Estimation method for karst carbon sinks on the basis of a concentration prediction model

Karstification can reduce the CO₂ concentration in the atmosphere/soil. Accurate estimation of karst carbon sinks is crucial for the study of global climate change. In this study, the Lijiang River Basin was taken as the research area. On the basis of the measured dissolved inorganic carbon (DIC) and dissolved organic carbon (DOC) concentration data from 14 consecutive months, the relationships of DIC and DOC to elevation, slope, aspect, rainfall, and temperature were established. Among six regression algorithms, the random forest (RF), boosted regression tree (BRT) and BP neural network (BP) were selected for stacking integration to construct DIC and DOC concentration prediction models, achieving accuracies of 91% and 83%, respectively. On the basis of these models, the spatial and temporal distributions of DIC and DOC concentrations in the Lijiang River Basin from 2000 to 2022 were predicted. The prediction results reveal that DIC and DOC concentrations have a stable spatial distribution, which is consistent with the lithology distribution in the basin. The solute load method was used to estimate the karst carbon sink in the Lijiang River Basin over 23 years. The carbon sink over 23 years showed an overall growth trend, although with significant fluctuations. On the basis of the estimation results of karst carbon sinks over 23 years, a time series prediction model is used to predict the Lijiang River Basin from 2023 to 2030. The prediction results continue the volatility and trend of the historical data. A comparison of the model verification results with related research findings revealed that the concentration prediction model constructed in this study has high accuracy and good applicability in the estimation of karst carbon sinks at the watershed scale.