Many ancient sedimentary basins are interpreted as endorheic basins, internally drained basins with no direct hydrological connection to the marine environment. Some of these endorheic basins are economically important because of the abundance of hydrocarbon resources. To date, many studies have been conducted on fluvial systems in endorheic basins; however, the fluvial architecture and facies distribution in ancient fluvial systems are not fully understood. Although they are an important key to rock record interpretation, modern terminal fluvial systems in semi-arid endorheic basins are rarely reported due to difficulties such as poor accessibility. The major objectives of this study are to: (1) investigate the development in space and time of the channel morphology and sediment distribution of a distal fluvial system in a semi-arid climatic setting, and (2) to build a quantitative data set for the construction of a 3D sedimentary architecture model. The study is carried out on a river terminus system, the Río Colorado, at the edge of the world’s largest salt lake, the Salar de Uyuni in Bolivia. This unconfined and largely non-vegetated river terminus provides the opportunity to acquire a large data set including field and satellite data that enables analysing the development of channel morphologies and sediment characteristics (e.g., avulsion history, splay morphology and surface dynamics). The data acquisition consists of daily precipitation data, a Global Digital Elevation Model (GDEM), a time-series Landsat imagery and high resolution WorldView-02 and QuickBird-02 satellite images, as well as surface and shallow sediment samples and high precision GPS data (Chapter 1). Changes in channel morphology of the terminal fluvial system are a function of the precipitation intensity in the catchment area (Chapter 2). The catchment area is characteristic of a mean slope of 0.0008 m/m with the highest slope near the margin and gradually decreasing slope downstream, as well as higher vegetation cover in the mountainous regions than that in the tributary delta and terminal fluvial fan. Ten peak discharge events with more than 50 m3/s have been pinpointed between 1985 and 1999. The peak discharges resulted in massive flood-out of water and sediment onto the floodplain in the very low gradient river terminus and with a cross sectional channel area of less than 80 m2. The development in space and time of crevasse splays and local avulsions was visualized by comparing Landsat MSS and TM images before and after peak discharge events. Crevasse splays expanded in peak discharge periods, and this led to amalgamation with adjacent crevasse splays by compensational stacking. The areal extent of the crevasse splays did not change in between peak discharge events. Multiple local avulsions were distinguished between 1975 and 2001. Crevasse splays and their crevasse channels can evolve over time to an entirely new river channel. Erosion greatly exceeds accretion on both banks of the river in the tributary catchment. The prominent river bank erosion is associated with an evolving channel planform such as meander morphology, channel morphology and river pattern development (Chapter 3). Normalized Difference Vegetation Index (NDVI) analysis and field investigations suggest that non-vegetation cover and abundance of desiccation cracks and burrows are the main contributors to bank erosion. Changes in channel planform are the product of continuous lateral migration and frequent overbank flooding. Shallow channels and poor development of levees in combination with in-channel accretionary benches result in frequent overbank flooding, which lead to a high density of crevasse splays over unconsolidated river banks and accretionary benches. Avulsion and chute channels together with reactivation of partially abandoned meanders and connection of headcuts and crevasse channels produce an anabranching pattern in the study area. Crevasse splays have been categorized into three classes based on their development in space and time: new crevasse splays (NCS), changing crevasse splays (CCS) and inactive crevasse splays (ICS) (Chapter 4). The occurrence of these three types of crevasse splays shows no relationship with distance along the stream. The local gradient also shows no correlation with the number of crevasse splays. By contrast, the number of crevasse splay shows an exponential increase as the cross-sectional channel area decreases. In addition, some crevasse splays are attributed to the topographic low between adjacent crevasse splays and fill in the depression by compensational stacking. The unconfined area of the system showed great variations in sediment composition downstream. The alluvial fan segment is characterized by gravel with a fining-upward sequence, whereas the upper coastal plain segment is typified by coarse sand and some fine gravel in the upstream area, grading to fine sand downstream. Silt and clay are the dominant sediments in the lower coastal plain segment, although there is also some very fine sand. It was found that the study area is characterized by a linear decrease downstream in bedload and correspondingly an increase in suspended load deposits (Chapter 5). Thus, in the upper coastal plain bedload deposits dominate, while in the lower coastal plain suspended load deposits are prominent. A spectral library has been established for four types of surface materials using Landsat surface reflectance in combination with Landsat CDR data and field data analysis (Chapter 6). Four types of surface materials are distinguished in the river terminus: A: salty surface; B: silt-rich surface; C, clay-rich surface; D: salt. The silt-rich surface has a weak correlation with the annual precipitation while the salty surface tends to be inversely proportional to the annual precipitation. There is no relationship between the clay-rich surface and the annual precipitation. The main geomorphological changes have been identified as the formation of crevasse splays and avulsions. High annual precipitation-induced avulsions are found to lead to increased silt-rich surfaces. Channel and splay morphodynamics, longitudinal sediment dispersion pattern in the system, and surface composition in the river terminus are the focus of this study. The expected preservation potential and sequence stratigraphy of the Río Colorado system is discussed. Future work on such dryland river systems could include investigations of the mechanisms of fluvial discontinuity at the boundary between alluvial fan and upper coastal plain, the interaction between aeolian dunes and fluvial systems in the Río Colorado river system, and groundwater regimes in the coastal plain (Chapter 7).