated.

Land deals database

The data on the land deals impacting the region were collected and compiled from the two major and updated open databases of land conflicts, Land Matrix (https://landmatrix.org/) and Environmental Justice Atlas (https://ejatlas.org/). The data extracted from these databases were complemented and cross-checked with reports from local and international organizations. The reports included are as follows: Land Grabbers of the Nacala Corridor (Grain & UNAC, 2015), Lords of the Land (Matavel et al., 2011), The Progress of Forest Plantations on the Farmers Territories in the Nacala Corridor: the case of Green Resources Mozambique (Calengo, 2016), Land Grab for Pulp: New mill project by Portucel Mozambique (Baffoni & Haggith, 2017), Oakland Institute Land Deals Report (Mousseau & Mittal, 2011).

The data was compiled by selecting the districts that have a direct allocation of infrastructure from the Nacala Corridor or that are within the scope of the ProSavana project. To avoid supranational analysis or the coincidence with other major scale development projects currently being implemented in the provinces of Cabo Delgado and Tete, the analysis was restricted to the Provinces of Nampula, Zambezia, and Niassa.

The compiled data table comprised 177 entries with information regarding the size of the land deal, the year when the deal was signed, when the operation started, the type of crops, and the location. The Land Matrix database differentiates the size of land deals into intended, under contract, and size in operation. The reported areas in operation were considered as a metric, however, when these data were not available, other categories were also used.

Much of the data were incomplete or lack spatial precision, 27 entries did not include the precise district where the conflict was reported; and for eight entries, the province was undetermined. Land deals are commonly dispersed in several districts and reports present aggregated data. Nevertheless, the data provided 86 entries regarding conflicts in the provinces within the study region impacted by the Nacala Corridor and the ProSavana program.

Abrupt NDMI land cover changes detection

A 30-year land change analysis was conducted using Landsat (LS) 5 and 7 imagery with 30 m spatial resolution together with the LandTrendr (Landsat-based detection of Trends in Disturbance and Recovery) algorithm in the Google Earth Engine. The algorithm was developed by Kennedy et al. (2010) to detect abrupt pixel-based changes in LS time-series by temporal segmentation. Representing pixel spectral trajectory values through line segments, disturbances and variances are identified as change events within timeframes. Besides abrupt changes within the timeseries, the results also include the year and magnitude of change, enabling the identification of specific timeframe of LUC changes and a horizontal comparison.

In the present case, the parameter analysed was the magnitude’s increase in NDMI levels in the dry season (from April to October) between the years 1994 and 2020. The NDMI index was applied because the types of crops implemented by companies involved in land deals were mainly cash crops with high utilization of inputs and irrigation systems, expressing a more distinguished wavelength spectrum, especially during the dry season. Studies have confirmed the effectiveness of the index for both forest disturbances (Jin & Sader, 2005) and irrigation detection (Chance et al., 2018).

The selected period, from 1994 to 2020, encompasses the new constitution and the growing efforts of economic liberalization, culminating in the concession of the Nacala Corridor to private international companies, the launch of the ProSavana program and the period of resistance against the project in more recent years (a summary of these events is presented below in Figure 10).

The results were later processed using QGIS software. A filter was applied to harmonize and decrease noises from non-representative areas along the time series. The isolated band with change detection pixels by years was aggregated by districts and the area in pixels was converted into hectares for the quantitative comparative analysis. Finally, a heatmap was created using the points of detected abrupt changes, which served as a basis to identify the hotspot areas of land cover changes.

EVI time series

The second detection method applied an EVI index time series from MODIS Surface Reflectance images with 250-m resolution. Single images were pre-processed and extracted from the Google Earth Engine with the mean pixel values calculated for each year during the dry season from the year 2000 to 2020. The images were stacked in a multidimensional single-band image collection and further analyses were conducted on ArcGIS Pro software.

This research opted to use Enhanced Vegetation Index (EVI) following studies on improved indexes for vegetation and crop classification (Hu et al., 2017; Ji & Peters, 2007; Jiang et al., 2008) considering the biophysics characteristics, the sudden densification of vegetation and the specific seasonal component.

