Spatial variability of topsoil δ13C across Qinghai-Tibet Plateau

<p>Soil carbon isotopes (&#948;¹³C) provide reliable insights at the long-term scale for the study of soil carbon turnover and topsoil &#948;¹³C could well reflect organic matter input from the current vegetation. Qinghai-Tibet Plateau (QTP) is called &#8220;the third pole of the earth&#8221; because of its high elevation, and it is one of the most sensitive and critical regions to global climate change worldwide. Previous studies focused on variability of soil &#948;¹³C at in-site scale. However, a knowledge gap still exists in the spatial pattern of topsoil &#948;¹³C in QTP. In this study, we first established a database of topsoil &#948;¹³C with 396 observations from published literature and applied a Random Forest (RF) algorithm (a machine learning approach) to predict the spatial pattern of topsoil &#948;¹³C using environmental variables. Results showed that topsoil &#948;¹³C significantly varied across different ecosystem types (p < 0.05).&#160; Topsoil &#948;¹³C was -26.3 &#177; 1.60 &#8240; for forest, 24.3 &#177; 2.00 &#8240; for shrubland, -23.9 &#177; 1.84 &#8240; for grassland, -18.9 &#177; 2.37 &#8240; for desert, respectively. RF could well predict the spatial variability of topsoil &#948;¹³C with a model efficiency (pseudo R²) of 0.65 and root mean square error of 1.42. The gridded product of topsoil &#948;¹³C and topsoil &#946; (indicating the decomposition rate of soil organic carbon, calculated by &#948;¹³C divided by logarithmically converted SOC) with a spatial resolution of 1000 m were developed. Strong spatial variability of topsoil &#948;¹³C was observed, which increased gradually from the southeast to the northwest in QTP. Furthermore, a large variation was found in &#946;, ranging from -7.87 to -81.8, with a decreasing trend from southeast to northwest, indicating that carbon turnover rate was faster in northwest QTP compared to that of southeast. This study was the first attempt to develop a fine resolution product of topsoil &#948;¹³C for QTP using a machine learning approach, which could provide an independent benchmark for biogeochemical models to study soil carbon turnover and terrestrial carbon-climate feedbacks under ongoing climate change.</p>