Ephemeral rivers are among the most striking features of arid landscapes. They flow only for a few days or weeks each year during the summer rainy season, yet beneath their sandy beds lie groundwater reservoirs that sustain entire linear oases of riparian forest. Along the Kuiseb, which rises in the Khomas Hochland and runs 560 kilometers across a drainage basin of 15,500 square kilometers, the dominant woody species include the camel-thorn tree Vachellia erioloba, the ana tree Faidherbia albida, Tamarix usneoides, Salvadora persica, and Euclea pseudebenus. These trees have evolved remarkable strategies for surviving extreme aridity: Faidherbia albida deploys an extensive root system that can tap water in various soil layers, with taproots recorded as deep as 34 meters in the Sahel, while also enriching the soil through nitrogen fixation. Vachellia erioloba goes even deeper, with rooting depths of up to 70 meters documented in the Kalahari, making its survival largely independent of sporadic rainfall.

The research team, working in two plots on the Kuiseb River near Gobabeb and a third plot in the Tsondab Valley, used a DJI Mavic 2 Pro drone flying at 90 meters over the Kuiseb plots and 45 meters over the Tsondab site. The resulting orthorectified images achieved resolutions of 3 and 2 centimeters per pixel, fine enough to delineate individual tree crowns and even fallen deadwood lying flat on the ground. More than 100 trees in the Kuiseb plots and over 60 moringa trees in the Tsondab plot were identified on the ground and located by GPS to validate the aerial interpretation. Crucially, the team constructed digital terrain models at 1-meter resolution, cutting out the trees themselves and interpolating the underlying ground surface, so that every tree could be assigned a precise vertical position relative to the deepest part of the river channel.

This vertical framing matters. Earlier work had analyzed tree distributions only in terms of lateral distance from the channel, using a coarse 30-meter terrain model and the geometric midline between the banks. But in a river like the Kuiseb, which can be 50 to 100 meters or more wide, the midline is not necessarily the deepest point of the bed. Because accessing groundwater is inherently a vertical challenge for a plant, the researchers argued that elevation above the riverbed is the more meaningful measure. They tested habitat associations using Berman tests, a spatial-statistical method that compares the distribution of a topographic covariate, here elevation, at observed tree positions against the distribution expected under complete spatial randomness. The Z2 variant of the test is sensitive to the shape of the deviation, with values beyond the range of -1.96 to 1.96 indicating significant negative or positive associations with elevation.

The results for the first Kuiseb plot, K1, were strikingly clear. Faidherbia albida, with 213 individuals mapped, grew on average 5 meters below the mean terrain elevation of the plot, yielding a strongly negative Z2 value of -14.889. Vachellia erioloba, represented by 102 trees, also showed a significant negative association with elevation, though a weaker one at -5.0989, and grew on average about 3 meters higher than the ana trees. This pattern confirms the first hypothesis of the study: the camel-thorn, with its exceptionally deep taproot, is less tightly bound to the low-lying riverbed than the ana tree, whose more lateral, shallow root system makes it dependent on regular access to groundwater and seasonal flooding. The two species thus occupy distinct hydrological niches, allowing them to coexist along the same stretch of river.

The second Kuiseb plot, K2, located only 4 kilometers away, told a very different story. There, Faidherbia albida’s negative association with elevation was no longer statistically significant, and Vachellia erioloba was actually significantly positively associated with elevation, growing above the mean terrain height. More alarming was the deadwood. In K1, only 54 localities of fallen deadwood were recorded, covering just 0.54 percent of the 33-hectare plot, while the two dominant tree species together covered 12.16 percent. In K2, by contrast, 489 deadwood localities were found, covering 12.08 percent of the 32-hectare plot, more than twenty times the proportional cover of K1, even though the living tree cover was nearly identical at 11.39 percent. The deadwood in K2 lay on average above the mean terrain elevation, with a highly significant positive Z2 value of 8.2196, indicating that tree mortality in that plot was concentrated among trees growing too far above the accessible groundwater.

