Natural uncertainty of spatial average aquifer recharge through atmospheric chloride mass balance in continental Spain

Abstract

In a previous paper, the atmospheric chloride mass balance (CMB) method for spatial average diffuse aquifer recharge by rainfall (R‾) in large and varied territories was evaluated. Continental Spain was chosen to show the reliability of this application. Two main sources of uncertainty (measured by the coefficient of variation, CV) affecting R‾, induced by the inherent natural variability of the variables (CVR) and from mapping (CVRK), were separated. While CVRK may be decreased with better data coverage, the part of CVR inferred by the variable length of yearly data series can be corrected by comparing them to existing long series. With the same data sets and methods as in a previous paper, a data-correction procedure to improve R‾ and CVR is presented. The critical balance period (N) to reach comparable steady (long-term) CMB averages and CV values (CVs) was defined. The correction considered the timing of short series to incorporate the yearly data oscillating trend from longer series, additional trends deduced from longer series having incomplete N-year cycles, and changing stationary parameters over space from several long series. In continental Spain, N = 10 years. This period coincides with the decadal global climatic cycles acting on the Iberian Peninsula. Corrected CMB averages and CVs were regionalized by ordinary kriging using the same 10 km × 10 km 4976-nodes grid used for mapping the original CMB components. Nodal R‾ values varied from 17 to 715 mm year–1, 90% ranging from 35 to 300 mm year–1. Data correction did not significantly change averages; corrected average R‾ was 2% higher. However, the CVs changed conspicuously. The average CVR was 0.29, more than twice the original 0.13. This more realistic estimation of CVs avoids the illusory precision obtained when the short series correspond to small change periods. The average CVRK was 0.09, similar to the original 0.07. The improvement is shown by the CV of CVR and CVRK, which notably decreased below 0.1, even when their absolute values increased.