Spatial variability in stable isotopes from Lesotho surface waters: insights into regional moisture transport

Precipitation in Lesotho is highly spatially variable, a feature of the high altitude and rugged topography. The hydroclimate dynamics, despite being critical to the water security of Lesotho and adjacent South Africa, are poorly understood. Ratios of oxygen and hydrogen isotopes in meteoric water are excellent tracers of hydroclimatic processes. This study presents the first analysis of stable isotopes from surface waters in Lesotho, and an investigation into the moisture sources. Our results demonstrate considerable variability in isotope values. There are statistically significant relationships between both oxygen and hydrogen isotopes and the altitude of the site and source of rivers sampled, and with hydrogen isotopes and longitude. The meteoric water line for the Lesotho samples is most closely aligned with that of the Global Network of Isotopes in Precipitation (GNIP) station at Harare, in Zimbabwe. The meteoric water line for Windhoek is more closely aligned to the Lesotho samples than the more proximate Cape Town or Pretoria meteoric water lines, which would more closely represent the South African winter-and summer-rainfall zones respectively. HYSPLIT back-trajectory air parcel analysis supports these findings, demonstrating a frequent continental anticyclonic track through southern Zimbabwe. Deuterium excess values vary widely, although are most likely related to processes during moisture transport rather than differences in moisture source. These findings are of particular importance in the context of the future water security of both Lesotho and South Africa, especially as the poleward displacement of the westerly moisture corridor has raised concerns for winter precipitation in the region.


