Correlation Analysis between Time Series of Precipitation and Soil Moisture under a Mediterranean Climate

Abstract

Rain-fed agriculture in Mediterranean-type environments relies on seasonal precipitation for agro-forestry production. Irregular precipitation rates (P) do influence the spatial-temporal soil-water (SW) distribution to a “critical soil depth” for a “quick” response of the active roots to tap water following a recent precipitation event. We investigated the statistical correlations between time-series of precipitation and soil-water data at depth using the Canonical Correlation Analysis (CCA) technique. Soil-water was measured in a Cambisol Dystic of an experimental plot, located on a smooth hillside in a non-ploughing Portuguese “montado” agro-forestry ecosystem, in a watershed of representative regional geomorphology and land-use of the Alentejo-region, southern Portugal. Soil-water was measured using three time-domain reflectometry (TDR) moisture-sensors since May 12th (2011) at 10-cm, 30-cm and 50-cm depths. Precipitation was measured with a standard rain-gauge (0.1 mm/tip), ca. 8-km of the TDR´s site. We performed CCA on the time-series of hourly data of P and SW recorded from May, 13th to November, 30th. Firstly, CCA involved P and the three variables (each per depth) for SW, considered all together in the model, and, then, using each SW-source variable at a time, for paired CCA´s. Secondly, the goodness of CCA´s was evaluated on how precipitation impacts correlates with soil-water, by either considering simultaneous triggering or by introducing time lags of 3 or 4 hours for all sensors. As to the initial condition SW=0.013 m3 m-3, results showed the best correlation between the temporal patterns of P “versus” that of SW at depth, by setting a common time lag of 4 h. Under the observed precipitation regime, the deeper TDR sensor did not respond at all to the precipitation forcing, probably due to insufficient accumulated P (196 mm) at 1.02 mm/day. Future work on this topic should focus on longer time-series, reinforcing the analysis of precipitation events as well.