Untangling the Uncertainties in Plant Water Source Partitioning With Isotopes

Abstract Plant water source partitioning suffers from multiple uncertainty sources including tracers, depleted xylem water deuterium, and mixing models, but the uncertainties have rarely been systematically evaluated. Taking apple trees of different ages as example, we combined four tracers of different characteristics ( 2 H, 3 H, 17 O, and 18 O), two xylem water deuterium bias correction methods (soil water excess and stem relative water content), and four mixing models (IsoSource, SIAR, MixSIR, and MixSIAR) to quantify the total uncertainty and the uncertainty of each component. For the total uncertainty, the contributions of the mixing models were the largest (37%), followed by those of tracers (28%) and interaction of tracers and models (27%), while the contributions of other sources were the smallest (8%). In particular, the xylem water deuterium bias correction methods had an uncertainty of 2%, implying its minor role in deviating plant water source partitioning. After evaluating the performance of the three components with four target functions, we recommend tracer combination of 2 H 18 O with MixSIAR as the best framework for water source partitioning of apple trees on the Loess Plateau. The decomposed uncertainties and recommended methods provide technical support and promote understanding of variable results in plant water source partitioning.