Introduction

1Urbanisation dynamics are a major cause of changes in land use. Globally, they occur mainly in countries in the South: between 1990 and 2015, 12.8 million ha were urbanised in Asia, Africa, and Latin America, as against 4.6 million in North America, Europe, and Oceania (Denis, 2020). They result in the building of housing and commercial zones, contributing to both urban sprawl and urban de-densification (Denis, 2020).

2Many works have explored urbanisation dynamics in various cities around the world. Results based on quantitative and spatial approaches have unsurprisingly shown that built-up areas increase at the expense of farmland, but also highlighted an increase in degraded land and wasteland (Follmann et al., 2018; Hou et al., 2016; Morshed & Zhang, 2017; Oda et al., 2018; Deribew, 2020). Researchers have focused on the forms of urbanisation and the factors driving it (Deribew, 2020; Morshed & Zhang, 2017; Oda et al., 2018; Newman & Kenworthy, 1996). Only a few have provided quantitative analysis of the types of agricultural recomposition associated with urban growth (Mesclier et al., 2014; Robineau et al., 2014; Bon et al., 2023). Finally, few works have provided qualitative and quantitative analysis of the recomposition of landscapes engendered by urban growth on built-up areas and farmland.

3Antananarivo, the capital of Madagascar, is a city of alternating hills and floodplains located on the central uplands. Since the seventeenth century, housing has been built around the historic villages on the main hills in the landscape. These settlements were placed on hilltops both to preserve the farmland in the plains, and to avoid risks of flooding, which is frequent in this territory hemmed in by three rivers and lying downstream of a large watershed (Esoavelomandrosoa-Rajaonah, 1989). During the colonial period and the first republics, the centre was urbanised in planned manner, under state oversight, and partly encroached on the plains (Defrise & Burnod, 2023). In more recent periods, urbanisation has intensified in the centre and surrounding communes, and in parallel along the hills and plains. Urbanisation forms are influenced by state policies funded by donors (road infrastructure projects), but mainly result from spontaneous dynamics by private operators and households (housing and commercial buildings) (op. cit.).

4As things stand, urbanisation is continuing to progress in this agglomeration of over 2.5 million inhabitants in response to the population’s needs for housing, mobility, and economic activities. Between 2003 and 2017, the urban footprint rose from 5,800 ha to 9,000 ha, with of 200 new buildings per year on average, and an annual growth rate of the built-up area of 3.2% (Defrise, 2020). The densification of built-up areas on the hills and the development of road infrastructure led to increasingly marked urbanisation on the plains, despite a legal ban on building there (Ranaivoarimanana, 2017; Defrise, 2020). The urbanisation of the plains had a negative impact on agricultural production conditions and on the profitability of farming (Defrise et al., 2019). At the same time, it increased the risks of flooding (PUDi, 2019). Urbanisation has taken place at the expense of farmland but, concomitantly, land under cultivation has expanded into hilly areas at the expense of savanna areas (Defrise et al., 2019). Agriculture is of key importance to the population. In 2017, cultivated land still made up 45% of the agglomeration’s surface area (Dupuy et al., 2020) and continued to provide crucial functions for the city, in terms of food, jobs, and above all protection against flooding (Aubry et al., 2012; Defrise, 2020).

5Against this backdrop, the purpose of this article is to better quantify, describe, and spatialise the urban dynamics and agricultural recomposition taking place across the Antananarivo agglomeration. Particular attention is paid to dynamics taking place on the plains, where urbanisation is progressing and of key importance for flood management.

6The following section presents this study’s quantitative and qualitative approach to the Antananarivo agglomeration. Section 3 updates various land uses based on a map produced in 2022. It then analyses the progression of urban space and agricultural recomposition conjointly over the period 2017-2022. It emphasises the scale of changes to farmland in this urbanisation process, particularly in the plains, as well as noting the diversity of land-use transitions and highlighting the flexibility of farming activities. Section 4 examines the factors encouraging or hindering urbanisation and the preservation of the plains for agriculture. The final section concludes on the importance of providing input for public debates on land-use planning to allow the roles played by agriculture to be taken into account in decisions about the city’s future.

