What is wrong with post-fire soil erosion modelling?

Wildfire patterns are shifting all over the world as a consequence, among others, of changes in land use and climate [1], which may entail remarkable social, environmental, and economic implications. The occurrence of wildfires is often linked to increased post-fire hydrological and erosive responses, which are hard to predict due to the complexity of factors involved [2]. Against this background, soil erosion models arise as a resourceful tool in the decision-making process for environments that are or could be affected by wildfires: from prevention to mitigation and from emergency actions to long-term planning. Nevertheless, the current soil erosion models were not originally developed for post-fire conditions, so they are not adapted to include fire-related changes into their predictions [3]. This work aimed to review the scientific advances in the last twenty years in post-fire soil erosion modelling research from a meta-analysis approach.

To this end, the Scopus database was searched using different combinations of the terms “model”, “modelling”, “fire”, “wildfire” “hydrology”, “erosion”, “runoff”, “burn”, “burnt”, “erosion”, “soil erosion”, “sediment” and “rill”. Afterwards, the following publications were excluded: a) reviews; b) journals without peer-review process; c) books or book chapters; d) reports; e) editorials; f) conference proceedings; g) works in which the modelling was conducted on individual processes; h) studies modelling debris flows and landslides; i) works that did not conduct post-fire and/or erosion modelling; j) works that are not in English. Then, it was identified whether authors included in the models important factors related to soil erosion in fire-affected environments such as changes in water infiltration, burn severity, and/or the application of post-fire mitigation treatments. The main modelling approaches used, the calibration and validation of predicted data, and the use of efficiency indexes were also evaluated. 

The screening resulted in 33 works (43 cases based on the model used) that were not homogeneously distributed worldwide, neither according to the model type used, nor by regions most affected by wildfires. For the calibration process, in 70% of the cases models were adapted to burned conditions but only in 25% of them, individual input parameters were improved to accommodate processes that were not previously represented. Additionally, burn severity and changes in infiltration were considered in 77 and 65% of the cases, respectively, whereas only 26% of the cases corresponded to studies where post-fire mitigation treatments were applied. It is noteworthy that only in 19% of the cases, the predicted data were validated with independent field datasets and uncertainty was assessed in 5% of the studies.

It is highlighted that further efforts are required on the adaptation of erosion models to burned conditions, evaluating the model performance in both calibration and validation stages for a wider variety of environments and scenarios, in order to accurately predict the hydrological and erosive response after fires.

[1] Andela et al. (2017). Science 356: 1356-1362. DOI: 10.1126/science.aal4108

[2] Larsen & MacDonald (2007). Water Resour. Res. 43: W11412. DOI: 10.1029/2006WR005560

[3] Vieira et al. (2018). Environ. Res. 165: 365-378. DOI: 10.1016/j.envres.2018.04.029