Background: Rift Valley fever (RVF) is a vector-borne disease affecting ruminants and humans. Madagascar was heavily affected by RVF in 2008–2009, with evidence of a large and heterogeneous spread of the disease. The identification of at-risk environments is essential to optimize the available resources by targeting RVF surveillance in Madagascar. Herein, the objectives of our study were: (i) to identify the environmental factors and areas favorable to RVF transmission to both cattle and human and (ii) to identify human behaviors favoring human infections in Malagasy contexts. Methodology/Principal Findings: First, we characterized the environments of Malagasy communes using a Multiple Factor Analysis (MFA). Then, we analyzed cattle and human serological data collected at national level using Generalized Linear Mixed Models, with the individual serological status (cattle or human) as the response, and MFA factors, as well as other potential risk factors (cattle density, human behavior) as explanatory variables. Cattle and human seroprevalence rates were positively associated to humid environments (p<0.001). Areas with high cattle density were at risk (p<0.01; OR = 2.6). Furthermore, our analysis showed that frequent contact with raw milk contributed to explain human infection (OR = 1.6). Finally, our study highlighted the easterncoast, western and north-western parts as high-risk areas for RVF transmission in cattle. Conclusions/Significance: Our integrated approach analyzing environmental, cattle and human datasets allow us to bring new insight on RVF transmission patterns in Madagascar. The association between cattle seroprevalence, humid environments and high cattle density suggests that concomitant vectorial and direct transmissions are critical to maintain RVF enzootic transmission. Additionally, in the at-risk humid environment of the western, north-western and the eastern-coast areas, suitable to Culex and Anopheles mosquitoes, vectorial transmission probably occurs in both cattle and human. The relative contribution of vectorial or direct transmissions could be further assessed by mathematic modelling.