1. Methodology

7Analysis is conducted both at the scale of the agglomeration of Antananarivo or Greater Tana (covering over 70,000 ha), and that of three smaller studies zones composed primarily of plains prone to flooding (ranging from 180 to 220 ha in size) (map 1). This provides a way of quantifying and specialising the differential urbanisation dynamics and agricultural recomposition taking place across Greater Tana, with the focus on the plains allowing us to explain the main factors at work.

8Greater Tana comprises 38 communes, including the Commune Urbaine d’Antananarivo (CUA) at its centre (map 1). Its total surface area is 76,000 ha, but for reasons of comparison using earlier maps, only 70,000 ha will be examined here. Greater Tana is composed of alternating hills and plains and is hemmed in by three rivers: the Ikopa and its two tributaries (the Sisaony and the Mamba) (map 1). Two of the studies zones are in the immediate periphery (Zones I and II), while the third (Zone III) lies right in the centre (in the CUA) (map 1). These zones were chosen as part of doctoral research by one of the authors, linked to a research project into resilience to flooding (Future Cities Lab – Antananarivo, FCL). The zones are located within the study area of the FCL project. They were selected because of their differential urbanisation dynamics, their vulnerability to flooding, and the presence of plains with varying proportions of farming.

9Changes in land usage refer to qualitative change (in ground type or usage) and quantitative change (an increase or decrease in an area dedicated to a specific usage or type of cover) over a given time interval (Briassoulis, 2020).

10In the wake of work by Dupuy and Defrise (Dupuy et al., 2020), our research combines image analysis and interviews. Changes to the built environment and to the categorisation and spatialisation of farmland are analysed using satellite images. Only images taken from 2003 onwards (Spot five) are used, as images captured as of the 1970s (Landsat) are not sufficiently precise for built-up areas. The period of study thus runs from 2003 to 2022.

11Analysis of the types of land use impacted by urbanisation is based on very high-resolution satellite imaging (50 cm – Pléiades) and reference data collected in the field (GPS point surveys, identification of cover, and digitisation). Data availability constrains the period analysed from 2017 to 2022. The same applies to agricultural changes. Analysis is based more specifically on two land-use maps made respectively in 2017 and 2022 using the MORINGA processing chain (Gaetano et al., 2019). For the 2022 land-use map, two data categories were used. The first included two types of satellite imaging: Pléiades imagery coverage of the agglomeration conducted on 3 April 2020 using Dinamis, and a time series of Sentinel-2 images (using 10m and 20m spatial resolution bands) from July 2021 to July 2022 (amounting to 110 images spread over 2 tiles: 38KQD and 38KQE). The second type is a reference database for 2022, obtained by updating the one produced in 2017. To detect the types of land use impacted by urbanisation, level III nomenclature was used (table 1, map 2). This was applied to the two land-use maps under analysis, providing the best level of results (with overall accuracy of 89.22 and a Kappa coefficient of 0.87). The various cover categories used are described in the following section.

Table 1: Nomenclature used for the 2022 land-use map.

Source: authors.

12Five categories were considered: i) built-up areas including residential, commercial, and industrial buildings, airports, car parks, and roads, ii) artificial surfaces concerning quarries (for stone and brick), embankments, building sites, landfill, and sports fields, iii) natural spaces including herbaceous, shrubby, and wooded savanna, iv) bodies of water (rivers, canals, marshland, lakes, and fishponds), and v) farmland, including all cultivated plots.

13To identify the common factors influencing urbanisation and agricultural recomposition, 170 stakeholders were interviewed in Greater Tana and the territories under study. These people were directly involved either in converting land into buildings (126 households and 10 companies) or in land management (26 local authorities, 6 institutions, and 2 people in charge of opening the sluices regulating water flow between the plains and rivers). Their discourse was analysed, and maps produced to cross-reference the nature and spatial impact of various factors in the study zones.

