Nature-based solutions in mountain catchments reduce impact of anthropogenic climate change on drought streamflow

The study addresses how Nature-based Solutions (NbS) can modulate the biophysical impacts of anthropogenic climate change (ACC) on extreme events, focusing on drought streamflow. Extreme events arise from climate phenomena interacting with the sensitivity of the biophysical environment. While mitigation progress has been limited, adaptation actions at local scales are emphasized, including ecosystem management and restoration. Despite wide promotion of NbS, there is little quantitative evidence on their ability to reduce impacts specifically attributable to ACC for realized extreme events. Event attribution science provides a framework to disentangle the ACC signal in meteorological extremes and trace it through to hydrological impacts. The paper investigates the 2015-2017 Cape Town 'Day Zero' drought to determine: (i) whether ACC reduced drought-period streamflow relative to a counterfactual Natural climate; (ii) whether invasive alien tree (IAT) clearing could have reduced or offset this impact; and (iii) whether pre-drought restoration avoided further exacerbation. The work is motivated by the critical role of mountainous headwater catchments for Cape Town’s water supply and the regional relevance of IAT management.

Existing NbS assessments have largely focused on climate mitigation (e.g., carbon removal, temperature effects) at global scales, which limits applicability to local adaptation decisions. Reviews document the role of NbS for adaptation and disaster risk reduction but generally do not isolate the ACC-attributable portion of impacts for specific events. Few studies jointly attribute biophysical impacts to ACC and evaluate NbS’ moderating role. In water-limited and winter-rainfall regions, paired-catchment and modeling studies consistently show that replacing native shrub/grass vegetation with trees (afforestation or IAT encroachment) increases evapotranspiration and reduces streamflow, especially during dry periods. Syntheses report decreased water yield in ~80% of cases following afforestation in ecosystems with shorter natural vegetation, with baseflows reduced in 63% of studies. Even where infiltration increases, higher transpiration can reduce aquifer recharge and low flows. Global proliferation of IATs and woody encroachment poses challenges, particularly in water-stressed regions. The literature highlights a gap in quantitative, event-specific, ACC-attribution studies that also evaluate locally relevant NbS (especially in the Global South).

