Sensitivity assessment of metafrontier data envelopment analysis for soil carbon sequestration efficiency

• Main Authors: Bacon, A.R (Author), Cropper, W.P., Jr (Author), Grunwald, S. (Author), Mizuta, K. (Author), Moss, C.B (Author), Phillips, M.A (Author)
• Format: Article
• Language: English
• Published: Elsevier B.V. 2021
• Subjects: agricultural soil; Agronomy; carbon sequestration; Data envelopment; data envelopment analysis; Data envelopment analysis; Efficiency; Efficiency benchmarking; Efficiency index; Florida; Florida [United States]; Forestry; Land use; Land use/land cover; Land-cover types; Metafrontier; Metafrontier model; Metafrontier models; Multivariate normal distribution; Multi-variate normal distributions; NDVI; Normal distribution; organic carbon; Organic carbon; pasture; Return to scale; Sensitivity analysis; soil analysis; soil carbon; Soil carbon sequestration; soil management; soil water; Soils; United States; Vegetation; Wetlands
• Online Access: View Fulltext in Publisher

 The data envelopment analysis (DEA) is useful for the measuring of efficiencies of functions in various research fields. In previous studies the soil carbon sequestration (SCseq) efficiency index (SCI) was assessed at regional scale; however, the effects of various calculation factors—including returns-to-scales (RTSs), selection of orientation (input or output), and global (or meta-)/local (or group) frontiers—on the SCI scores are still unclear. Thus, in this paper, the sensitivity of DEA frontier models with various types of RTSs and orientations was investigated. The models examined the SCseq rates as target outputs in three land use/land cover (LULC) types (pinelands, improved pastures, and wetlands). The rates were calculated based on lab-measured soil organic carbon concentration collected from the topsoil (~20 cm depth) between 1965 and 2009 in Florida, United States. Pedogenic, hydrologic, and environmental variables relevant to the carbon sequestration function were chosen as inputs, including soil available water capacity, sand concentration, pH, soil total nitrogen and phosphorus, Enhanced Vegetation Index, and Normalized Difference Vegetation Index. The input and output variables for the three LULC types informed multivariate normal distributions that served to simulate the local frontiers. Results show that the SCseq rates in wetland soils were statistically lowest (p<0.05), while the SCI scores were highest. This suggests that the efficiency of the soil carbon sequestration function was distinctively different from the SCseq rates. The SCI scores also demonstrated to identify internal causes (i.e., local management practices) for inefficiency in the SCseq in pasture soils and suitability of the function among the LULC types. A higher ability in output orientation to discern efficient from inefficient sites for the SCseq function was also observed than the input orientation. The preferred choice of RTSs in either conservative or progressive approaches to SCseq efficiency management was also identified in the study. Sensitive scores reflected regions that are in need for efficient soil management (such as no-tillage) to improve the performance of individual soils in terms of SCseq efficiency. SCseq efficiency scores enable land resource managers to refine their approaches to optimize soil carbon and possibly other soil and ecological functions. Further research is invited to identify potential cause(s) that determine(s) the local/global frontiers. © 2021 The Authors