2. Urbanisation and the recomposition of agricultural areas

2.1. A highly agricultural agglomeration

14In 2022 Greater Tana was still highly agricultural: farmland covered 40% of the agglomeration (28,203 ha, cf. map 2). The vast cultivated areas, often on the plains, are mainly located to the north, north-west, and south-west of the agglomeration. Wetlands, used mainly for market gardening and rice and watercress growing, account for 24% of the agglomeration’s surface area and over half of the area under cultivation (16,750 ha) (map 2). Market gardening, traditionally located on lower slopes, is also to be found on the plains. Rainfed crops grown on hillsides include cassava, maize, and legumes. Fruit crops are found on lower slopes, mainly orange groves. Although not captured by satellite imagery analysis, pig, dairy, and poultry farming are very present in the agglomeration and are developing rapidly (Defrise, 2020; Orbell et al., 2023).

15Natural areas cover one third of the agglomeration (22,439 ha). These are composed of herbaceous and shrubby savanna and are found on the many hills marking the agglomeration’s landscape, increasing in extent as one moves away from the town centre; they are used for cattle grazing. Wooded savanna and forest plantations (10,867 ha and 15% of the urban area) predominate on the eastern side of the agglomeration.

16Built-up spaces in fact only cover 17% of the agglomeration (11,728 ha). They are mainly found in the CUA, at the centre of the agglomeration, and stretching along the old and more recent main roads in ribbon developments (map 3).

17Artificial surfaces (brick quarries, embankments, landfill, and building sites) only occupy 6% of the agglomeration (4,550 ha). Identified as precursors to urbanisation in other cities (Follmann et al., 2018; Morshed & Zhang, 2017), in Antananarivo they are mainly located in the north-west and south-west of the agglomeration. Some of them are large in scale, while others are smaller and result from artisanal activities.

18Bodies of water, amounting to 3,191 ha (5%), are mainly found in the western part of the agglomeration. Numerous small ponds and permanent or temporary residual or developed natural marshes are used for fish farming (either by adding fry, or by capturing fish brought in by river flooding).

2.2. Accelerating urbanisation at the expense of farmland

19Built-up areas are mainly for residential and commercial purposes, and to a lesser extent for industrial, military, and administrative uses. Over a period of nearly twenty years (from 2003 to 2022), the built-up area in Greater Tana doubled (table 2, map 3). Its growth accelerated particularly between 2017 and 2022. Over these five years, the surface converted to building nearly equalled that converted over the fifteen previous years (2,903 ha of additional buildings between 2003 and 2017, as against 2,377 ha between 2017 and 2022). The annual growth rate of the built-up area went from 3.2% between 2003 and 2017, to 5.08% between 2017 and 2022. It is thus close to figures observed in other African capitals (averaging 5.37% for Bamako and 4.99% for Nairobi over the period 2000-2014, Hou et al., 2016).

20Urbanisation occurred mainly by converting farmland (figure 1 and table 3), largely explaining the drop in the area under cultivation. Across Greater Tana as a whole, between 2017 and 2022, over half the newly built-up areas (55%) used to be farmland. More specifically, built-up areas developed mainly in wetlands used for rice growing during the season and for market gardening and fish farming out of season (27% of the surface area), followed by land at the bottom of slopes used for market gardening (15%), and finally hillsides (referred to locally as tanety, used for rainfed crops (12%) (figure 1). Built-up areas also spread onto hillsides composed of savanna and grazing land (27%) and into areas which already had artificial surfaces (18%).

21Urbanisation pressure, although limited in absolute terms, appears to be greater on farmland than other types of land use (4% of total farmland was converted, against only 2% of natural spaces and pasture, and 1% of artificial surfaces) (table 2).

Table 2: Land by type in 2017 converted to built-up use by 2022 and the conversion rate.

Land use	Surface area in 2017 (ha)	Share of Greater Tana surface area in 2017 (%)	Areas converted to built-up use between 2017 and 2022 (ha)	Conversion rate over the period (%)
Artificial surfaces	4,318	6	412	1
Natural areas	23,254	33	640	2
Bodies of water	1,662	2	8	0
Farmland	31,526	45	1,317	4
Total surface area converted to built-up use between 2017 and 2022 (ha)			2,377

Source: authors.