Introduction
The eastern Lesotho highlands are often referred to as the 'water tower' of southern Africa (UNEP 2010;Hitchcock 2015;Geppert et al. 2022).The hydroclimate of the region is crucial in supplying water to the subsistence farms of Lesotho, and to the economic and industrial hubs of adjacent South Africa through the export of water through the Lesotho Highlands Water Project (Matete and Hassan 2006).While the region currently experiences some of the largest volumes of mean annual precipitation in southern Africa, palaeoclimate records indicate that the eastern Lesotho Highlands have experienced a reduction in surface water availability during the Holocene (Fitchett et al. 2017;Norström et al. 2018).Climate change projections indicate likely reductions in both precipitation and soil moisture for the southern African region under 1.5 -4 °C warming scenarios to 2080-2100 (Archer et al. 2018;IPCC 2021).For the eastern Lesotho highlands, the projections are less consistent (Morris 2017;Mukwada et al. 2020), but considering both climatic change and future water demand in Lesotho and South Africa through population growth (GCRO 2019), the region is anticipated to face water scarcity before the end of the century, under all modelled climate change scenarios (Pryor et al. 2022).Western Lesotho, which serves as the economic hub of the country, experiences lower total annual rainfall, with dynamics more consistent with those of South Africa (Hydén 2002).The moisture sources of precipitation in both eastern and western Lesotho remain poorly understood (Hydén 2002).
Lesotho is located within the summer rainfall zone of southern Africa, with 70-80% of the mean annual precipitation falling between the months of November and March (Sene et al. 1998;Fig. 1).During these months the precipitation comprises orographic rainfall along the escarpment and across the rugged topography (Sene et al. 1998;Xulu et al. 2020), and more short-lived local convection (Dedekind 2015) associated with the Tropical Temperate Trough (TTT) that extends from Namibia to the southeast coast of South Africa (James et al. 2020).Less than 10% of the mean annual precipitation falls during the winter months, predominantly as snowfall, induced by mid-latitude cyclones and associated Cut-off Low (CoL) systems (Stander et al. 2016;Wunderle et al. 2016;Grab et al. 2017).Although a small contribution to the mean annual precipitation, this winter precipitation is likely to diminish with the progressive poleward displacement of the westerlies (Sousa et al. 2018).The TTTs, influenced by ENSO and South Indian Ocean sea surface temperatures (Manhique et al. 2011), are also likely to change in their dynamics under climate change.Key to projecting the hydroclimate future of Lesotho is understanding the moisture transport associated with the synoptic features that drive precipitation in the region, and their relative contributions to seasonal and annual precipitation.The paucity of ground-based meteorological data for Lesotho, and the complexity of the rugged topography present challenges in this regard.
Hydrogen and oxygen Isotope ratios ( 2 H/ 1 H and 18 O/ 16 O respectively) of precipitation and resultant surface waters can serve as important tracers of moisture sources and hydroclimatic processes (Sengupta and Sarkar 2006;Xie et al. 2011;Li et al. 2016;Ren et al. 2017;Geppert et al. 2022).Isotope ratios of oxygen and hydrogen in the water molecule are generally expressed as deviations, in parts per thousand (‰), relative to a standard and expressed in delta (δ) units: Where R = 18 O/ 16 O or 2 H/ 1 H and δ = δ 18 O or δ 2 H and the relevant standard for water is VSMOW.
The relationship between δ 2 H and δ 18 O for meteoric waters worldwide is described by the Global Meteoric Water Line (GMWL), given by: (1) (2) 2 H = 8.17 ± 0.906 x 18 O + 10.35 ± 0.65 Fig. 1 A Topography and major drainage features of Lesotho along with location of the sampling sites for this study.Inset shows location of Lesotho and of the 4 GNIP stations within southern Africa.Rozanski et al. (1993).Geographical variations in the δ 2 H and δ 18 O relationship form Local Meteoric Water Lines (LMWLs), the slope and intercept of which are controlled by moisture source and various atmospheric processes.The intercept of the relationship is referred to as the deuterium excess (d-excess), where: Dansgaard (1964).The deuterium excess is a second-order parameter controlled by sea-surface conditions (temperature and especially humidity) at the site of primary evaporation, as well as by the evaporation of raindrops below the cloud base and any addition of 'recycled' moisture to the atmosphere from the land surface (Fröhlich et al. 2002).
Water-isotope data are used extensively to investigate moisture transport ranging from individual rainfall events over very short periods of hours (Soderberg et al. 2013;Good et al. 2014), through to long-term rainfall dynamics over seasons to years (Vystavna et al. 2021;Li et al. 2022a).Many of these studies use data from the International Atomic Energy Agency (IAEA) Global Network of Isotopes in Precipitation (GNIP) in their analyses.While this network comprises more than 1000 sites across 125 countries, it is sparse in southern Africa (Geppert et al. 2022).Only four sites (Cape Town and Pretoria in South Africa, Windhoek in Namibia, and Harare in Zimbabwe; Fig. 1) have long-term records.There are no GNIP stations in Lesotho.A small dataset of surface water analyses from Katse Dame and other unspecified locations in Lesotho is reported as part of the study of de Wet et al. (2020) into the isotopic composition of South African tap water.Analyses of rainwater isotopes from Cape Town (Harris et al. 2010), Mossel Bay (Braun et al. 2017), Thohoyandou (Durowoju et al. 2019) and Johannesburg (Leketa and Abiye 2020), and from sites across Namibia (Kaseke et al. 2016(Kaseke et al. , 2018)), confirm the variability in stable isotope ratios across the different rainfall zones that characterise the subcontinent (Geppert et al. 2022), and in turn the value in tracing their moisture sources.
In this study we present a spatial analysis of surface water samples obtained from rivers, dams, springs and snowpack in the northern half of Lesotho in a reconnaissance study, coupled with HYSPLIT moisture back-trajectory modelling.The study has two goals (1) to assess the utility of surface-waters for tracing moisture source and (2) to provide a baseline for future, more detailed, work. (3)