An absolute pixel value change detection was done by subtracting the values from extreme sides of the raster dataset (values from 2000 to 2020). An increase of at least two standard deviation were considered as a threshold and selected as a unique class of pixel changes. The areas where detected changes were concentrated were highlighted by a density map and used for later comparison in both other methods to establish the regional hotspots of land cover changes.

These hotspots were used to identify temporal modification patterns in five samples from features that coincide with the operational concept of land grabbing, as shown in Figure 6 in the Results section.

Results

Data from reported land conflicts

Nationally, 150 cases of land conflicts were reported, with most of the reports concentrated in the provinces of Zambezia and Nampula, with 44 and 27 conflicts, respectively. Including Niassa, the three provinces impacted by the ProSavana program made up 58% of the cases of reported land conflicts in the country.

At the province level, Niassa and Zambezia registered reported cases of land deals involving timber exploitation. In Niassa, there were no occurrences regarding food crops, but were occurrences in cases regarding timber and non-food crops. Nampula and Zambezia have the highest concentration of deals aimed at food crop production (Figure 4).

Figure 4. Reported crop types by districts and coloured grouped by provinces. Source: author’s compilation of open databases, 2023.

At the district level, most of the reported cases are concentrated in Lichinga, Gurue, and Monapo. Monapo with most cases regarding food crops, Gurue with food, non-food, and timber in high proportion, and Lichinga with reported cases of non-food and timber. Other districts have fewer cases, with instances regarding timber plantations prevailing.

Spatially, the occurrences of land conflict cases reported were concentrated in specific districts. Gurué and Monapo had the most occurrences with over 10 cases reported, followed by Lichinga/Chimbonila with five reported cases. Other districts with fewer cases were situated around those previously mentioned (Figure 5).

Figure 5. Location of land conflict occurrences aggregated by districts. Spatial reference: Tete, UTM Zone 37S. Base map source: ESRI, Maxar, Earthstar Geographics, and the GIS user community.

The temporal analysis of reported land conflicts was concentrated between 2008 and 2017, with peaks in 2008, 2012 and 2014. The Niassa province presented fewer cases overall but showed constant reports starting in 2008. In Nampula, cases were registered especially in 2008, 2010, with a remarkably high register of seven cases in 2014. Districts with the highest number of reported cases are Sanga, Monapo, and Namarroi in 2008; Gurue in 2009,2012, and 2016; Lichinga in 2012; and Lalaua and Ribaue in 2014 (see Figure 6).

Figure 6. Reported and detected hectares by district based on the NDMI abrupt change detection and reported cases.

LandTrendr algorithm for abrupt land cover change detection

The data extracted from the LandTrendr algorithm presented three major concentrations of results. One before the year 2000, another in 2005/06, and the most representative between 2010 and 2018 (Figure 6).

An outstanding concentration of change detection occurred in the district of Chimbonila/Lichanga in 2014. The region is the location of the final stop of the Cuamba-Lichinga railway and part of a less-developed road connection to the province of Cabo Delgado.

Sanga, Namarroi, and Lago had the most detected pixels in the second decade of the 2000s, which coincides with registered conflicts for timber plantations. Chimbonila/Lichinga registered timber, food, and non-food crops while Ribaue and Monapo registered conflicts regarding the implementation of food crops and biofuels. The latter had its change detection peaked in the years 2005 and 2010.

Figure 7 shows the detected changes in land cover using the method of Landtrendr with NDMI (a) and EVI time-series (b), from the results three hotspots’ regions with coincidence of land cover changes were extracted for sample analysis.

Figure 7. Location of hotspots detected that coincide in three main areas of analysis: (a) Detections resulted from LandTrendr algorithm with NDMI extracted from landsat imagery; (b) Detections resulted from MODIS EVI time-series subtraction. Spatial reference: Tete, UTM Zone 37S. Base map source: ESRI, Maxar, Earthstar Geographics, and the GIS user community.

