This master thesis investigated the origin of the seemingly perennial (base) flow in a hydrological complex, Inselberg (a type of mountain) dominated catchment in the Municipality of Messica, Manica Province, Mozambique. Smallholders rely on these perennial flows, since they have three growth seasons, two of them during the dry period. This year-round agriculture is of pivotal importance for dwellers’ (economical) situation and contributes significantly to Mozambique’s food supply. However, there are many (hydrological) unknowns. How much groundwater is stored where, and how much water can be expected during the growth season. More knowledge will improve plant and irrigation schemes, minimize losses and prevent over-exploitation. The Messica Irrigation Pilot Project (MIPP) searches for hydrological answers and tries to provide practical and theoretical insights. Therefore, students performed measurements (e.g., precipitation, water levels, discharges, cross-sections) in the 220km2 large Messica Catchment since December 2012 up to August 2014.Most measurements have been performed at Godi Downstream (a headwater from the mountainous ridge). This perennial stream is small and abundantly vegetated. Another important measurement point is a free flow three-gated bridge downstream in the River Messica. Unfortunately, limited measurements are done here. Overall, the entire Messica Catchment is parsimoniously measured. Scientists also strive to improve ’Prediction in Ungauged Basins’ (PUB). This research took place in recent years. (Parts of) its knowledge and insights are used in this study. This master thesis continued on earlier theses, together they contribute to MIPP’s initiatives on three points. With this insights and practical methods, one might predict headwater baseflow one to two growth seasons ahead. This study proposes (1) the ’conveyance method’ to predict baseflow recession using cross-sectional corrected water levels (conveyances). Conveyances are measured continuously and are obtained directly, as opposed to discharge measurements. This method is based on uniform flow and linear (groundwater) reservoir assumptions together with a known cross-section. The conveyances C and discharges Q are linear related. The bottomslope and roughness are combined in one constant a. From this, we derived groundwater response timescales Ks , and hence, estimated the baseflow recession. Godi’s recession timescale is about 550 days. Together with some additional discharge measurements we are able to calculate discharges two growth seasons ahead. Inselbergs are very steep mountains, ridges or isolated hills in gently sloped areas. They are omnipresent in the Manica province, just as in other parts of East and West Africa. Normally they consist of solid, barely erodible TTG (Tonalite, Trondhjemite, or Granites) materials, as it is the case for the Kalahari Craton, where Messica is located on. Just East of Messica, the Kalahari borders the Barue Complex. Water storage in these solid Inselbergs is not obvious. However, literature study showed that Messica’s mountain ridge belongs to the later deposed Gairezi Sedimentary Group on top of the Kalahari Craton. It forms a 220km long belt from Sussendenga in the South to Guru in the North. The Gairezi consists of quartszites and metapelic schists, with internal westward-dipping inclinations. These mountains contain internal cracks and fissures, where water storage can take place. We developed two Flexible Topography driven, conceptual rainfall runoff (FlexTopo) models (FlexA and FlexC) with an additional output conversion model (OCM) to calibrate on conveyances. FlexA serves as a lumped benchmark model which incorporates all the predefined relevant hydrological processes. FlexC is semi-distributed and distinguishes three different HRUs (hydrological response units), which are entities that show similar hydrological responses. We use two objective parameters to classify the area: Head Above Nearest Drainage (HAND) (as function of the flow accumulation) and hill slope values. FlexC’s HRUs are: Wetlands (with saturated overland flow and exfiltration of deeper groundwater), Inselbergs (with perennial baseflow from its large groundwater storages, fast runoff processes due to steep, forested and/or rocky slopes) and Flatlands. The latter covers a wider scope from irrigated areas, grass and shrub lands to rain-fed fields, but contributes also (delayed) to base and peak flows. Every HRU has its own slow groundwater response reservoir. Furthermore, we applied altitude corrected precipitation to correct for orographic effects. Both FlexA and FlexC are calibrated on the River Godi and validated on the RiverMessica. FlexC showed that (2) Messica’s Inselberg ridge most likely contains an over-year storage. It predicted a groundwater storage of about 800mm in 2013-2014. This led to a specific discharge q of 1.45 mm/s at the beginning of the dry period. In general, this storage guarantees a continuous streamflow during the dry season and makes the area more resilient for dry years. Moreover, we showed with FlexA and FlexC that (3) conveyances can be used to calibrate year-round discharge variations (both base and peak flows). It also proved that FlexModels are transferable, i.e. discharge variations for the (ungauged) downstream locations (Messica Downstream Station) can be predicted based on calibration on a (gauged) subcatchment (Godi Downstream Station). Where FlexA showed better peak flow predictions according to objective functions, FlexC narrowed down uncertainty intervals around predictions on other locations and gave a more consistent model performance. Given the smaller uncertainty intervals, the landscape dependent FlexC is better transferable than the lumped FlexA. However, FlexC seemed to overestimate peak flows, for which several explanations and improvements are given. In general, the conveyance method appeared accurate to predict baseflow recession and to calibrate on low flows. It performed less well during peak flows, since the uniform flow assumption was violated. This was mostly due to the abundant (water) vegetation. The vegetation was even denser above the average water level. Additional work is needed in order to make the method fully functional for the whole flow regime.