Study site
Lesotho is a small landlocked country, surrounded entirely by South Africa, spanning 28°34'36.923"S-30°40'56.806"S and 27°0'28.134"E-29°27'47.735"E,covering a land surface area of 30,340km 2 (Mbata 2001;Fig. 1).Lesotho has the highest low point of any country in the world, at 1400 m.asl in the western Lowlands, and a maximum elevation of 3482 m.asl at Thabana Ntlenyana in the eastern Highlands (Grab and Nash 2009).Mean annual precipitation decreases with distance from the Indian Ocean, from approximately 1600 mm in the eastern Highlands along the escarpment (Sene et al. 1998), to 735 mm in the lowlands (Hydén and Sekoli 2000).While the majority of rainfall is experienced during summer months (Sene et al. 1998), snow falls approximately 8 times per annum in the highlands, and once every 2.3 years in the lowlands, predominantly during the colder months of April through September (Grab et al. 2017).Mean annual temperatures range from approximately 6 °C in the highlands to 15 °C in the lowlands, reaching maxima of 29 °C in the lowlands mid-summer, and minima of -6.1 °C in the highlands in winter (Grab and Nash 2010).
The source of the Senqu-Orange River, the longest and one of the largest rivers in southern Africa, is located in the eastern Lesotho Highlands (Putteman et al. 2011).The Senqu basin, referring to the catchment within the political boundaries of Lesotho, covers approximately two thirds of the land surface area of the country (Pryor et al. 2022).Major tributaries of the Senqu River, including the Mohokare River, flow from the highlands to the lowlands (Chatanga et al. 2019).Despite the uniform basalt geology across the eastern Lesotho Highlands, and sandstone in the western Lesotho Lowlands, the river morphometrics are highly spatially variable (Knight and Grab 2018).Although mean annual precipitation is relatively high, there is only one known natural lake in Lesotho, Letšeng-la Letsie, in the central southern part of the country (Rose et al. 2020).There are, however, a number of small rockbottom tarns and grass pans in the Sehlabathebe National Park, located along the eastern escarpment (Dunnink et al. 2016).A number of small dams have been built to provide water to cities in the western Lesotho Lowlands (Bates and Haacke 2003;Gwimbi and Selimo 2020).Major dams that were built to form part of the Lesotho Highlands Water Project, which export water to South Africa, include the Katse, Muela and Mohale Dams, while the Polihali Dam which forms part of stage II of the project is still under construction (Tanner et al. 2009;Vinti 2022).

Field water sampling
Fieldwork was conducted in Lesotho during the winter season between 2 and 5 June 2022.Surface water samples were collected from snowpack, springs, rivers and dams along an approximate E-W transect spanning the eastern Lesotho Highlands and the western Lesotho Lowlands, and a N-S transect along the western border of the country (Fig. 1).Due to the rugged topography and the sparse road network, a convenience sampling approach was used, traversing the major road networks of Lesotho, the majority of which are untarred.Permission for the study was granted by the Lesotho Ministry for Tourism, Environment and Culture.
Water samples for isotope analysis were collected in 50mL polyethylene sample bottles, which were completely filled, sealed and stored in a cool box to prevent evaporation.In the case of snowpack, sample bottles were filled with compacted snow, which later melted out.All samples were later subsampled into 20mL duplicate bottles, and re-sealed for transportation.At each site, GPS coordinates and altitude were recorded using a Garmin eTrex handheld device.For record-keeping, air temperatures were recorded using a Kestrel 5000 handheld device.For samples obtained from rivers, springs and dams, water quality variables including water temperature, pH, electrical conductivity (EC) and dissolved oxygen (DO) were measured using a calibrated YSI multiparameter probe.For the river samples, the river source altitude was obtained from Google Earth, tracing each river to its source, and identifying the altitude from overlaid contours at 20 m intervals (Table 1) following the approach of Lachniet and Patterson (2002).
Water samples were transported to the University College London Bloomsbury Environmental Isotope Facility (BEIF) for analysis.Prior to analysis, aliquots of each sample were filtered and transferred to 2 mL glass Thermo vials.Samples were then analysed for oxygen and hydrogen isotopes by Cavity RingDown Spectroscopy (CRDS) using a Picarro L2130-i analyzer.Results are expressed in δ notation relative to the VSMOW standard.The errors, based on multiple analyses of an in-house standard and the variability on the repeat analyses of the 9 injections for each sample, were better than 0.17 and 1.1‰ for δ 18 O and δ 2 H respectively.