Study area: Two mountainous headwater catchments in South Africa critical to Cape Town’s supply: Upper Berg (78 km²) and Du Toits (46 km²). The Mediterranean climate features winter rainfall; geology and soils are characteristic of the Table Mountain Group with highly leached sandstone-quartzitic soils. During 2015–2017, observed daily streamflow declined to 45% (Upper Berg) and 67% (Du Toits) of long-term means. IAT infestation during the drought was ~9% (Upper Berg) and ~40% (Du Toits). Nature-based Solution: Invasive alien tree (IAT) management via clearing. Three landscape states were defined: Current (C) invasion (9% Upper Berg; 40% Du Toits), Cleared (CL) with all IATs removed, and fully Invaded (I) representing spread to areas available for invasion (90% Upper Berg; 98% Du Toits). Joint event attribution framework: Four paired climate–IAT states were simulated for the drought period: (1) Natural climate + Current IAT (NC; reference), (2) Actual climate + Current IAT (AC), (3) Actual climate + Cleared IAT (ACL), (4) Actual climate + Invaded IAT (AI). This allowed attribution of ACC on drought-period streamflow and the moderating or exacerbating roles of IAT clearing or spread. Climate attribution inputs: Three climate model experiment ensembles provided daily rainfall and reference evapotranspiration (RefET) inputs: (i) Weather@home (HadRM3P nested in HadAM3P-N96), 68-member AGCM ensemble; (ii) C20C+ ECHAM5.4, 50-member AGCM ensemble; (iii) CMIP5 multi-model CGCM ensemble, 27 models. Total per climate state (Actual, Natural) = 145 simulations; overall 290 climate simulations. AGCM Actual runs used observed SST/SIC and contemporaneous forcings for 2015–2017; Natural runs used detrended/naturalized SST/SIC and pre-industrial GHG/aerosols. For CMIP5, analog droughts were constructed by selecting the driest consecutive 3-year periods from 1869–1899 (Natural) and 2001–2031 (Actual) for each model. Bias correction and derivations: Daily precipitation and min/max temperature were extracted at nearest grid points to local stations (15 years for AGCMs, 31 years for CGCMs). RefET was derived using Hargreaves-Samani. A quantile–quantile bias correction with a 20-day moving window aligned model distributions to station observations; Natural-state daily quantiles were aligned to the Actual-state distributions before mapping to observed quantiles. Model evaluation against CRU and ERA5 indicated satisfactory seasonal and spatial rainfall reproduction; bias correction improved correlations. Hydrological modeling: The MIKE SHE physically based model coupled with MIKE HYDRO River simulated daily catchment hydrology for each climate–IAT state. Spatial discretization was 60 m, with 32,400 cells (Du Toits) and 48,400 cells (Upper Berg). Three vertical computational layers represented: unsaturated soil zone (1.5 m), a perched/weathered saturated zone (15 m) to mimic piston flow, and a deep Table Mountain Group aquifer (average depths ~700 m Berg; ~900 m Du Toits). Overland flow used a diffusive wave Saint Venant approximation; unsaturated flow via Richards’ equation; groundwater via 3D Darcy; ET via Kristensen & Jensen; river routing via MIKE HYDRO using fully dynamic Saint Venant equations. Land cover (including IAT distributions) derived from Sentinel-2 imagery guided overland parameters. Meteorological inputs used local station rainfall (interpolated with elevation lapse rate) and observed RefET for baseline/hindcast; climate-driven runs used bias-corrected climate inputs. Model performance: <9% difference in daily mean streamflow vs observations; Nash–Sutcliffe efficiencies >0.58 (log 0.74), correlations r >0.77, percent bias between -4 and 10. The model reproduced IAT hydrological impacts consistent with paired-catchment evidence. Attribution metrics and uncertainty: For each catchment and climate experiment, ensembles of 145 hydrological simulations per climate–IAT state were generated. Two primary metrics quantified impacts: (i) QR% = (Q_comparator / Q_reference) × 100, where the reference is NC; its inverse % Change in Q = ((Q_comparator − Q_reference)/Q_reference) × 100; (ii) QR% point difference = QR%(ACL/NC) − QR%(AC/NC) for clearing benefit, and QR%(AI/NC) − QR%(AC/NC) for invasion impact. Uncertainty was assessed via bootstrap percentile 95% confidence intervals by resampling drought-period mean streamflow 1,000 times per state and experiment, propagating through metric calculations. A multi-experiment synthesis combined medians with equal weighting and pooled standard deviations to derive overall 95% confidence intervals.

- ACC reduced drought-period streamflow: Multi-model synthesis QR% was 83% (95% CI: 78–88) for Upper Berg and 78% (95% CI: 71–85) for Du Toits, corresponding to -17% (95% CI: -22 to -12) and -22% (95% CI: -29 to -15) relative to Natural climate with current IAT levels. - Streamflow impacts exceeded rainfall reductions: ACC reduced drought-period rainfall to 85–93% (95% CI across catchments) of Natural, while streamflow reductions were larger (71–88%). Hydrological processes amplified the meteorological ACC signal. - Small but attributable increase in reference evapotranspiration: Multi-model RefET ratio was 101.84% (95% CI: 101.69–101.99) under ACC. Sensitivity tests indicated no confident detection that RefET changes materially altered the ACC impact on streamflow for the drought period. - Clearing IATs ameliorates ACC impact where invasion is substantial: In Du Toits (40% invaded), pre-drought clearing would have increased QR% from 78% (95% CI: 71–85) to 87% (95% CI: 80–94), a +9 percentage-point gain (95% CI: +3 to +15), i.e., an attributable NbS effect. In Upper Berg (9% invaded), clearing yielded a non-detectable change: +1% (95% CI: -4 to +6). - Lack of maintenance and full invasion would have greatly exacerbated reductions: Under Actual climate with fully invaded catchments, QR% was 69% (95% CI: 64–74) for Upper Berg and 57% (95% CI: 52–63) for Du Toits, i.e., -31% (95% CI: -36 to -26) and -43% (95% CI: -48 to -37) relative to NC. Relative to current invasion levels with ACC (AC), full invasion would further reduce realized Natural-climate streamflow by -14 (95% CI: -19 to -10) and -21 (95% CI: -27 to -14) percentage points for Upper Berg and Du Toits, respectively. - NbS cannot fully offset ACC: Even with complete clearing, ACC effects remain: QR% with clearing was 84% (95% CI: 79–90) for Upper Berg and 87% (95% CI: 80–94) for Du Toits; 100% is not reached within 95% CIs. - Drivers of hydrological drought: Reduced precipitation, not increased RefET, was the dominant ACC driver of the 2015–2017 drought-period streamflow deficits, consistent with prior regional analyses of meteorological drought.

