Digital soil carbon mapping in diseased Agathis australis forests using site-specific and remote sensing data

Forest soils contain over 40% of the total organic carbon stored in terrestrial ecosystems. Information on the distribution of soil carbon and its influencing factors is important for assessing forest ecosystem functions and potential impacts of disturbances. Forest disease and dieback events are increasing, but investigations into the geospatial distribution and the factors influencing soil carbon in diseased forests are lacking. Agathis australis (kauri), a culturally and ecologically significant species of northern New Zealand, is threatened due to kauri dieback caused by Phytophthora agathidicida. Using soil samples and tree survey data collected across diseased kauri forests in the Waitākere Ranges Regional Park, Auckland, New Zealand, we quantified and estimated a set of 26 variables, including site-specific (location, tree and soil related characteristics), topographic, and remote sensing-derived variables. Random forest regression models were employed to predict soil carbon concentration using three different sets of variables. The model using all variables was the most successful (RMSE=1.64, R2=0.99, adjusted R2=0.98), followed by the model using all site-specific soil variables (RMSE=1.52, R2=0.99, adjusted R2=0.99). The model using only topographic and remote sensing-derived sources performed the worst (RMSE=11.37, R2=0.25, adjusted R2=0.17). Predictions ranged from low soil carbon concentration (< 10%) along the west coast to high soil carbon concentration (up to 49% C) in the centre of the Waitākere Ranges Regional Park. Visualising the results of these models across the regional park illustrated that soil carbon concentration increased with increasing tree size and forest floor depth. Our findings will help to inform forest management efforts in the Waitākere Ranges Regional Park.