Conservation Agriculture (CA) was introduced at the lake Alaotra, in Madagascar, in the 2000’s in a context of traditionnal mining upland agriculture and silting-up of lowlands rice fields. Land tenure pressure linked to the attractiveness of the area lead to the progressive colonization of surrounding upland hills (Tanety), very sensitive to erosion. Conservation agriculture tackles with a double challenges: i) maintain and/or increase household income and ii) preserve natural resources through sustainable agricultural practices in the long term. This paper assesses the economic impact of CA adoption on farmers’s income trough modeling representative farms selected according to a local typology,based on the last 5 years with a prospective analysis for the next 5 years. The BV-lac Project Field database highlighted a light increase of yield according to the age of CA systems. A buffering effect on climate hazards has been as well identified trough production stability over the years leading to adoption as part of a risk limiting strategy. Elements of the CA techniques are adopted spontaneously within surrounding farming systems leading to improvment of conventionnal tillage based systems. Smallholders agricultural practices evolution displays a high capacity for innovation.Modeling with a dedidated tool (Olympe is a budget analysis oriented tool) has highlighted that CA systems improve significantly net farm income in the midterm (5 to 10 years) and gross margin at plot scale. For farm holdings with few irrigated rice fields, mainly relying on upland agriculture, CA systems increase farming systems resilience to climatic events and price volatility as well as sustainable agricultural practices maintaining localand fragile ressources

L’objectif de ce document de travail est de montrer les résultats des enquêtes faites sur les systèmes conventionnels(non SCV) dans paysans de la zone Alaotra. Les données sont les suivantes:  Extraction des données de 2007 à 2010, par type de culture non SCV à partir de la base de donnée RFR_BV_Lac avec scenarii 27_11_2010 48 exploitations RFR de 2007 à 2009 15 exploitations RFR en 2010 Document de référence: BASE DONNES NON SCV_Sarra_Colomban_2011.xlsx La détermination de classes par culture est basée sur lesrendementsdes campagnes 2006/2007, 2008-2009 et 2009/2010, issue de la base de donnée du RFR BV-Lac. Nous avons calculé des rendements moyens par classe sur trois zones confondues au lac (BRL NE, BRL VSE, et AVSF) et sur les trois campagnes.