Pollen-based reconstruction of total land-cover change over the Holocene in the temperate steppe region of China: An attempt to quantify the cover of vegetation and bare ground in the past using a novel approach

•A novel strategy proposed to reconstruct the regional land-cover from pollen data.•The cover of bare ground was reconstructed as the main land cover type.•Actual plant cover and bare ground were reconstructed first-time using pollen data.•Fossil pollen percentages differ significantly from the reconstructed plant cover. Fossil pollen data are essential for reconstructing ancient vegetation and land-cover changes. Sugita’s REVEALS model is the best method to estimate regional plant cover (in percentage cover) using pollen data from lakes. Such reconstructions imply that the sum of all plants’ cover is 100%. However, land cover is not always represented by vegetation alone, the area of bare ground can be significant in many types of biomes, e.g., in alpine or steppe regions. Here we define “total land cover” as the sum of vegetation cover (VegC) and bare ground (BareC). In this study, we use the relationship between tree pollen percentages and both tree cover (TreeC) and VegC (=TreeC + herb cover (HerbC)) based on a dataset of modern pollen assemblages and related total land cover. This relationship is applied to estimate past “actual” vegetation cover (a-VegC) from fossil pollen percentages using the Modern Analogue Technique (MAT). The REVEALS (RV) model can then be applied to the same fossil pollen records to estimate regional cover of individual plant taxa (RV PlantC; e.g., RV PinusC, etc.), total tree cover (RV-TreeC) and total herb cover (RV-HerbC). These cover values can then be converted into RV a-PlantC, RV a-TreeC and RV a-HerbC using the MAT-reconstructed a-VegC (e.g., RV PinusC × MAT a-VegC = RV a-PinusC; RV-TreeC × MAT a-VegC = RV a-TreeC, etc.). The results of leave-one-out cross-validation indicates that the MAT reconstructions using the modern pollen assemblages provide values of a-TreeC, a-HerbC and BareC mostly very similar to the modern vegetation data. We further tested the method using pollen assemblages from lake surface sediments of 11 lakes and the results also suggest a good performance of MAT-based reconstruction. We then applied the proposed method (MAT-REVEALS) to four Holocene pollen records available from the study area to evaluate the feasibility of the strategy to reconstruct past actual plant cover. The results suggest that the method provides plausible estimates of vegetation cover for the sub-regions within the study area. The results from Lake Daihai over the last 10,000 years BP are interpreted and discussed in more deta