Tropical forests are significant carbon sinks and their soils’ carbon storage potential is immense. However, little is known about the soil organic carbon (SOC) stocks of tropical mountain areas whose complex soil-landscape and difficult accessibility pose a challenge to spatial analysis. The choice of methodology for spatial prediction is of high importance to improve the expected poor model results in case of low predictor-response correlations. Four aspects were considered to improve model performance in predicting SOC stocks of the organic layer of a tropical mountain forest landscape: Different spatial predictor settings, predictor selection strategies, various machine learning algorithms and model tuning. Five machine learning algorithms: random forests, artificial neural networks, multivariate adaptive regression splines, boosted regression trees and support vector machines were trained and tuned to predict SOC stocks from predictors derived from a digital elevation model and satellite image. Topographical predictors were calculated with a GIS search radius of 45 to 615 m. Finally, three predictor selection strategies were applied to the total set of 236 predictors. All machine learning algorithms—including the model tuning and predictor selection—were compared via five repetitions of a tenfold cross-validation. The boosted regression tree algorithm resulted in the overall best model. SOC stocks ranged between 0.2 to 17.7 kg m^-2, displaying a huge variability with diffuse insolation and curvatures of different scale guiding the spatial pattern. Predictor selection and model tuning improved the models’ predictive performance in all five machine learning algorithms. The rather low number of selected predictors favours forward compared to backward selection procedures. Choosing predictors due to their indiviual performance was vanquished by the two procedures which accounted for predictor interaction.

The reconstruction of Holocene thermokarst landform evolution is important to understand the potential impact of current global climate change on permafrost regions. A multi-proxy approach was applied to analyse the sedimentological and biogeochemical characteristics as well as pollen and lacustrine microfossils of a core profile drilled in a small pingo within a large Central Yakutian thermokarst basin (alas). Age–depth modelling with macrofossil ¹⁴C ages reveals high thermokarst deposit sedimentation rates and a complete thermokarst sequence spanning about 900 years during the mid-Holocene between ~6750 and 5870 cal. yr BP. In total, three stages of thermokarst landscape evolution have been identified. Thermokarst processes were initiated at ⩽6750 to 6500 cal. yr BP. Terrestrial conditions changed quickly to lacustrine conditions, and a thermokarst lake rapidly emerged and grew to an estimated size of 120–600 m diameter and 7.5–15 m depth during only ~150 years between ~6500 and 6350 cal. yr BP. The decline of thermokarst processes and lake decrease may have been affected by local hydrological conditions between ~6350 and 5870 cal. yr BP but ceased completely after 5870 cal. yr BP, likely due to climatic changes. Clear evidence for long-lasting and stable lacustrine conditions was not obtained. The study emphasises that short-term warming led to very active permafrost degradation and rapid but locally variable modification of alas and thermokarst evolution.

The use of remote sensing techniques to identify (geo)archaeological features is wide spread. For archaeological prospection and geomorphological mapping, Digital Terrain Models (DTMs) on based LiDAR (Light Detection And Ranging) are mainly used to detect surface and subsurface features. LiDAR is a remote sensing tool that scans the surface with high spatial resolution and allows for the removal of vegetation cover with special data filters. Archaeological publications with LiDAR data in issues have been rising exponentially since the mid-2000s. The methodology of DTM analyses within geoarchaeological contexts is usually based on “bare-earth” LiDAR data, although the terrain is often significantly affected by human activities. However, “bare-earth” LiDAR data analyses are very restricted in the case of historic hydro-engineering such as irrigation systems, mills, or canals because modern roads, railway tracks, buildings, and earth lynchets influence surface water flows and may dissect the terrain. Consequently, a "natural" pre-modern DTM with high depth accuracy is required for palaeohydrological analyses. In this study, we present a GIS-based modelling approach to generate a pre-modern and topographically purged DTM. The case study focuses on the landscape around the Early Medieval Fossa Carolina, a canal constructed by Charlemagne and one of the major medieval engineering projects in Europe. Our aim is to reconstruct the pre-modern relief around the Fossa Carolina for a better understanding and interpretation of the alignment of the Carolingian canal. Our input data are LiDAR-derived DTMs and a comprehensive vector layer of anthropogenic structures that affect the modern relief. We interpolated the residual points with a spline algorithm and smoothed the result with a low pass filter. The purged DTM reflects the pre-modern shape of the landscape. To validate and ground-truth the model, we used the levels of recovered pre-modern soils and surfaces that have been buried by floodplain deposits, colluvial layers, or dam material of the Carolingian canal. We compared pre-modern soil and surface levels with the modelled pre-modern terrain levels and calculated the overall error. The modelled pre-modern surface fits with the levels of the buried soils and surfaces. Furthermore, the pre-modern DTM allows us to model the most favourable course of the canal with minimal earth volume to dig out. This modelled pathway corresponds significantly with the alignment of the Carolingian canal. Our method offers various new opportunities for geoarchaeological terrain analysis, for which an undisturbed high-precision pre-modern surface is crucial.