Deforestation amplifies climate change effects on warming and cloud level rise in African montane forests

Tropical montane forest ecosystems are pivotal for sustaining biodiversity and essential terrestrial ecosystem services, including the provision of high-quality fresh water. Nonetheless, the impact of montane deforestation and climate change on the capacity of forests to deliver ecosystem services is yet to be fully understood. In this study, we offer observational evidence demonstrating the response of air temperature and cloud base height to deforestation in African montane forests over the last two decades. Our findings reveal that approximately 18% (7.4 ± 0.5 million hectares) of Africa’s montane forests were lost between 2003 and 2022. This deforestation has led to a notable increase in maximum air temperature (1.37 ± 0.58 °C) and cloud base height (236 ± 87 metres), surpassing shifts attributed solely to climate change. Our results call for urgent attention to montane deforestation, as it poses serious threats to biodiversity, water supply, and ecosystem services in the tropics.


Estimating cloud water interception efficiency loss due to deforestation
We computed the spatial distribution of cloud water interception efficiency (CWI) loss using Katata et al. (2008) method, which was developed for montane cloud forest.The CWI, which refers to the deposition velocity of cloud water, depends on the vegetation structure (Leaf area index and canopy height) and can be estimated using the below equation.The equation was reported to have good correlation (R 2 =0.928) against numerical experiment (Katata et al., 2008).

Supplementary Figure 3 :
Contribution of climate change and deforestation to maximum air temperature (ΔTair) change in tropical montane forest of Africa across latitude from South to North at ~ 10 km interval.For climate change-induced ΔTair, 30 years of ERA5-land Tair data between 1992 to 2022 were used.Deforestation-induced changes were between 2003 and 2022.The blue, red, and black line show the average ΔTair due to deforestation, climate change, and combined effect from these two factors.The shaded region show mean ± standard deviation.

Supplementary Figure 4 .Supplementary Figure 9 .
area density in m2/ m3; LAI= leaf area index (m2/ m2); H= canopy height (m).We used H data at 30 m spatial resolution for the year 2000 and 2020 fromPotapov et al. (2022).The H data was prepared by integratig Landsat analysis-ready data, which is valid for time series analysis, and Global Ecosystem Dynamics Investigation (GEDI) Lidar based on machine learning ensemble approach(Potapov et al., 2021(Potapov et al., & 2022)).For the corresponding forest loss pixels indetified in the main manuscript during 2003 -2022, we calculated ΔH= H (2000) -H (2020).Since our forest loss anaysis excluded pixels with tree cover losses occured before 2003, the H (2000) is applicabe to H (2003).However, there could be underestimation in ∆CWI due to the two years difference between H (2020) and the forest loss period (i.e., 2003 -2022).For leaf area index (LAI), we used the 2003 average LAI from Moderate Resolution Imaging Spectroradiometer (MODIS) MCD15A3H Version 6.1 product, which is a 4-day composite product combined from Aqua and Terra sensor at 500 m resolution (Myneni et al., 2021).We assumed near zero (0.1) LAI after forest loss (i.e., in 2022) to reduce uncertainty arising from the LAI product (Pu et al. 2020).Panel (a) shows deforestation-induced cloud water interception efficiency (CWI) reduction across montane forest in Africa.Higher value shows stronger reduction in CWI.Montane forest extent (see Methods in the main manuscript) is used as a background.Panel (b) shows the deforestationinduced mean ΔCWI and standard deviation (SD) across 1° latitude intervals.Closer views of ΔCWI over Ethiopia (c), Kenya (d), and central to western parts of eastern Africa (e) are displayed for boxes (1-3) in Panel (b).Administrative boundary data are from Global Administrative Areas (GADM) (2015 -2022).Comparison of predicted and observed monthly minimum air temperature (Tmin) for each month between 2003 and 2022 in montane forest in Africa.Model training and validation data is from the global summary of the day (GSOD) station data over Africa (NOAA National Centers of Environmental Information ,1999).

Table 1 .
Supplementary Figure10.Comparison of predicted and observed monthly dew point temperature (Tdew) for each month between 2003 and 2022 in montane forest in Africa.Model training and validation data is from the global summary of the day (GSOD) station data over Africa (NOAA National Centers of Environmental Information ,1999).Model performance tested against validation data for Tmax using R 2 , RMSE, and MAE.

Table 2 .
Model performance tested against validation data for Tmin using R 2 , RMSE, and MAE.

Table 3 .
Model performance tested against validation data for Tdew using R 2 , RMSE, and MAE.