Hot spot change detection with EVI time series analysis

The hotspot change detection method resulted in a similar pattern as the one using the previous method. Both methods have concentrated hotspots in the same locations with some small disperse variances in both cases. The area of analysis was then divided into three hotspot-areas, named by the province where most detections were registered.

The hotspot-area Nampula is located in the eastern part of the area of analysis mainly in the districts of Monapo and Mossuril, where established and denser settlements and the harbor are located. The hotspot-area Zambezia is in the southern region mainly in the districts of Zambezia Province – Gurue, Ile, and Namarroi. The third hotspot-area in the Niassa province is located in the northern part of the area of analysis, around the capital Lichinga, in the districts of Chimbolina, and the southern part of Muembe, Sanga, and Lago districts.

Two areas with detected changes were not considered in the analysis. Although pixel changes were identified in the northern region, the results were presented in a very dispersed pattern and not coincide with the LandTrendr results, and were thus excluded from this study. Additionally, although the western borders of Mecanhelas have shown a concentration of changed pixels, this region was not considered since the detected pixels were located inside water bodies, which could not be characterized as a human-driven land modification.

Every sampled location within these three hotspots-areas presented coherent patterns of intensification of EVI reflectance (see Figure 8). The hotspot Nampula presented a flatter variation with a slightly increasing trend of EVI values. There were no abrupt changes in land cover but a continuous increment of EVI reflectance value. One exception is the sample collected in the western region of Monapo, where there was an abrupt change between 2008 and 2010.

Figure 8. Variation of EVI levels in sample areas detected in each hotspot.

In Zambezia province, the sampled locations showed a pattern of a steep increase in reflectance between 2015 and 2017. There were two exceptions, which are the ones collected in the district of Gurue. One sample showed an overall high reflectance value oscillating with the lowest values detect in 2009 and 2017. Another exception is also from Gurue with a steady increase of reflectance values with no abrupt changes.

The Niassa province showed a similar pattern across all samples. A steady value was depicted until major changes occurred between 2013 and 2016 presenting a strong increment in values.

Discussion

The results show that the temporal analysis of RS data is in line with the major political, infrastructural, and economic dynamics of the region. Moreover, the general tendencies of the detected spatial changes by both methods coincide spatially and, to a lesser degree, temporally with reports published concerning land rights violations and land grabbing. The spatial precision, however, was hindered by a lack of spatially and temporally disaggregated information on land concessions and operations.

Data regarding the size of land deals are often unreliable or not fully disclosed. Official and government sources present significant gaps between the operations and official reports (Bey & Meyfroidt, 2021). A study from the World Bank reports that from 2004 to 2009 up to 2.7 million hectares were allocated to investors but close to 50% were not fully used (Matavel et al., 2011). Matavel et al. (2011) present several similar cases regarding timber plantations in which the actual planted areas were below 10% of the government concessions.

The results confirm this discrepancy, as the overall reported land grabbed was much higher than the detected LUC change. Only in one district, Alto Molocue, was there an agreement in these figures, namely, 200 ha reported, and 227 ha detected. This overall result can be due to the detection methods that overlooked LUC changes and/or the unprecise information from the land grabbing databases. Also, as pointed out before, some reports may have displayed the amount of land conceded or negotiated and not the amount under operation. The present research targeted LUC changes at a regional scale; however, further research should be done to investigate methods specifically aimed at calculating areas and border detection to determine the location and amount of land more precisely.

The data show that the most prominent changes in the landscape started in the middle of the first decade of the 2000s, reaching the highest rates in the middle of the second decade. This trend is confirmed by both the reported cases of land deals and by the detection of land cover modifications utilizing both RS methods.

In general, though the ProSavana master plan states the encouragement of production diversification (MINAG, Fundação Getulio Vargas, Oriental Consultants Co Ltd., & NTC International Co Ltd, 2013), results showed the concentration of land with similar temporal reflectance patterns, indicating a concomitant implementation of LUC operations. As reports present few types of crops and the introduction of industrial and monoculture systems, it is plausible to infer that the images present the emergence of a less diverse landscape.

