Rainfed farming in semi-arid India is marked by its vulnerability to climate variability, accelerating resource degradation, and asset constrained populations. With climate change poised to exacerbate existing stresses on smallholder farming, there is a need to understand the factors constraining and enabling farmer adaptation. Integrated watershed development has emerged as a policy instrument to encourage and institutionalise sustainable natural resource use, to diversify rural livelihoods, and to build local capacity and propel rural development. Against this backdrop, this study had three main objectives: to examine farmers ' perceptions of water scarcity and climate variability in a semi-arid rainfed region of India and see whether these perceptions are reflected in meteorological records; to understand why some farmers are more vulnerable than others to these stressors; and to examine what strategies farmers undertake in response to perceived risk, and understand the decisionmaking process behind their choice of certain strategies. The study draws from the actorcentric vulnerability framework, which places the human system (here, the farming household) at the centre and explores vulnerability through its three determinants: exposure to a stressor (water scarcity and climate variability), system sensitivity, and adaptive capacity of the system. Situating the fieldwork in Pratapgarh, a predominantly tribal district in the semiarid state of Rajasthan, data was collected over one agricultural year (2011 to 2012) covering the monsoon (kharif) and winter (rabi) seasons. Differential vulnerability was explored between villages, between households within a village, and within households to develop a clear picture of local vulnerability. Towards this, data was collected in two sites: one with a watershed intervention operational for five years and the other with no intervention except for State-run public welfare schemes. A blend of household surveys (semi-structured interviews), focus group discussions, direct observation, open-ended key informant interviews, and in-depth case histories were used to collect data. Farmer narratives demonstrated that households interpret, experience and respond to climatic and non-climatic changes concurrently. The drivers of household vulnerability were an ensemble of highly localised and individual factors (intra-household dynamics and local socio-cultural norms) and macro-scale forces (global market demand, national policies, regional climate variability). These drivers were experienced together to inform farmer response decisions and livelihood strategies. This study found that tribal farmers in Pratapgarh were far removed from the caricature of passive victims of climate change and made proactive and reactive responses to changes in their environment. However, household-level response decisions were constrained by local and cross-scale factors, as well as factors perceived as beyond the decision maker's control. The thesis demonstrates how an understanding of livelihood trajectories and dynamic vulnerability pathways that incorporates views from the vulnerable, can allow agricultural and development policy to incorporate differential vulnerability, especially in the context of increasingly interdependent and multiple-scale drivers.

This research develops a bio-economic modelling framework for the assessment of agricultural and water policy change implications for irrigated agriculture. It aims to contribute towards the implementation of the Water Framework Directive within the European Union, particularly in regard to the principles of cost recovery of water services and water demand management for irrigation.

In Bangladesh flood mitigation strategies have concentrated on embankments. However, due to a perceived lack of success with this strategy and high returns to irrigation development, the emphasis in water management switched towards small scale irrigation. Serious floods in 1987 and 1988 have renewed interest in flood mitigation. However, there is a lack of detailed evidence on the impacts of embankments. This study provides just such evidence. The study comprises a detailed post-evaluation of a major flood control, drainage and irrigation project, the Chandpur Irrigation Project. The results show that the project has been successful in reducing normal monsoon water levels, with consequent changes in cropping patterns to higher yielding varieties. However, a with-without comparison revealed that expected yields (allowing for risks of flooding and drainage problems) were less than predicted due to drainage problems within the project, whereas non-project expected yields are as predicted. Hence agricultural output has increased substantially, but output and its value are less than anticipated. Household incomes in the project are 25&37 higher (on average) than in control areas, but this is because larger landowners have gained from greater returns to agriculture (for example, from preferential access to irrigation), and diversifying into other occupations. Some households have gained by selling land and obtaining salaried jobs elsewhere or by becoming traders. By comparison small landowners may be slightly less prone to losses in flood years but have not moved into surplus. Landless labourers have similar incomes inside and outside the project; real wages are unchanged, and increased work has not compensated for population growth. Inside the project changes in landholding category have been fewer than outside. Open water fish populations have been reduced by the embankment, particularly affecting poor households. However, fish cultivation has expanded more than in flood prone areas, benefiting pond owners and enabling professional fishermen to maintain their incomes (at the cost of greater dependence on larger landowners). Improved methods of post-evaluation (which have implications for appraisal of new projects) are developed to incorporate flood risks and consequent yield reductions and non-agricultural damages, and to standardise economic valuation. Applying these methods revealed that the Chandpur Project has been, despite an uneven distribution of benefits, relatively successful (an economic rate of return of at least 5%). Although there are some project specific reasons for this success, this project has suffered many problems typical of embankment in Bangladesh. Hence embankments are potentially important for the future economic development of Bangladesh - by reducing damage and disruption and increasing agricultural output. The study concludes with recommendations for improved flood plain planning in Bangladesh, using detailed flood loss data and improved appraisal methods, to ensure that choice of flood mitigation strategies and investments are rational, and that all public investments take flood risks into account.

