Soil organic carbon stocks controlled by lithology and soil depth in a Peruvian alpine grassland of the Andes

Abstract

The soil is the largest carbon (C) pool in the terrestrial ecosystem, and soil organic carbon (SOC) stocks play an important role in global C dynamics. Alpine grasslands of the Andes are characterized by high SOC stocks. Quantifying SOC stocks and unraveling key factors controlling SOC stocks, is necessary to obtain a better understanding of the dynamics of the large C stocks in this environment. However, most studies on C dynamics of the Andes focus on volcanic-ash soils, whereas information about non-volcanic ash soils in this region is scarce. Our objectives were: (i) to estimate SOC stocks in an alpine grassland of the Peruvian Andes (7° 11′S, 78° 35′W) with parent materials other than volcanic ash, and (ii) to identify the underlying soil formation and environmental (SFE) factors and soil properties explaining observed patterns of SOC stocks. We sampled 69 plots up to the parent material to measure soil properties and to calculate SOC stocks, in relation to lithology, land use, grazing intensity, slope angle, slope position and altitude. We applied linear models to identify key factors controlling SOC stocks. Our results showed that total SOC stocks had a mean value of 215 ± 21 T ha−1, whereas SOC stocks of the upper 10 cm and 40 cm comprised 29.3% and 80.0% of total SOC stocks respectively. The variation of the total SOC stocks was mainly explained by soil depth and soil moisture. When soil depth and soil moisture were controlled as conditional variables, lithology became the key factor controlling the total SOC stocks. For the SOC stocks of the upper 10 cm, soil moisture explained a large part of the variation, whereas lithology, grazing intensity and altitude were also significant predictors. Our results also show that when soils are sampled with limited depths instead of the entire soil profile, SOC stocks can be underestimated, and the effects of the SFE factors on SOC stocks can be overestimated.