Identifying moisture sources of precipitation in Lesotho
To identify the various moisture sources and air-mass pathways of precipitation in Lesotho, the National Oceanographic and Atmospheric Association (NOAA) Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Stein et al. 2015) was used.Following the methodology detailed by Li et al. (2022a) for moisture trajectory analysis in alpine river watersheds, back trajectories of all air parcels were modelled in HYSPLIT from the central field-sampling GPS coordinates of 29.11°S, 28.47°E for the 72-hours preceding each day of precipitation at the same coordinates, at 6-hourly resolution, and at three levels of 500 m, 1000 and 2000 m above ground level.Because the isotopic composition of streams and rivers often approximates the mean weighted precipitation isotope value (Mook 2006), we conducted this analysis for each precipitation event in the centre of our sampling region for a full calendar year prior to sampling.As there is a paucity of meteorological data in Lesotho (Sene et al. 1998), rainfall days were estimated using the Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) gridded rainfall data (Funk et al. 2015).Days with snowfall were identified from Snow Report South Africa.Input data for HYSPLIT include cloud height, wind direction, temperature and surface pressure (Bagheri et al. 2019), which were obtained from the Global Data Assimilation System (GDAS) global 1° resolution meteorology data.Dominant air-mass pathways were identified, and their proportional representation calculated, through cluster analysis of all moisture tracks calculated from each of the model runs (Rose et al. 2020;Li et al. 2022a).A second set of HYSPLIT runs were modelled for 72-hour back trajectories of the same set of precipitation dates, with the overlay of relative humidity from the GDAS global 1° database (Rapolaki et al. 2020), to identify those trajectories that had air parcels saturated with moisture along or in the terminal phase of their path.

Results
A total of 35 samples were collected from 33 sites (Table 2) including rivers, snowpack (LS3, LS5, LS11, LS12), springs (LS7, LS8) and dams (LS27, LS28, LS32).Snow samples were collected at a maximum altitude of 3205 m. asl (LS3), while river sample sites ranged in altitude from 3063 m. asl (LS4) to 1461 m.asl (LS33; Table 1).River source altitudes ranged from 3350 m.asl (LS14) to 1690 m.asl (LS29; Table 1).For the river samples, DO ranged from 86.6% (7.98 mg/l, LS21) to 127.8% (12.51 mg/l, LS26), while generally higher levels were recorded for the dam samples (Table 2).The EC was generally much lower in the alpine regions, as would be expected for first order streams, increasing in western Lesotho lowlands, with values ranging from 191.1 µS/cm (LS21) to 317.5 µS/cm (LS32, Table 2).The EC for the dams was representative of that of the nearby rivers, and considerably higher than the high-altitude streams (Table 2).Both air and water temperature varied both by altitude and time of day (Table 2).
The isotope values for the dataset as a whole range from + 0.40 to -18.53‰ and + 1.5 to -138.3‰ for δ 18 O and δ 2 H respectively (Table 3).The lowest δ 18 O and δ 2 H values are recorded for the samples from the snowpacks, ranging from − 18.53‰ (LS11-2) to -5.85 (LS3-2) for δ 18 O: corresponding values for δ 2 H are − 138.3 (LS12) and − 28.9‰ (LS3-2).This snow remained on the land surface following a snowfall event on 23 May 2022 and there had been no further precipitation events that date until the time of sampling.The d-excess values for snowpack varied from 17.9‰ (LS3-2) to 9.4‰ (LS5).
The isotopic values for the two samples from springs did not differ markedly from those of the river samples (δ 18 O values of -4.11‰ -4.70‰, δ 2 H values of -23.9‰ and − 26.5‰ and d-excess values of 9.2‰ and 11.1‰ were obtained for samples LS-7 and LS-8 respectively) whereas the dam samples show slight heavy-isotope enrichment, with δ 18 O and δ 2 H values ranging from + 0.40‰ and + 1.5‰ respectively (LS28) to -2.76‰ and − 13.6‰ respectively (LS32), with wide variation in d-excess.
The δ 18 O and δ 2 H values for all water and snow-pack samples are strongly correlated and described by an ordinary-least squares regression given by: for the entire dataset excluding the lake samples, which appear to lie on an evaporation line.
(4) 2 H = 7.8 x 18 O + 9.9 R 2 = 0.99 Following the analysis of Lachniet and Patterson (2002), the correlations between the isotopic ratios for the river samples against a range of geographical characteristics were calculated using Pearson's correlation coefficient.For both δ 18 O and δ 2 H, statistically significant correlations are calculated for the relationships between isotope values and the altitudes of both the sample site and of the river source (Table 4; Fig. 2 -shown for δ 18 O only).The strongest correlation is calculated for δ 2 H and site altitude (r = 0.71, p = 0.0001).Notably, for both δ 18 O and δ 2 H, stronger correlations are calculated for the site altitudes than for the source altitudes of the rivers.For δ 2 H, a statistically-significant correlation is also calculated between the isotope values and longitude of the site.Given the position of Lesotho, longitude additionally serves as a proxy for distance from the Indian Ocean.No statistically significant correlations are calculated between isotope values and latitude (Table 4).No statistically significant relationships are calculated between d-excess and any of the geographical variables (Table 4).However, d-excess shows variability with δ 18 O.For the snowpack samples, albeit few in number, there is an increase in d excess with δ 18 O whereas for the other samples, d-excess decreases sharply with increase in δ 18 O (Fig. 3).

