CHARACTERIZING FLOW BEHAVIOUR IN KARST SYSTEMS USING TIME SERIES ANALYSES – BETARI VALLEY - BRAZIL

In the southern mountains of São Paulo State, Southeastern Brazil, karst aquifers occur on plateaus developed on folded Precambrian rocks. The climate in this region is classified as warm and wet, with no dry season, the mean annual rainfall is 1575 mm. The carbonate plateaus covered by Atlantic Rainforest are characterized by polygonal karst and receive important allogenic recharge from surrounding psammitic crests. In this region, Betari River valley cuts limestone belts and is the discharge zone of karst aquifers.

The water level of four springs (Santana – ST, Couto – CT, Água Suja – AS and Zezo – Z), and the rainfall are being monitored in Betari valley. Preliminary results (summer-autumn, 2009-2010) about aquifer’s hydrodynamics are presented using time series structure checking tools, like autocorrelation, lag plot and spectral Fourier analysis. The cross correlation function (CCF) is used to assess the relationship between rain impulse and the spring response.

According to the autocorrelation, spectral Fourier and lag plot analyses, rainfall data don’t show visible periodicities, the signal appears fairly random. The same analysis applied to the water level in the springs indicates the presence of an autoregressive model, there is high association between an observation and the succeeding one. The systems’ memories obtained with autocorrelation (time to reach r=0.2, daily data) are the following: Z - 27, ST - 35, AS - 36, CT – 40. The absence of structure in the rainfall signal and its existence in springs indicate that the system works as a filter. In this sense, the degree of transformation of the input signal provides information on the nature of the flow in the system; inertial systems have greater storage capacity.

The CCF between rainfall and water level in springs gives an estimation of the particle travel times through the aquifer. The spring CCF signal is stronger for AS and CT (0.34 and 0.33), with response times of 5 h and 4 h respectively. For ST and Z, the CCF signal value and response time are respectively 0.27, 10 h and 0.21, 18 h.

By the end of the monitoring stage, the hydrological time series will be analysed with tracer tests results and geological and geomorphic attributes; this information will be important to define protection zones for the karst aquifers.

Research sponsored by FAPESP Proc. 2009/05115-5