22In many agglomerations the urbanisation of farmland also involves intermediary phases when it lies fallow, during which the land is left “latent” or else used for brick production. Wastelands indicate both agricultural decline and new areas of economic interest (property transactions and speculation, etc.), becoming built-up areas over the medium or long term (Bon et al., 2023; Deribew, 2020; Follmann et al., 2018; Hou et al., 2016; Morshed & Zhang, 2017). In the case of Greater Tana, artificial surfaces include stone quarries, landfill, building sites (levelling on hillsides), embankments (in wetlands), brick quarries, and bare land (figure 2). Contrary to observations in other cities (Follman et al., 2018; Hou et al., 2016), two points stand out for Antananarivo. First, there is rarely any agricultural wasteland: all farmable land is cultivated, indicating a strong need for farmland, not decline. Second, the transformation of farmland into brick quarries is not an irreversible process and does not lead in linear fashion to urbanisation. Lowlands are used for brick production, but are also reused for rice growing. It is only after several cycles of alternating between rice growing and brick production that plots cease to be cultivated and buildings are erected (Aubry et al., 2012; Brouillet et al., forthcoming). This land take also occurs at the expense of farmland and, by order of decreasing importance, on wetlands, market gardens, and rainfed cropland (respectively 629 ha, 136 ha, and 123 ha).

Table 3: Changes to built-up areas in Greater Tana between 2003 and 2022, and changes to other spaces between 2017 and 2022.

	Surface area (ha)			Variation (ha)		% of the territory as a whole
	2003	2017	2022	2003-2022	2017-2022
Built-up areas	6,447	9,351	11,728	5 281	2,377	17
Artificial surfaces	nd	4,318	4,550	nd	232	6
Natural areas	nd	23,254	22,439	nd	-815	32
Bodies of water	nd	1,662	3,191	nd	1,529	5
Farmland	nd	31,526	28,203	nd	-3,323	40
Total surface area of Greater Tana (ha)			70,111	Total		100

Source: authors.

2.3. Recomposition of farmland

23Although farmland still accounted for a sizeable area in 2022, it dropped by 3,323 ha between 2017 and 2022 (table 3, table 4). Irrespective of this decrease in surface area under cultivation, the types of farming evolved. Lowland farming (rice farming and market gardening) dropped by 5,132 ha and 531 ha respectively, while hillside agriculture was up 1,074 ha for rainfed crops (maize, cassava, legumes, etc.), up 936 ha for fruit growing, with fallow land up 330 ha (table 4). Greater Tana is experiencing an original, synchronous development on its hillsides where urbanisation is occurring alongside land being newly put under cultivation (Defrise et al., 2019). This recomposition of farmed areas is because households continue to farm, and their number is on the rise (op. cit.).

Table 4: Changes to farmland in Greater Tana by types of crop, between 2017 and 2022.

Land use	Surface area (ha)		Change between 2017 and 2022 (ha)
	In 2017	In 2022
Wetland farming	21,882	16,750	-5,132
Market gardening	4,813	4,282	-531
Rainfed crops	4,218	5,292	+1,074
Fallow	100	430	+330
Fruit growing	513	1,449	+936
Total (ha)	30,929	28,203	-3,323

Source: authors.

2.4. Varied land-use transitions

24The case of Greater Tana highlights varying types of transition occurring in tandem to urbanisation (figure 1, figure 2). First, the recomposition of farmland, and second, the artificialisation of natural and agricultural spaces. Rice-growing areas on the plains have been transformed into market gardens (1,677 ha), bodies of water converted to fish farming (1 317 ha), and land put to use for fruit growing (350 ha). Plots previously used for market gardening have been converted to fruit growing (290 ha), and savanna brought under cultivation for rainfed crops (684 ha). Farmland (765 ha) and natural spaces (123 ha) have been turned into quarries (figure 2).

3. Factors jointly influencing urbanisation and agricultural recomposition

25Various factors may stimulate or hinder urbanisation, and, when combined, may cause farmland to be preserved. The focus here is on agricultural plains, the farmland most affected by urbanisation.