Because the two plots share the same climate, the same species composition, the same conservation status within the Namib-Naukluft Park, and the same floodwater inputs, the researchers attribute the stark difference in mortality to local topographic heterogeneity rather than to upstream water management. The Kuiseb around Gobabeb failed to flow in the summers of 2007, 2010, 2013, 2016, and 2019, and during prolonged droughts trees stranded on elevated terraces simply cannot reach the water table. The team notes that floods are a double-edged sword for riparian vegetation: they deliver the germination and recruitment that ana trees depend on, yet heavy discharges can uproot mature trees and drown or flatten younger ones. Why so many trees in K2 came to grow on higher ground remains uncertain; the river may have shifted its course over years or decades, gradually separating the elevated plateau from the active channel, though the authors flag this as speculation.

Paradoxically, the surviving trees in K2 appeared to be thriving. Between February 2020 and April 2025, fifty sampled trees per plot were tracked for changes in crown area. In K1, mean crown area grew from 86.6 to 103.7 square meters, a total area increase of 851 square meters, with six trees declining. In K2, mean crown area rose from 87.8 to 114.9 square meters, a total increase of 1355 square meters, with only three trees declining, roughly 1.6 times the growth observed in K1. The authors suggest that intensified successional dynamics under strong spatial heterogeneity, possibly including asymmetric competition that favors taller individuals, may explain why growth and mortality can both be elevated in the same plot. Deadwood itself may even help: studies in South Africa’s Kruger National Park have shown that fallen wood acts like a sponge, storing water and nutrients through the dry season and creating regeneration niches for young trees. The crown growth findings also align with satellite analyses showing a 33 percent increase in vegetation cover along the Kuiseb near Gobabeb between 1984 and 2019.

The third study site, a 200 by 200 meter plot in the Tsondab Valley some 23 kilometers southwest of Solitaire, examined the phantom tree Moringa ovalifolia, the only moringa species native to Namibia. This succulent tree stores large volumes of water in trunks that can reach a meter in diameter, and it typically occupies rocky mountain slopes, though it occasionally forms small plains forests, most famously the so-called Phantom Forest in Etosha National Park. Despite this impressive drought buffering, the 116 mapped moringa trees showed a significant negative association with elevation, with a Z2 value of -3.2711, clustering along the low-lying drainage lines that traverse the plot. Even a terrain range of only 3 meters proved ecologically meaningful, because drainage lines concentrate disproportionate amounts of runoff during rainfall events. The finding demonstrates that population-level habitat associations reflect adaptations to potential drought even in species with substantial water storage, and that the native moringa, which grows more slowly in height than its Indian relative Moringa oleifera, likely invests heavily in underground root biomass as a drought strategy.

The broader implications reach well beyond the Namib. Climate projections indicate that Namibia will receive less rainfall, endure higher temperatures, and face longer dry seasons, while upstream dam construction has already driven severe dieback of Faidherbia albida in the nearby Swakop River, where mortality was worst along drier stretches deprived of tributary flow. The new study shows that mortality rates of riparian trees can differ enormously between inventory plots separated by just a few kilometers, purely because of fine-scale topography. For scientists extrapolating tree mortality, biomass, or carbon fluxes across regional gradients, and for conservationists deciding where to monitor these fragile linear oases, the message is unambiguous: in desert rivers, a few meters of vertical separation can be the difference between a thriving tree and a fallen one, and no representative survey can afford to ignore the ground beneath the canopy.

Subject of Research: Elevational habitat associations and mortality of riparian tree species along ephemeral desert rivers in the Namib

Article Title: Elevational Habitat Associations of Tree Species and Deadwood in Ephemeral Streams of the Namib Desert

Article References: Messirek, F., & Getzin, S. (2026). Elevational Habitat Associations of Tree Species and Deadwood in Ephemeral Streams of the Namib Desert. Ecology and Evolution, 16(10), Article e74451. https://doi.org/10.1002/ece3.74451

Image Credits: AI Generated

DOI: 10.1002/ece3.74451

Keywords: Namib Desert, ephemeral rivers, Kuiseb River, Faidherbia albida, Vachellia erioloba, Moringa ovalifolia, drone mapping, Berman test, deadwood, groundwater, habitat associations, climate change