2.2 Sensors, housing units, and deployment
We selected a waterproof temperature logger with the capacity for battery replacement by the user for longevity (company: Onset, Bourne, MA, USA; model: HOBO Pendant, MX2201; diameter: 3.35 cm; temperature range: -20 ˚C to 50˚C; temperature precision: ± 0.5˚C; cost: $54.00 USD; data retrieval: Bluetooth, battery: user replaceable CR2032 3V lithium). Our study region is remote with rugged terrain, and deployed equipment is exposed to variable weather, UV exposure, and potential tampering from humans, wildlife, or livestock. The intermittent ponds in our study region are visited frequently by cattle, and equipment must be able to withstand trampling or tampering. With this in mind, we designed a rugged housing unit to protect temperature loggers from damage and ensure long-term durability (Figure S1). We placed a logger inside a PVC junction box (hereafter called the housing unit) with two nuts between the box and the lid for increased air or water flow. The logger moved freely inside the housing unit to increase the chance that it remained submerged (i.e., fell to the lowest point within the housing unit) if disturbed after deployment. The housing unit was connected to a concrete tie or other secure post (e.g., a metal fence post marking edges of allotments) via a 3/32” (2.381 mm) galvanized, uncoated steel cable strung through the holes of the junction box. We fastened the cable by swaging a crimping sleeve. We provide a complete list of specifications for tools and materials in Table S1.
At each of the 16 ponds, we deployed one logger at the approximate deepest point of fill within the tank (the pond logger) and one logger approximately 10 m outside of the high-water mark for the pond (the control logger). Where possible, we placed control loggers in sunny, shade-free areas in order to most closely match conditions and exposure of the pond logger. If the pond basin consisted of fine clay or silt, we placed the housing unit on a flat rock partially buried to sit flush with the ground and to avoid it becoming buried in silt upon pond inundation. We then secured the housing unit to an existing fence post (typically a metal T post) or to a concrete tie using steel cable looped through the housing. We used a mallet to drive concrete ties completely into the ground for protection of livestock. Finally, we covered units with loosely stacked rocks to minimize livestock tripping risk and to help camouflage units to avoid tampering (Figure S2). Loggers recorded temperature at 15-minute intervals with Bluetooth set to manual (i.e., not continuously seeking a signal), resulting in an estimated 3.2-year battery life for each logger. We visited ponds three times after sensor deployment: 31 Jul – 2 Aug 2018, 31 Mar - 3 April 2019, and 21 – 27 June 2019 (time of data retrieval). During each site visit, we evaluated logger function, cleared any mud or sediment in the rugged housing units, and replaced disturbed rock piles (Figure S3).