Cette thèse développe et discute différentes approches micro-économiques de modélisation de l’agriculture pour représenter l’effet de changements globaux et de politiques de gestion de l’eau sur l’adaptation de l’agriculture et sur les ressources en eau. Après un chapitre de synthèse et une revue de la littérature, quatre essais sont présentés. Le premier essai décrit la représentation du comportement de dix exploitations agricoles en Alsace et en Bade (Allemagne) à partir de modèles de programmation linéaire qui intègrent la prise en compte du risque. Après extrapolation, les résultats de simulation sont couplés à une chaîne de modèle plante-sol et de transfert hydrogéologique afin d’estimer la concentration future en nitrate dans l’aquifère du Rhin supérieur. Les simulations des trois scénarios de changements - tendanciel, libéral et interventionniste - suggèrent que les concentrations en nitrates baissent dans les trois cas par rapport à la référence. Le second essai explore l’effet de l’incertitude de changements globaux sur les ressources en eau par des simulations Monte Carlo pour le modèle alsacien (premier essai) et un modèle de demande en eau agricole (Sud-Ouest). Plusieurs niveaux de dépendance entre les paramètres incertains sont caractérisés. L’analyse des résultats montre que les objectifs environnementaux peuvent être déterminés avec suffisamment de précision malgré l’incertitude forte. Le troisième essai développe un modèle agricole régional de programmation mathématique positive avec élasticité de substitution constante entre l’eau et la terre afin d’explorer comment l’agriculture, partiellement irriguée, de Beauce s’adapte à une baisse de la disponibilité en eau. La réponse du rendement à l’eau est calibrée à partir d’information agronomique. Les adaptations à la baisse de disponibilité en eau sont distinguées selon qu’elles correspondent à des baisses de dose d’eau d’irrigation ou de changement de culture. Environ 20% de la réduction est due à la baisse des doses d’eau (marge intensive). Le dernier essai présente un modèle hydro-économique “holistic” de l’agriculture et de l’aquifère de Beauce afin d’évaluer plusieurs politiques de gestion quantitatives de l’eau ainsi que d’évaluer le cas où l’accès à la ressource n’est plus régulé. Des simulations dynamiques sont réalisées à l’horizon 2040 en tenant compte de l’incertitude liée au changement climatique. La politique actuelle de quotas annualisés semble être plus coût-efficace que les autrespolitiques testées (taxes, transferts etc.). / This thesis develops and discusses agricultural-supply modeling approaches for representing the adaptation of farming to global changes and water policies: their effects on agricultural economics and water resources comprise critical information for decision makers. After a summary and a review chapter, four essays are presented. The first essay describes a representation of the behavior of ten typical farms using a risk linear programming model connected to a plant-soil-hydrodynamic model chain, to assess the future level of nitrate contamination in the upper Rhine valley aquifer. The baseline, liberal, and interventionist scenarios for 2015 all result in lower nitrate concentrations. The second essay explores the effects of the economic uncertainty of global changes by means of a Monte Carlo approach distinguishing various levels of dependence on uncertain parameters. Analyses for a nitrate-oriented and a water-use model (in Alsace and southwestern France) show that the environmental objectives can be targeted withsufficient confidence. The third essay develops a flexible specification for positive mathematical programming - constant elasticity of substitution with decreasing returns - to explore how irrigated farming adapts to increased water scarcity in Beauce, France. The possibility of adjusting the application of water per hectare accounts for about 20% of the response. The last essay presents the development of a holistic hydro-economic model of Beauce’s agriculture and aquifer under climate-change uncertainty, so as to evaluate various water policies, as well as the open-access case, up to the year 2040. The results show that the baseline policy is more cost-effective than the other instruments tested (tax, transfer,etc.).