Local Meteoric Water Lines
As the slope and y-axis intercept are indicative of moisture source, physical setting and climatic conditions, and because the isotopic composition of surface waters can reflect weighted annual precipitation in some circumstances, local meteoric water lines can be used to investigate the source of surface water.As there are no GNIP stations in Lesotho, there is no published meteoric water line based on precipitation for comparison.We therefore present the local meteoric water line for our samples, and the meteoric water lines for each of the four southern African GNIP stations at Harare, Windhoek, Cape Town and Pretoria (Fig. 4).We note that LMWL from a two-year-long daily rainfall isotope record from Johannesburg (Lakete and Abiye 2020) is very similar to that from the GNIP station at near-by (~ 60 km) Pretoria, as would be expected.We also note the close similarity between our local meteoric water line for Lesotho and a best-fit line through oxygen and hydrogen isotope values for Lesotho surface waters reported in de Wet et al. (2020;their Fig. 11).
The two GNIP stations closest to Lesotho are Cape Town Airport and Pretoria.Notably, Cape Town is located within the contemporary winter-rainfall-zone of southern Africa, with the westerlies forming the primary moisture source (Roffe et al. 2022).Pretoria and Johannesburg are located within the summer-rainfall-zone, and similar to Lesotho, the majority of rainfall is convective in nature, associated with the TTT (Todd and Washington 1999;Crétat et al. 2012).Based both on proximity and on the synoptic conditions, we would anticipate that the Lesotho MWL would sit between the Pretoria and Cape Town MWLs, likely with a slope more similar to that of Pretoria.However, both the slope and the y-intercept are most similar to the Harare MWL, and are more similar to Windhoek than to either Pretoria or Cape Town (Fig. 4).Notably, even the snow samples, which are from a single, known precipitation event, and which are understood to be a feature of the CoL system associated with the westerlies, predominantly sit on the Harare and Windhoek MWL, rather than the Cape Town MWL.The isotopic composition of the spring samples from Lesotho is consistent with that of streams, although since only two springs were sampled, we cannot draw further conclusions based on spring composition.