3.1. Demographic growth, housing and income needs

26Urbanisation is linked to demographic growth: each year, Greater Tana takes in between 100,000 and 150,000 new residents (World Bank, 2017). This demographic growth is caused both by the annual population increase and by migration (op. cit.). It increases the population’s need for housing and jobs.

27To meet housing needs, various stakeholders invest in the built environment. As in other cities in the South (Denis, 2016), households in Greater Tana are the main drivers of urbanisation through dense building on micro-plots. This accounts for 85% of the built-up areas (our topographic analysis). They construct buildings on small plots (of between 32m² and 290m²) for commercial or residential purposes (for themselves and for tenants), or for mixed usage (residential and commercial). Private companies and the state are also involved in this urbanisation process by constructing buildings for commercial, administrative, industrial, and residential uses on larger plots (of between 400m² and 40,000m²).

28Concerning jobs, the secondary and tertiary sectors are relatively underdeveloped (PUDi, 2019). Households thus continue to farm as a source of income (Aubry et al., 2012). In the absence of a recent census, it has been estimated that 20% of households in the agglomeration have an agricultural activity (Defrise et al., 2019). Agriculture is a major safety net in this city marked by high unemployment and rising impoverishment (Defrise, 2020).

3.2. Risks of flooding and water management

3.2.1. Flood risks

29The agricultural plains, in addition to being in low-lying areas (map 4), tend to be close to watercourses and sometimes lie between two rivers (map 1). When there is heavy rainfall, a characteristic of the wet season in this part of the country, the agricultural plains receive river water as well as rainwater from nearby upstream watersheds. They naturally act as a natural buffer basin, but are also used as spillways by hydraulic infrastructure (dikes and sluices) and by hydro-agricultural facilities (irrigation channels). In study zones I and II, the agricultural plains are surrounded by dikes; each year, via a system of opening and closing sluices, they receive water from the River Sisaony to prevent it compounding flooding from the River Ikopa and causing further inundation in the CUA. Water levels can rise rapidly and significantly (the level of the Ikopa can rise 0.23m in three hours at Anosizato in zone II). These plains may receive additional unforeseen amounts of water in the event of dike breaches, which occur frequently given the lack of maintenance (in 2015 and in 2017 in zone 1, for example).

30These very real risks of flooding (given the frequency and height of water) have severely restricted urbanisation in the most exposed areas. They influence farming conditions (given the risk of crop loss), but farmers also adapt their plots and cropping systems accordingly.

3.2.2. Water management and farming conditions

31Flood risks severely limit urbanisation in the most exposed areas. Such risks may be better managed in other zones via hydro-agricultural infrastructure and embankments. Farming conditions are thus a decisive factor in keeping land under cultivation or else its conversion into built-up areas.

32Poor farming conditions prompt households to produce bricks, fill in a plot and build on it (figure 2), or, for lack of means, sell it to a builder. These poor conditions result from two water-related problems. The first results from insufficient maintenance of hydro-agricultural infrastructure. Since the 1980s, the state has no longer been able to manage this infrastructure properly due to insufficient state budgets (occasionally bailed out by development aid), lengthy tendering procedures, the incompetence of certain of the companies hired, and so on (plate 1). Water users’ associations, which are meant to help with maintenance, are similarly unsuccessful. In the rainy season, water availability outstrips agricultural needs. Rainwater and river water are discharged onto the plains, which are rapidly flooded. Due to the lack of infrastructure maintenance, the water cannot be drained in timely manner, and thus damages crops (even though rice can withstand flooding for 5 to 7 days).

33The second problem is access to water in the dry season. As a result of climate change and increased competition for water resources due to urban growth, famers on the plain can no longer access water in the right quantity at the right time. In the dry season, water drawn from dam lakes is allocated first and foremost to urban consumption, and only secondarily to irrigating farm produce. Between 2019 and 2022, between one fifth and, at best, half of farmers’ water needs were met.

34In zone III, farmers do not have enough water in the dry season, and cannot use it properly in the rainy season as they are unable to manage it (due to the deterioration of drainage channels, and to water inlets and outlets being blocked by buildings or waste) (plate 1). They turn to brick production (figure 2). After several soil extraction cycles, they sell the land or build on it themselves, provided other favourable factors are present.

