Can we benchmark annual ground cover maintenance?

Abstract

The capacity for rangeland stakeholders, including land managers, financiers and regulators, to regularly assess impacts of management practices on grazed landscapes has potential benefits. This paper describes the development of ground cover maintenance (GCM) spatial layers for a large study area in the catchment of the Great Barrier Reef in Queensland, Australia. GCM layers are an experimental product designed to benchmark the direction and strength of annual change in remotely sensed total ground cover (ΔTGC). This was achieved by predicting ΔTGC per pixel in a multivariate model, then using the quantile of the observed ΔTGC within its modelled prediction interval to benchmark observed ΔTGC. Under this approach, pixels with higher quantiles are those with a more positive annual observed ΔTGC after rainfall and other predictors in the multivariate model are taken into account. We then mapped these quantiles annually (2011–2021) across the study area and the annual spatial distribution of these quantiles is what we call the GCM layers. We identified two important issues to be addressed in future iterations of this work, namely, the potentially confounding impact of fire on GCM layers and their interpretation, and a need for more predictive skill in the underlying random forest model. Because management variables were not part of the underlying multivariate model but management practices can affect ΔTGC, we were interested in whether patterns in the mapped GCM values correlated with any known management practices or management-practice effects in the study area. We tested this idea on three datasets. In one, we compared GCM values from 12 well managed and 12 poorly managed grazing sites, finding no significant differences between the two groups. Another analysis looked at the relationship between grazing land condition and cumulative GCM values at two sets of sites (n = 110 and n = 189). Land condition and cumulative GCM values correlated significantly, although in only one of these data sets. Overall, we conclude that the developed GCM layers require further refinement to fit their desired purpose, but have potential to produce a number of benefits if current limitations can be addressed.