Global patterns of soil available N production by mineralization-immobilization turnover in the tropical forest ecosystems.

Available N (N) is important to nurish plant-microbial system and sequestrate carbon (C) in terrestrial ecosystems. For forest ecosystem, N is usually calculated as the sum of N fixation (N), N deposition (N) and soil available N production (N), in which N determined the N production and its temporal changes. While, there is still a lack of N estimation at the global and regional level due to the temporal and spatial variability of influencing factors, such as climate and soil physicochemical properties. By assembling a dataset of gross rates of soil N mineralization (GR), immobilization of ammonium (NH) (GR) and nitrate (NO) (GR), as well as their corresponding geographic information, climate and main soil physicochemical properties, the N produced from organic N (N) mineralization and inorganic N (N) immobilization turnover (MIT) was calculated via building a random forest (RF) model in global tropical forests. The results revealed a good fit between the observed and predicted GR (R = 0.76), GR (R = 0.77) and GR (R = 0.67). We further estimated that the total mineralized N, immobilized NH and NO was 23.97 (10.48-37.46), 17.98 (5.81-30.15) and 4.86 (1.46-8.26) Pg N year, respectively, leading to the total N of 1.13 (-0.95-3.21) Pg N year. Referring to the reported average density of N and N, the total N and N was 0.03-0.05 and 0.01 Pg N year, respectively, by producting density and square meter of global tropic forest. Then the total N of global tropic forest ecosystem was 1.18 (-0.91-3.27) Pg N year (N + N + N). According to the tight stoichiometric relationship between C and N in the production of gross primary productivity (GPP) and soil respiration (R), C:N ratio of 31.8-41.9 and 22.7-48.2 was calculated, respectively, which all fall into the C:N ratio range of plants and litter (13.9-75.9) in tropical forest ecosystem. These results confirmed the prediction of N production from MIT was in line with theoretic estimates by applying RF machine learning. To our knowledge, this is the first estimation of N and the results provide the theoretical basis to evaluate soil C sequestration potential in tropical (e.g. southern America, southeast Asia and Africa) forest ecosystem.