1. Introduction
For water-dependent organisms, hydroperiod – or the amount of time a
lentic waterbody holds water – plays a critical role in population and
community dynamics (De Meester et al., 2005). Dispersal decisions
(Tournier et al., 2017), fitness (Johnson et al., 2013; Rogers &
Chalcraft, 2008), reproductive success (Ryan & Winne, 2001), survival
(Acosta & Perry, 2001), and source-sink dynamics (Ruetz III et al.,
2005; Werner et al., 2007) are all influenced by hydroperiod.
Hydroperiod is also an important predictor of community composition
(Razgour et al., 2010; Skelly, 1997; Waterkeyn et al., 2008) and
diversity (Schriever et al., 2015; Schriever & Williams, 2013; Stendera
et al., 2012), and thus may influence ecosystem stability. Specific
components of hydroperiod, such as inundation timing or stability,
influence species density and richness (Florencio et al., 2020; Kneitel,
2014) and may play a key role in determining reproductive success of
aquatic organisms such as amphibians (Paton & Crouch III, 2002).
The wide-ranging effects of hydroperiod on individual organisms,
populations, and ecological communities necessitate tools to enable
fine-scale measurement and monitoring of the timing, frequency, and
duration of hydroperiod events in temporary lentic waters. Such tools
will play an important role in predicting how hydroperiods may change in
response to future water use and climate scenarios – and how organisms
that rely on these habitats will fare. Satellite remote sensing tools
such as Synthetic Aperture Radar (Bourgeau-Chavez et al., 2005; Hong et
al., 2010) and Landsat imagery (DeVries et al., 2017; Díaz-Delgado et
al., 2016; Murray-Hudson et al., 2015) enable wetland hydroperiod
assessment over multi-year or multi-decade periods and covering 10s to
1000s km2. The accuracy of these methods continues to
improve with advances in image analytical techniques that provide
information on surface water presence and area at a sub-pixel level
(Halabisky et al., 2018). Unmanned aerial systems (i.e., drones) can
also provide high-resolution spatial and temporal data, particularly for
specific regions (e.g. Levy & Johnson, 2021). Despite these promising
advances, the temporal grain of remotely sensed data remains coarse for
most remotely sensed datasets. For example, Landsat captures images at a
spatial resolution of 30 meters every 16 days (Irons et al., 2012;
Ozesmi & Bauer, 2002). In many regions, inundation of intermittent
lentic habitat may occur over hours or days. Temporal resolution on the
order of 2-4 weeks may miss important fine-scale differences in pond
inundation timing, particularly in regions with unpredictable spatial
patterns of precipitation that drive a patchwork of inundation dates.
More recently, new earth observation products continue to refine the
temporal grain size of satellite data, with some providing daily or
sub-weekly observations (Lefebvre et al., 2019). Spatial resolution of
new commercial products offer sub-meter resolution in some cases (e.g.,
Planet Team 2021). Though these products do not have the historical
breadth of Landsat, they offer much higher spatial and temporal
resolution for more recent years. However, challenges remain. Some
remotely-sensed hydroperiod data can be obscured by cloud and/or canopy
cover, decreasing temporal resolution from a few days to longer periods.
This may be particularly challenging if inundation timing co-occurs with
periods of high cloud cover. Though commercial products are often
available at no charge for researchers with particular affiliations,
data in these cases is often limited in format, spatial extent, or
resolution. For full access, project budgets often must accommodate
product costs.
Temperature and conductivity sensors are used increasingly in both lotic
and lentic systems to provide fine-scale spatial and temporal
hydroperiod measurements (Anderson et al., 2015; Arismendi et al., 2017;
Jaeger & Olden, 2012). Daily temperature variance is typically lower in
water than in air, and comparison of daily temperature variance provides
a reliable proxy for inundation state (Sowder & Steel, 2012). A rapid
drop in daily temperature variance can reliably measure the precise
timing of an inundation event (Anderson et al., 2015; Arismendi et al.,
2017). For example, Anderson et al. (2015) tested the ability of
temperature sensors to accurately predict inundation states both in
natural wetlands and in controlled mesocosms. The authors deployed
temperature sensors for two six-month periods in ponds over a 7140 ha
area that varied in size and depth. They demonstrated that daily
temperature variance reflected pond filling and drying events, with
higher variance in dry ponds and in control sensors placed on the ground
outside of ponds, and they determined an approximate variance threshold
to predict inundation states. Arismendi et al. (2017) placed paired
temperature sensors and electrical resistors in temporary streams and
found that using daily temperature standard deviation more accurately
predicted inundation states than mean hourly or daily temperature
measurements.
Hidden Markov models (HMMs) can be used to identify shifting trends in
time series data (e.g. high temperature variance associated with dry
states and low temperature variance associated with wet states) while
accounting for temporal autocorrelation and are useful tools for
modeling climatic data (reviewed in Srikanthan & McMahon, 2001).
Arismendi et al. (2017) demonstrated the ability of these algorithms to
predict wetland inundation states and found that fitting 2-state HMMs to
temperature standard deviations led to accurate predictions of shifts
from wet to dry states in ephemeral streams.
Here, we describe methods for deployment and data analysis of an array
of temperature loggers to monitor inundation state of intermittent ponds
in the San Rafael Valley of Arizona, USA. These methods are applicable
to a range of temporary lentic habitats, especially where logistical
challenges may necessitate use of on-the-ground measurements rather than
remotely sensed data. The objectives of this study were: 1) design a
sturdy, low cost, and low maintenance housing unit for temperature
sensor deployment in remote and rugged terrain; 2) deploy paired sensors
(one within the target pond and one outside the pond) to monitor
hydroperiod inundation states in temporary ponds; 3) evaluate inundation
states using HMMs, comparing inundation date inference between 1-logger
(pond only) and 2-logger (pond + control) experimental design; 4)
compare observed and inferred inundation state recorded during in-person
visits to ponds. Overall, our findings point to the utility of
temperature loggers as a cost-effective, low profile tool in uncovering
ecologically relevant spatiotemporal differences in intraregional
inundation timing. This is particularly useful in regions with highly
localized precipitation events that drive small-scale differences in
spatiotemporal hydroperiod dynamics.