Although commencing in the middle of the 1990s, the institutional arrangement in the region towards an export-oriented environment, was fully established in early 2000 with the formalization of the partnership and partial privatization of the state-owned railway company leading to infrastructural changes such as the renewing of the Nacala Port, the paving and renewing of main roads, and rebuilding train stations along the logistic corridor. This process coincides with a steep increase in FDIs that started in 2005 and peaked in 2013, on which, from 2010 to 2013, the growth in FDIs reached over 530%.⁴ Also, during the same period, global commodity prices reached unprecedented high levels (Figure 9). As pointed out by Lambin et al. (2001) socio-economic changes, such as in this case the fluctuation of commodity prices, are the underlining causes of LUC changes.

Figure 9. Commodity prices time series. Scale: Index units. Source: Primary Commodity Price System (PCPS) from http://data.imf.org.

The inflow of capital to the Nacala corridor, as for the rearrangement of international capital flows during the financial crisis (Harvey, 2011), was directed to unleash the productive potential of the region, taking advantage of the high prices in the main global commodities – coal, metal, food, and non-food. This coincidence – the increase in commodity prices and land grabbing – confirms what was globally observed by authors since the 2007/08 food crisis (Edelman et al., 2013).

Every spatial change is interlinked with multiple levels of institutional and technological changes that have a direct and indirect impact on local decisions (Tilzey, 2018). In this sense, the Nacala corridor cannot be seen as separated from several other plans and projects being implemented by the national government targeting poverty reduction and agricultural development (Baloi, 2018) and the multilateral bodies. One of them is the establishment of the African Union Development Agency (AUDA) and the implementation of development projects such as the Southern Africa Hub Port and Rail Program, and the Program for Infrastructure Development in Africa – PIDA,⁵ of which the Nacala corridor is part of, together with many others spatial development initiatives. Within this international and regional context, it is important to place the ProSavana program as a key axis (Kato, 2019), which reinforces the idea that changes in land use frequently follow the integration of a region into the global economy (Lambin et al., 2001).

Figure 10 shows a summary of the main occurrences in the Nacala Corridor, and global and national events that impacted the implementation of the project. This timeline suggests a coherent alignment of processes in multiple scales that induce the intensification of land uses for export-oriented commodity production.

Figure 10. Timeline of most impactful events on regional, national, and global levels.

The institutional setup was also shaped by efforts to implement the ProSavana project. The cooperation program impacted the 19 districts established in the master plan unevenly. The described aim was to implement competitive agriculture with highly productive inputs and techniques. Therefore, the program also indicated the improvement of the offer to better access to inputs and strengthen financial and extension schemes, nonetheless, it relied mainly on the introduction of productive infrastructure (MASA, 2015). Although the railway line has a clear starting operation time, road improvements cannot be clearly defined, in the sense that a semi-finished road already improves connectivity and can be a pushing factor for land-use changes.

The results showed that the timeframe of new infrastructure implementation is aligned with the major landscape changes detected. In the districts analysed in Zambezia, major changes were detected between 2015 and 2017. These years coincide with the conclusion of the rehabilitation of the railway stretch Cuamba-Nancala (2015). This area, especially the districts of Gurue and Ribaue, is a long-time agricultural hub for both food and non-food commodities. This explains the divergent results of the sample extracted in Gurue, which does not present an abrupt change but a gradual increment of seasonal variations from a productive area. Nonetheless, all other samples presented a similar pattern of abrupt increment from 2016 to 2018.

Similarly, the stretch connecting Lichinga to the main railway network was concluded in 2016. The results, however, showed that the first detections of changes occurred before its conclusion, around the years 2013–2015, meaning that investments for land use modifications that resulted in LUC modifications started before the completion of the infrastructure. A report describes that one company’s first investment in the Lichinga plateau started in 2005 by requesting 140,000 ha; however, in 2015, only 17,000 ha were under operation with planted eucalyptus (Calengo, 2016). Kronenburg García et al. (2022) thoroughly narrate the historical process of investment waves in the region indicating a decline of investments starting in 2014. The results, however, show that there are still concentrated spots of land cover changes impacting the region.

