METHODS
This study presents an approach to extract, quantify, and explain three temporal patterns within SC time series, which are related to streamwater responses to rain events. These patterns are found in chemographs produced from in situ sensors across numerous New Hampshire lotic systems (Inserillo et al. 2017). The first of these patterns is a rapid increase in SC near the beginning of a precipitation event, on the rising limb of a hydrograph, known as a solute flush (Figure 1). The second is the subsequent dilution of SC, as low SC water (i.e., precipitation) mixes with higher SC water (i.e., groundwater); and the third is the recovery of SC as that low conductivity water leaves the catchment and the streamwater returns to a SC resembling groundwater. Although flushing responses do not occur in all streams, dilutions and their subsequent recoveries after a precipitation event are common (e.g., Inserillo et al. 2017).
We developed an algorithm to consistently and objectively locate these temporal patterns in streamwater SC, quantify the FSCand DSC behavior, and calculate the RSCthat follows. This process was sequential: a flush following the onset of rain, a dilution following a flush, if one was present, and a recovery following a dilution, if one was present. We then compared those streamwater responses to various environmental conditions surrounding the periods in which they occurred, to understand how such independent variables related to each chemograph pattern. The environmental conditions examined were based upon precipitation, antecedent moisture, and the season in which the precipitation event occurred.
Precipitation variables included the total amount of precipitation (PT), the intensity at which it fell (IP), and maximum precipitation intensity per 20-minute interval (IP,max). Precipitation that occurred after a flush reached its peak would not influence the FSC; likewise, with the maximum dilution and DSC. So, for both flushing and dilution responses, each precipitation variable was analyzed only from the beginning of the storm to the point in time at which the response occurred. Analysis of the RSCincluded the precipitation of the entire event, as the tail end of a storm presumably impacts the rate at which SC recovers, and as the recovery is an expression low SC water leaving the system or becoming chemically similar to stored water.
Antecedent moisture variables included the length of the inter-storm period (ISP) and the cumulative vapor pressure deficit (ΣVPD) during that period of time. The ISP was determined by the amount of time between the first measurement of rain for a precipitation event and the last measurement of rain for the previous event. The ΣVPD for that period of time was indicative of the intensity of drying pressure and was determined by summing the vapor pressure deficit (VPD) over this ISP. VPD was calculated by subtracting vapor pressure (VP) from saturated vapor pressure (VPS). VP was calculated by multiplying relative humidity (RH) by VPS. VPS was calculated using
where T was air temperature in °C, and the resulting VPSwas in kPa (Bolton 1980). These temperature and RH data were measured at 20-minute intervals, and all subsequent calculations maintained this time step.
Variables of seasonal impact were the day of year (DoY) on which the precipitation event began and the median air temperature (TA) of the day on which the event began. In both instances the day on which the event began was the calendar day that the first rain measurement of that precipitation event occurred.