Moisture source and air transport modelling
Averaging the 72-hour back trajectories of all air parcels for the 70 precipitation days, 11 dominant air parcel transport pathways are evident.Approximately 13.5% of the airmasses originate from south of the southern African subcontinent, tracking east and then north over the Indian Ocean, before recurving west to Lesotho (Fig. 5a).A further 12.9% originate further north over the Indian Ocean in the Mozambique Channel, curving northwest, before recurving south to Lesotho.The largest proportion of tracks, at 19.6%, originate over central Mozambique, and move west through southern Zimbabwe before recurving southwards to Lesotho.This dominant pathway is consistent with the alignment in the isotope records with the MWL for Harare.A similar anticlockwise track path is followed by air parcels originating over southeastern Botswana, tracking across Botswana (6.9%); central south Botswana, tracking due south to Lesotho (9.6%); western Botswana, tracking south along the Botswana-Namibia border (4.8%); and central Namibia, tracking southeast to Lesotho (4.1%).The latter two correspond with the alignment of the isotope values with the MWL for Windhoek.Approximately 17.7% of the tracks originate over the Atlantic Ocean and follow a due-easterly path to Lesotho.A further 2.7% originate south of southern Africa, tracking west before recurving north and then east over the Atlantic Ocean.A small proportion of tracks are very short in length, originating over land in close proximity to Lesotho from the south (2.0%), and north (6.2%) respectively (Fig. 5a).
Inspecting the individual 72-hour back-trajectories of all air masses for each storm event, there are no notable differences between the tracks, and in turn the broad climate dynamics, of snow and rain events collectively, nor any distinct shifts in the different types of tracks between seasons.There are, however, more frequent due-east tracks off the Atlantic Ocean during the winter months, and more common anticyclonic tracks off the Indian Ocean during the summer months.It is important to note that these refer to all air parcel trajectories, not all of which would be moisture bearing, or indeed sufficiently moisture bearing to contribute to a precipitation event along their trajectory or in the region of interest, Lesotho (Rapolaki et al. 2020).Exploring only those air masses that are close to (> 70% relative humidity) or at saturation (100% relative humidity), there are broad similarities in the distribution of clustered air masses (Fig. 5b).Following the work of Li et al. (2022a) in an alpine setting, we argue that an analysis of all air masses is likely to provide a more comprehensive overview of both the synoptic climate systems responsible for precipitation events, and of the pathways of moisture transport.This is because the rugged topography induces orographic saturation over  The entire set of water and snowpack samples and B for all samples excluding snowpack.
The regression line for this study describes all data except samples from lakes.Local meteoric water lines from the four southern African GNIP stations are also shown.For explanation, see text much shorter timescales than the 6-hourly trajectories, and air masses that are at comparatively low relative humidity along their track to Lesotho may reach saturation due to the considerable uplift.This is demonstrated in particular by the precipitation events in Lesotho for which none of the air mass back trajectories have a modelled relative humidity of > 50% for the 72-hour period prior to precipitation.

