Aquatic DOM, a major carbon pool in inland waters, plays a key role in the global carbon cycle and is influenced by human activities and environmental factors. While watershed surveys have identified some drivers of aquatic DOM changes, national-scale knowledge remains limited. Spectral monitoring data from 721 aquatic DOMs in inland China were collected to predict and quantify the spectral properties of aquatic DOMs and the coupled effects of human activities and environmental factors by complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), random forest (RF) and structural equation modeling (SEM). Our results show that the dataset is relatively evenly distributed in terms of flow regimes, and the data are adequately representative of various types of water in China. Although the performance of CEEMDAN is affected by the skewed distribution of data sampling points (453 and 449 for eastern regions and eutrophic waters, respectively), CEEMDAN effectively processes the spatial sequences of aquatic DOM spectra, which significantly improves the reliability of the machine learning model (by 13%–68%) and reduces the error (by 63%–92%). CEEMDAN-RF and SEM results indicate that longitude and latitude are the most important environmental factors affecting the spectral properties of aquatic DOM through temperature, light and land cover differences, reducing the biogenic properties of aquatic DOM. Unlike observations in small watersheds and estuaries, environmental factors of season, precipitation and salinity have weak effects on aquatic DOM. Furthermore, the biogenic character of aquatic DOM is enhanced by urban human activities, represented by urbanization, population, and impervious land fraction, which have long-lasting and strong impacts on aquatic DOM by driving water eutrophication and enhancing phytoplankton and microbial activity. Our study informs the prediction and quantification of coupled environmental and anthropogenic impacts on aquatic DOM at large scales.
Read full abstract