35Conversely, good or average agricultural conditions enable producers to carry on farming plots of land, even if they explain in interviews that their children will eventually be able to build on them.

3.2.3. Adaptation of farming systems

36The disappearance of farming in an urban environment (resulting, for instance, from opportunities to sell the land, and from deteriorating growing conditions due to the proliferation of buildings) is not necessarily inevitable. As in Bobo Dioulasso, Saint-Louis, and Manila (Robineau et al., 2014), agriculture in Greater Tana is being reconfigured. As an essential source of income for farming households, it gives rise to innovation, adaptation, and intensification.

37The first such adaptation concerns rice production, with the use of direct seeding and techniques generally used in flood-recession rice farming. Farmers have abandoned transplanting for direct seeding and opted for varieties that are more resistant to flooding (practised in zones I and II). A second adaptation consists in developing a plot to allow both farming and fish farming (in all three zones). Farmers dig out large volumes of land to create raised strips on parts of the plot. On these raised areas, they grow market-garden crops in the dry season and rice in the wet season. These raised gardens are irrigated in the dry season using water drawn from the dug-out areas, and are less constrained by flooding in the wet season (the water only partly covers the young rice plants, rather than fully). The channels formed between the raised beds are fed by floodwater which brings in fish (showing as black stripes on aerial photos of zones I and II, plate 2). These channels may be blocked at either end and turned into temporary fishponds. On occasions, the need for earth for brickmaking or for developing adjacent land leaves deep holes on certain plots (Brouillet et al., forthcoming), hindering their reverting to rice growing or market gardening. Such plots are then converted into permanent fishponds (showing as black spaces on the aerial photos of zone III, plate 2).

3.3. Plot accessibility

38A third key factor encouraging urbanisation at the expense of farming is plot accessibility. Even for parcels prone to flooding or suitable for farming, the proximity of a carriageway or permanent pedestrian access will encourage farmers to embank them and build on them, or else sell up for lack of means. The sale price of accessible plots is far higher than those located further out on the plains (in zone III, a plot far from the road is 6 €/m², while a roadside plot of similar size is 30 €/m²).

39Given the lack of available, affordable land on the hills near the urban centre (the area with services and job opportunities), businesses and households purchase land on the floodplains (in white on the maps in plate 3). For private companies, road access is a key criterion. Households can make do with pedestrian access. Their dense, contiguous roadside buildings can limit car access in certain places, and even pedestrian access back from the road, thus hampering urban expansion. In zones I and II, built-up parts—in pink on plate 3—are to be found only along main roads. In zone III, they extend into low-lying areas via narrow pedestrian corridors (where plank walkways are erected in the event of flooding), resulting pockets of wasteland being urbanised (plate 3).

40The urbanisation of plots is closely linked to their accessibility, but also depends on the need for embankments of varying size and cost depending on the flood risks. Most households cannot undertake such earthworks. The minority who do, do so on the edge of the plain, using material to hand (waste, water hyacinth, leftover brick, or wood chippings). They elevate their houses (made of wood, sheet metal, or brick) and live surrounded by water for several months of the year. The others often end up selling their plot of land, after being approached by intermediaries or private operators’ representatives. The only people who build embankments using massive quantities of earth and motorised equipment are private operators, with the networks required to pay for transporting the earth and for administrative authorisations (despite the legal ban on building).