Thus, activities that resulted in LUC changes relied first on the institutional setup trusting in the future implementation of a proper physical environment, instead of on the operationalization of the infrastructure. In this case, although the literature agrees that infrastructure is an important driving factor of LUC changes (Lambin et al., 2003; Zheng et al., 2021), the fact that policies were set up preparing the implementation of roads and rail renewal indicates that an institutional framework played a more important role. The results are in accordance with Harvey (2011), which affirms that relations between capital, labour, and nature, are mediated by both implementations of technologies and organizational forms. The Nacala corridor enabled an innovative setup in both ways to provide the means for spatial fixes.

Conclusions

In conclusion, the analysis of LUC at a regional scale along the Nacala Corridor confirms the temporal trend of LUC modifications following the implementation of infrastructure and economic liberalization policies. LUC changes and modifications do not precisely reflect the implementation dates of SDI but follow trends of policies and institutional arrangements that guarantee a safer environment for investments.

The combination of data from local reports with remote-sensing data indicated that an increased number of LUC disturbances in the landscape occurred during the first years of implementation of the Nacala corridor. The wave of reported cases, however, started before the detected changes – some years prior to the publicization of the ProSavana program. This study supports the scholarship that reinforces the multidisciplinarity of land systems, and the need to methodologically address the complexity of socio-ecological systems (Barbier, 2000; Erb, 2012; Lambin & Geist, 2006). Also, built up Sassen’s (2013) argument of the national territory being subject of non-national systems of regulation and interest, in our case, infrastructure implementation for extractivist industry leading to massive land investments.

Several issues are combined factors that could have led to the modification of the landscape along the corridor – local projects, individual initiatives, and push-pulling factors from other regions, among many others. Nevertheless, since the corridor is the biggest and most ambitious SDI currently under implementation in the country, the major LUC modifications located in the same timeframe can be, with a certain degree of certainty, directly connected to international massive land deals.

Regarding the question of whether the detected LUC changes were implemented through land grabbing. On a regional level, temporal and spatial data are in agreement, confirming the tendency pointed out by the selected reports, meaning that the RS results spatially concurs with documented cases, with majority of changes in locations where more conflicts were reported. However, the data did not coincide with the number of hectares under land deals. A limitation of this type of analysis is that the numbers reported by The Land Matrix and other organizations might be based on unprecise sources, sometimes biased by what involved parties decide to publicize, leased areas not put into operation, or data from the government which must be taken cautiously as real figures.

The time-series analysis, however, presented a clear temporal indication of LUC modifications; which, if combined with a case-by-case analysis, has the potential to indicate the implementation of new land operations; and therefore, reported land-grabbing cases. Finally, this research demonstrated the possibility to detect spatial trends and patterns that are aligned with major political-economic changes and, in combination with multiple sources of information, indicate regions where land grabbing and land conflicts occur.

Acknowledgments

We acknowledge support by the DFG Open Access Publication Funds of the Ruhr-Universität Bochum

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data regarding land conflicts were extracted from Land Matrix (https://landmatrix.org/) and Environmental Justice Atlas (https://ejatlas.org/), other sources were mentioned in the body of the text. Every image collection was accessed via open-sourced databases. Specific data used in the research as well as partial results can be shared upon request.

Additional information

Funding

This research was developed with funds from the DAAD Graduate School Scholarship Program.

Notes

¹ https://landmatrix.org/

² https://ejatlas.org/

³ For a comprehensive description of the corridor trade project, see World Bank (2018)

⁴ Data from World Bank database: https://data.worldbank.org/indicator/BX.KLT.DINV.CD.WD?end=2020&locations=MZ&start=1992&view=chart

⁵ List of projects from AUDA-NEPAD in https://www.nepad.org/countries/mozambique