The study demonstrates, via a joint event attribution framework that links climate and hydrology, that ACC measurably reduced drought-period streamflow in key headwater catchments supplying Cape Town, and that managing invasive alien trees can moderate this impact. The ACC signal in rainfall is amplified through catchment processes into larger proportional reductions in streamflow, highlighting the sensitivity of mountainous, water-limited systems to meteorological droughts. Clearing IATs yields detectable benefits where invasion is moderate-to-high (e.g., Du Toits at 40% invasion), whereas benefits are smaller or not statistically detectable where invasion is already low (e.g., Upper Berg at 9%). Preventing spread and maintaining low invasion levels avoids substantial additional losses during droughts. However, even complete clearing cannot fully counteract ACC impacts, indicating NbS should be integrated within broader adaptation portfolios for water security. The findings reinforce global evidence that afforestation or woody encroachment in ecosystems naturally dominated by shorter vegetation often reduces water yields, cautioning against extensive tree planting in such regions without careful assessment. Strategically focusing NbS on critical headwater catchments and heavily invaded areas is likely to yield higher hydrological dividends. The dominance of precipitation changes over evapotranspiration in driving the 2015–2017 deficits aligns with regional and global projections indicating precipitation-driven increases in drought risk in the Western Cape.

The paper provides quantitative attribution of ACC impacts on drought-period streamflow during the 2015–2017 Cape Town drought and shows that NbS—specifically invasive alien tree clearing—can reduce, but not eliminate, these impacts. By operating at decision-relevant catchment scales and integrating climate and hydrological modeling, the study demonstrates that vegetation management in headwater catchments is an effective adaptation measure for mitigating hydrological consequences of ACC-driven meteorological droughts. The results have broader relevance for water-limited regions facing woody encroachment, supporting caution against extensive tree cover expansion in ecosystems with naturally shorter vegetation. Future research should assess how the contribution of NbS evolves under progressively warmer and drier (or wetter) climates, identify conditions under which human influence on climate exceeds the potential of NbS to reduce hydrological drought impacts, and evaluate diverse NbS types across contexts. NbS must be integrated into comprehensive adaptation portfolios to sustain water security as climate change intensifies.

- Scope and context specificity: The analysis focuses on two mountainous catchments in the southwestern Cape; NbS outcomes are context-specific and may not generalize across all regions or ecosystems. - Evapotranspiration representation: The study assesses reference evapotranspiration (derived via Hargreaves–Samani) and does not simulate actual evapotranspiration; sensitivity tests suggest limited influence of RefET changes on drought-period streamflow for this event. - Climate model resolution and analogs: Climate model inputs are much coarser than the hydrological model grid and require quantile–quantile bias correction. For CMIP5, the drought was represented using driest 3-year analogs (1869–1899 vs 2001–2031) rather than the exact 2015–2017 sequence. - Residual ACC signal: Even with complete clearing, ACC-attributable reductions persist, indicating NbS alone cannot fully offset impacts under current conditions. - Model-based inference: Hydrological and climate modeling with bootstrapped confidence intervals underpins results; uncertainties remain inherent to model structures, parameterizations, and input data.