Statistically robust monitoring of threatened populations is essential for effective conservation management because the population trend data that monitoring generates is often used to make decisions about when and how to take action. Despite representing the highest proportion of threatened animals globally, the development of best practice methods for monitoring populations of threatened insects is relatively uncommon. Traditionally, population trend data for the Nationally Endangered New Zealand grasshopper Brachaspis robustus has been determined by counting all adults and nymphs seen on a single ~1.5 km transect searched once annually. This method lacks spatial and temporal replication, both of which are essential to overcome detection errors in highly cryptic species like B. robustus. It also provides no information about changes in the grasshopper’s distribution throughout its range. Here, we design and test new population density and site occupancy monitoring protocols by comparing a) comprehensive plot and transect searches at one site and b) transect searches at two sites representing two different habitats (gravel road and natural riverbed) occupied by the species across its remaining range. Using power analyses, we determined a) the number of transects, b) the number of repeated visits and c) the grasshopper demographic to count to accurately detect long term change in relative population density. To inform a monitoring protocol design to track trends in grasshopper distribution, we estimated the probability of detecting an individual with respect to a) search area, b) weather and c) the grasshopper demographic counted at each of the two sites. Density estimates from plots and transects did not differ significantly. Population density monitoring was found to be most informative when large adult females present in early summer were used to index population size. To detect a significant change in relative density with power > 0.8 at the gravel road habitat, at least seventeen spatial replicates (transects) and four temporal replicates (visits) were required. Density estimates at the natural braided river site performed poorly and likely require a much higher survey effort. Detection of grasshopper presence was highest (p_g > 0.6) using a 100 m x 1 m transect at both sites in February under optimal (no cloud) conditions. At least three visits to a transect should be conducted per season for distribution monitoring. Monitoring protocols that inform the management of threatened species are crucial for better understanding and mitigation of the current global trends of insect decline. This study provides an exemplar of how appropriate monitoring protocols can be developed for threatened insect species.

Plot and transect searches for B. robustus were conducted at Patersons Terrace during the austral summer (November to March) for three consecutive seasons (2015–2018). Three 100 m x 1 m (100 m²) transects spaced ~1 km apart were marked out along the centre of the gravel road. Along each 100 m transect, four evenly spaced 5 m x 5 m (25 m²) plots were defined on alternating edges of the road using metal corner pegs. Henceforth, a group of 4 plots (equal to 100 m²) is referred to as a “plot unit”. The plots and transects remained in the same location over the duration of the study.

Transect searches were conducted at Snowy River during the austral summers of 2016–17 and 2017–18 only. Three 100 m x 1 m transects spaced > 200 m apart were set up longitudinally along the riverbed in locations where grasshoppers were known to occur. Transects were marked with plastic pegs placed at 20 m intervals. In the third season, two additional 100 m x 1 m transects were set up at the site. The three original transects were set up as close as possible to their initial location. However, changes in channel morphology resulted in minor deviations of less than 8 m.

Monitoring took place from November to March each summer on days of suitable weather (ground temperature > 13.6°C and not during gale force winds or precipitation). Each transect and plot unit was searched on at least 6 days per month when feasible (Patersons Terrace, mean visits per month = 6, range = 2 to 8; Snowy River, mean = 5.5, range = 3 to 7) but the number of monthly searches achieved and duration between visits (1 to 11 days) varied according to the occurrence of favourable weather. All plots and transects were searched by the same observer throughout the study. Both sites were searched on the same day except where weather was not permitting (e.g., rain at one of the sites). Prior to commencing a search, barometric pressure, air temperature at 1 m above the ground, and ground surface temperature in the shade were measured using a Kestrel 3500 Pocket Weather Metre (GeoSystems New Zealand Ltd). Cloud cover (categories: none; high cloud, when cloud was high in the sky but did not cause shadows; patchy cloud, when clouds were lower in the sky and caused shadows; overcast) and start time were recorded for each search. During the search, the observer walked slowly sweeping their front foot over the ground in front of them and moved in a direction such that their shadow fell on the area already searched. This method made visual detection possible by eliciting a jump response from the grasshopper. When detected, grasshoppers were captured, their body length (from the top of the head to the tip of the abdomen) and femur length measured, and sex and transect location recorded. Each grasshopper was then released behind the observer to ensure it was not re-counted, and the remainder of the area was searched. The minimum time required to complete a search of a 100 m² transect was 5 mins, and for each plot was ~1.5 mins (~6 mins per 100 m² plot unit including time taken to walk between plots) but this increased with the number of grasshoppers found. During this study, adulthood was determined by measuring hind femur length. Females with a femur length of ≥ 15 mm, and males with a femur length of ≥ 9 mm were considered adults. Individuals less than 8 mm in body length were excluded from the study because the risk of causing a fatal injury during capture and handling was too high.

For usage notes, please refer to README_Schori_et_al_2020_dataset.txt.