3.4. Urban development plans and land tenure

3.4.1. Urban development plans and rules

41While legislation and urban development plans influence builders’ strategies, they have not managed to halt the urbanisation of the plains. Since 1926, various plans have been drawn up for the city’s development (Esoavelomandrosoa-Rajaonah, 1989; Ranaivoarimanana, 2017; Defrise, 2020). The latest is the 2019 Plan d’Urbanisme Directeur (PUDi). In this latest plan, the plains are mainly classified as no-build zones (plate 4), with decision-makers hoping to use these areas as a buffer against flooding. These agricultural plains are also the subject of CUA and Greater Tana ministerial orders prohibiting embankments. Comparative analysis of the 2019 PUDi and constructions carried out between 2017 and 2022 shows that these regulations (plan and decrees) have not been respected. In all three zones, built-up areas have advanced into areas classified as no-build. Households have generally built without authorisation, and local authorities have not objected, being aware of the housing needs of these economically fragile populations. The authorities regularise the situation after the event by granting permits, issuing local authorisation, allocating house numbers, and levying taxes, or else by extending buildable zones on the detailed urban development plans. Businesses are compelled by tighter government oversight to obtain the necessary authorisations (before or during construction). They often achieve this through bribery (Ranaivoarimanana, 2017). The only zone where the ban on building was respected is that with the highest risk of flooding (zone I). Potential builders are discouraged by the scale of the embankments needed, while the authorities exert stricter oversight of these heavily flooded areas.

3.4.2. Land tenure status and insecurity

42In the agglomeration of Antananarivo, many landowners do not have up-to-date property deeds or documents that fully secure their rights (Burnod et al., 2020). Obtaining land title is complex (with many stages to complete and documents to provide), expensive (due to the cost of producing documents, land inspections, and corruption), and sometimes perceived as risky (due to the rights-holders’ fear that the authorities will formally attribute their plot to a third party) (Ranaivoarimanana, 2017; Defrise, 2020). In most communes in the agglomeration, it is not possible to obtain a land certificate—a legal document of ownership introduced in 2005 under the decentralised land management system as an alternative to title deeds (Defrise, 2020; Burnod & Bouquet, 2022). Producing legal documents is also made difficult by complex land tenure situations: certain plots are jointly owned by several generations, or else the owner of the land may not be the same as the owner of the building, etc. (Burnod et al., 2020).

43Some owners have up-to-date land deeds, but they are in the minority. It is wealthy companies and well-off households who produce them to obtain building permits or to get connected to water and electricity networks. Other owners may be in possession of legal documents (either title deeds or intermediary documents), but the latter are in the name of former owners (ascendants in the case of inheritance, or the vendor in the case of purchase). In practice, most households produce semiformal documents to prove their rights and retrace the history and subdivision of their plots, such as affidavits drawn up by the family, deeds of sale validated by the Commune, plans drawn up by private surveyors, or tax receipts.

44Many households feel insecure about their plots and fear their rights will be challenged by a third party (neighbours, elites, the land administration, or the state). This has an impact on urbanisation and on preserving land for agricultural purposes. One strategy for securing land is to mark ownership by ensuring the land does not lie bare. Depending on households’ projects and resources, the environment of the plot (the risk of flooding, farming conditions, and accessibility), and, finally, independently of the land tenure documents they hold, households opt to build on land or to keep it for farming (by cultivating it themselves, by sharecropping, or by allowing others to farm it in exchange for warding off squatters) (zone I). In the event of complex land tenure, the most frequent choice is to farm it to avoid any disputes over ownership or the right to build (zone II). For reasons of topography, particularly the frequent risk of flooding, the types of farming selected to mark land ownership are market gardening and rice growing, with trees being less appropriate.

Conclusion

45This article has quantified, spatialised, and explained urbanisation processes in Greater Tana, as well as forms of agricultural recomposition there. Over the past five years, urbanisation has progressed increasingly rapidly, at the expense of farmland, particularly in lowlands. In parallel, agriculture has spread to hillsides and diversified in the plains. The types of transition in land use are many and varied, revealing inhabitants’ needs and constraints. Agriculture may be used to mark ownership (with market gardening and rice growing), or to provide households with sources of income given the lack of alternative employment and the high levels of poverty. It thus takes up much of the plains which thereby to continue to act as a buffer against flooding. Nevertheless, the preservation of agricultural usage depends on the progression of urbanisation. This article has emphasised that the key for effectively regulating urbanisation dynamics—above and beyond devising rules and drawing up urban development plans—is to intervene on hydro-agricultural infrastructure, farming conditions, the location of access roads, and to control the erection of embankments. All these findings may feed into public debate on land-use planning and the place of agriculture in the city’s resilience (in terms of food, jobs, and revenue for vulnerable populations, as well as protection against flooding).