Spatial variations in surface-water isotope composition
The stable isotope values from the sampled surface water and snowpack in Lesotho demonstrate considerable spatial variation.The greatest variation is between the snow and water samples, as snow forms in clouds under non-equilibrium conditions (Lamb et al. 2017;Bagheri et al. 2019).Moreover, snow typically forms at low temperature and is thus heavy-isotope depleted in contrast to river waters, which tend to approximate the isotopic composition of weighted annual precipitation.There is notably less variation between the river, spring and dam samples than has been found elsewhere in southern Africa (Geppert et al. 2022).Despite the small number of samples, the dam waters may lie on a line with lower gradient than the rivers (Fig. 4), which is consistent with evaporative enrichment, commonly seen in lakes.
The spatial variability in stable isotope values is consistent with the spatial heterogeneity in surface moisture and precipitation across Lesotho (Armstrong et al. 1993), a result of the rugged topography, particularly in the eastern Highlands (Sene et al. 1998).The negative correlation between both δ 18 O and δ 2 H and altitude (Fig. 2) is an expression of the well-known altitude effect, which arises from heavy-isotope depletion as air is forced to rise by the topography, leading to adiabatic cooling.The relationship with longitude, which is similar to that for altitude (although only statistically significant for δ 2 H) (Fig. 2), is also a reflection of the altitude effect, because the topography rises from west to east.These relationships are stronger than those found for Costa Rican surface waters (Lachniet and Patterson 2002) in a similar topopgraphic setting, where the strongest relationship was found between δ 18 O values and altitude (r = 0.38).Lachniet and Patterson (2002) interpret the weak correlations found in their study to represent complex and multi-variate influences of geographical variables.In the case of Lesotho, there likely are multivariate influences, and these statistically significant relationships for altitude and longitude do not preclude this.Moreover, as the difference in isotopic values between the Highlands and Lowlands (as defined by a 2000 m.asl threshold altitude) is weak but statistically significant, it is also likely that the relationships are not linear.Thus, while linear statistical tests have been employed in the interests of comparability of data, future analysis should explore more complex statistical models.Deuterium excess in the Lesotho surface waters is likely to be strongly related to the d-excess in precipitation, which itself is controlled by conditions at the source of vapour formation (sea-surface temperature and, especially, relative humidity: Pfahl and Sodemann 2014) and modifications during vapour transport (Fröhlich et al. 2002).Given that the surface waters in our samples from Lesotho reflect the Fig. 5 A all proportional airmass trajectories, and B all saturated airmass trajectories, for Lesotho for each precipitation event over the period 30 May 2021-1 June 2022, modelled using HYSPLIT averages of moisture from several different sources, variations in source are unlikely to explain the range of d-excess values in Fig. 3. Instead, the negative correlation between δ 18 O and d excess may be explained by enhanced evaporative loss of raindrops at higher temperature (higher δ 18 O values) possibly associated with lower relative humidity.Possible evaporative enrichment of dam (lake) samples seems not to have modified these, so that they fall within the general trend of the surface waters.The snow samples, by contrast, show a positive correlation between δ 18 O and d-excess.As noted, the snow packs samples accumulated during a single event and the variations may simply reflect variations in atmospheric temperature during that event.
The working understanding of the hydroclimate dynamics of Lesotho has been that the summer precipitation derives predominantly from the Indian Ocean, transported by the easterlies, and frequently takes the form of convective storms (Sene et al. 1998;Hydén 2002).Winter precipitation, often in the form of snowfalls in the eastern Lesotho Highlands, is understood to be associated with mid-latitude cyclones and associated CoLs, originating from the Southern Ocean and South Atlantic (Nel and Sumner 2008;Mulder and Grab 2009).On the basis of the dichotomy between easterly-derived summer precipitation, and westerly-derived winter precipitation, and of proximity to the sites, it was hypothesised that the meteoric water line for the Lesotho surface water samples would sit between that of Pretoria, representing the southern African easterly-derived summerrainfall zone, and Cape Town, representing the westerlyderived winter rainfall zone, with a likely closer affinity to that of Pretoria.The results demonstrate a meteoric water line and d-excess that is most similar to that of the Harare GNIP station, and with greater similarity to the Windhoek station than to Pretoria or Cape Town, indicative of a primary moisture source north of South Africa.This is important when considering the impact that the progressive southward displacement of the westerlies (Sousa et al. 2018) may have on precipitation and water security in Lesotho in the future.
The role of TTT cloud bands, stretching from Namibia through to the southeast coast of South Africa, in inducing rainfall in subtropical southern Africa during summer months, is increasingly well understood as a synoptic phenomenon (e.g.Vigaud et al. 2012;Ratna et al. 2013;James et al. 2020).Moreover, the role of the TTTs in explaining convective rainfall in Lesotho has been documented (Crétat et al. 2012;Hart et al. 2013), but little interpretation has been conducted on the dynamics of this feature across the country.The TTTs and the South Indian Ocean Convergence Zone (SICZ) interact with a range of synoptic features including the Mascarene and St Helena Highs, the Angolan Low, the Congo Air Boundary, and the westerly storm tracks (James et al. 2020).Notably, a greater number of TTTs occur during La Niña years (Ratna et al. 2013);the year prior to sampling, which arguably represents the reservoir period for the surface water samples, followed a strong 'triple' La Niña (Jones 2022;Li et al. 2022b).Three dominant moisture sources have been identified for TTTs, namely a northerly flux from the Indian Ocean, a northwesterly flux from equatorial southern Africa and the far eastern equatorial Atlantic, and a southerly cyclonic flow around the TTT (Todd and Washington 1999: 941).These three fluxes align relatively well with the outputs of the HYSPLIT back-trajectory modelling for precipitation days in Lesotho.The dominance of the anticyclonic circulation track, which originates within the 72-hour window over the northeastern interior, likely originally from the Indian Ocean, and tracks through Zimbabwe, explains the strong coherence with the meteoric water line for Harare.Likewise, circulation through the TTT of moisture from the Atlantic Ocean would explain the similarities with Windhoek.The advection of air parcels in a south-easterly direction, along the SICZ, has been referred to in the case of smoke transport as the "river of smoke" (Swap et al. 2003: 1), and air-parcel trajectory analyses of air pollutants is consistent with those modelled for this study (Zunckel et al. 2000;Diab et al. 2004;Rose et al. 2020;Kok et al. 2021).To the best of our knowledge however, moisture transport along this "river of smoke" has not been explored in detail.
As there is no GNIP station, nor known rainfall isotope records, for Durban or similar sites along the east coast of southern Africa, it was not possible to identify a direct Indian Ocean rainfall isotope signal.This forms an important direction both for future research on surface water and rainfall isotopes in South Africa and Lesotho, and for higher resolution rainfall isotope data collection through GNIP.It will also be valuable to establish regular rainfall isotope monitoring within Lesotho, to contribute to the relatively sparse network of GNIP stations across southern Africa.Further analyses of the isotope record to investigate the source water of moisture in Lesotho would also benefit from a more extensive sampling of surface waters across Lesotho, repeated across different seasons.

