Madagascar houses one of the Earth’s biologically richest, but also one of most endangered, terrestrial ecoregions. Although it is obvious that humans substantially altered the natural ecosystems during the past decades, the timing of arrival of early inhabitants on Madagascar as well as their environmental impact is still intensively debated. This research aims to study the beginning of early human impact on Malagasy natural ecosystems, specifically on Nosy Be island (NW Madagascar) by targeting the sedimentary archive of Lake Amparihibe, an ancient volcanic crater. Based on pollen, fungal spore, other non-pollen palynomorph, charcoal particle and diatom analyses combined with high-resolution sediment-physical and (in)organic geochemical data, paleoenvironmental dynamics during the past three millennia were reconstructed. Results indicate a major environmental change at ca. 1300 cal BP characterized by an abrupt development of grass (C4) dominated and fire disturbed landscape showing the alteration of natural rain forest. Further, increased soil erodibility is suggested by distinct increase in sediment accumulation rates, a strong pulse of nutrient input, higher water turbidity and contemporaneous increase in spores of mycorrhizal fungi. These parameters are interpreted to show a strong early anthropogenic transformation of the landscape from rain forest to open grassland. After ca. 1000 cal BP, fires remain frequent and vegetation is dominated by forest/grassland mosaic. While natural vegetation should be dominated by rain forest on Nosy Be, these last results indicate that human continuously impacted the landscapes surrounding the lake. At a local scale, our data support the “subsistence shift hypothesis” which proposed that population expansion with development of herding/farming altered the natural ecosystems. However, a precise regional synthesis is challenging, since high-resolution multi-proxy records from continuous sedimentary archives as well as records located further north and in the hinterland are still scarce in Madagascar. The lack of such regional synthesis also prevents precise comparison between different regions in Madagascar to detect potential (dis)similarities in climate dynamics, ecosystem responses and anthropogenic influences at the island’s scale during the (late) Holocene.