Abstract

We compiled maps for the topsoil (0–30 cm) organic carbon (SOC) stock and its prediction uncertainty in Hungary at 100 m resolution for the years 1992 and 2010 using a machine learning algorithm, namely, quantile regression forest. 10-fold cross-validation was used for checking the accuracy of the spatial predictions and uncertainty quantifications for both years. The performance of the spatial predictions and uncertainty quantifications was appropriate, which was verified by the computed biases (0.15 and 0.30 for 1992 and 2010), root mean square errors (21.99 and 21.39 for 1992 and 2010), accuracy plots and the G statistics (0.96 for both years) as well. Based on the compiled SOC stock maps, we assessed the spatio-temporal change of SOC stocks on the territory of Hungary. A scheme was elaborated based on the quantified prediction uncertainties for identifying and delimiting significant and tendentious changes of SOC stock during the respective period. The total SOC stock in the topsoil was found to be 424.41 Tg (1 teragram = 1012 grams) in 1992 and 451.59 Tg in 2010. Thus SOC stock increased by 27.18 Tg over the respective period. On those areas where the land use types did not change, we observed that the SOC stock increased under forests (by 16.29 Tg) and pastures (by 2.48 Tg), decreased under wetlands (by 0.49 Tg) and did not change under agricultural areas. On those areas where the land use has been changed during the 18-year period, we found that afforestation has increased the SOC stock, whereas cultivation of pastures has decreased it. Due to soil sealing 34,000 ha of soil have been lost resulting in approximately 1.7 Tg carbon loss. We compared our own total SOC stock estimate and map referring to 1992 with other estimates and maps provided by global and continental initiatives. The comparisons have pointed out that the SOC stock map of 1992 outperformed these maps. We recommend applying the SOC stock map of 1992 as a baseline for Hungary.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call