Conclusion
This study presents the first analysis of hydrogen and oxygen isotopes from surface waters in Lesotho.The analysis of these isotopes reveals considerable spatial variability, with strong correlations between the isotope values and the site and source altitude of the river samples.For δ 2 H strong relationships are also found with longitude, which due to the position of Lesotho and the Great Escarpment of southern Africa, represents a proxy for distance from the Indian Ocean and is also related to altitude.The meteoric water line for the Lesotho samples aligns most closely with that of the isotope data from the Harare GNIP station.This can be explained through the back-trajectory analysis of air parcel movement for the 72-hours prior to each precipitation event over the year leading up to sampling, with the greatest proportion of air mass trajectories circulating in an anticyclonic motion over continental southern Africa, passing through Zimbabwe.The alignment with the meteoric water line for Windhoek is less clearly supported by the results from HYSPLIT, but likely relates to the role of the TTT cloud band, and moisture transport that may be associated with the "river of smoke".More detailed analyses of these isotope records would benefit from a higher resolution rainfall isotope network in southern Africa, and in Lesotho.More detailed sampling of surface waters in Lesotho should focus on investigating any seasonal variations.This work is of value in understanding the contemporary synoptic systems responsible for precipitation in Lesotho, which would inform climate modelling and hydroclimate projection efforts in determining the future water security of both Lesotho, and South Africa's economic hub.

Fig. 2
Fig. 2 δ 18 O (A and B) and δ 2 H (C and D) versus altitude and latitude for river samples only Fig. 4 δ 18 O v δ 2 H for A. The entire set of water and snowpack samples and B for all samples excluding snowpack.The regression line for this study describes all data except samples from lakes.Local meteoric water lines from the four southern African GNIP stations are also shown.For explanation, see text

Table 1
Details of the four long-term GNIP sites in southern Africa (IAEA/WMO (2023) Global Network of Isotopes in Precipitation.The GNIP Database.Accessible at: https:// nucle us.iaea.org/ wiser)

Table 2
Location, sample details and physical variables of water samples

Table 3
Isotope values

Table 4
Correlations between stable isotope ratios for river