Análisis de la severidad y la capacidad de recuperación de un matorral atlántico tras el paso del fuego en estacionalidad verano-otoño

Laburpena

Either of natural or anthropogenic origin, fire plays a prominent role in both the structure and dynamics of ecosystems. The Atlantic-Cantabrian region is characterised by its high frequency and recurrence of forest fires. The evolution of fire behaviour in the last decades has been a concerning matter for firefighting services across the globe. This dynamic finds its origins in the increment of fuel loads and fire continuity, as well as prolonged drought periods. All of which give rise to a higher fire intensity, secondary fire emissions over long distances, and the proliferation of economic and environmental impacts. In the Atlantic peninsular region, specifically in the urban-rural juncture, there is a growing frequency of episodes which present unprecedented numbers of burnt areas. In order to provide solutions which improve the trajectory of the last decades, it is imperative to investigate and acquire the necessary knowledge to implement changes. In Asturias, 10,000 ha of forest are burnt every year, 72 % of which are burnt during the winter-spring months, and 80 % of which constitutes shrubland. The initial proposal of this thesis investigated the comparison between fire behaviour and fire severity, as well as the regeneration capacity of an Atlantic shrubland measured in the course of both winter-spring and summer-autumn seasonalities. During the summerautumn interval of 2017, the first controlled burn was carried out. However, a week after, a forest fire burnt the area adjacent to where the first controlled burn had been conducted; this was the location for the second controlled burn, which was scheduled for the winterspring period. This unforeseen event, along with the weather conditions of the following months, prevented the execution of controlled burns during the winter-spring seasonality, all of which replaced the specific aims of this thesis to the following ones: i) identifying the severity caused by the passage of fire during the summer-autumn seasonality according to fire behavioural patterns, both in a controlled burn and a forest fire (Chapter I), ii) evaluating the feasibility of images acquired by a multispectral sensor boarded on an unmanned aerial vehicle (UAV), aiming to estimate the severity of vegetation and soil burning by employing an artificial neural network (ANN) classifier, iii) determining the regeneration capacity of an Atlantic shrubland after a forest fire and a controlled burn respectively, identifying whether there are any differences between each type of fire (Chapter III). Chapter I covers the comparison between a high-intensity controlled burn and a forest fire in the Atlantic shrub. A controlled burn of 6.7 ha was conducted on the 8th of October 2017, whereas the adjacent plot, comprehended of 6.8 ha and also scheduled for a controlled burn during a different seasonality, was burnt by a wildfire on the 15th of October 2017, affecting a total 3700 ha of land. This study assesses fire behaviour and its severity, in conjunction with the reduction of low fuel loads, for both controlled burns and wildfires. The comparative analysis of fire behaviour required weather stations, K-type thermocouples, unmanned aerial vehicles, photographic and audiovisual material as well as pre and post-fire field inventories. Additionally, fire severity was determined by using the Composite Burn Index (CBI), the difference in pre and post fire apical diameter and remote sensing, employing for the latter the Normalized Burn Ratio (dNBR) from Sentinel-2 imagery and the Relativized Burn Ratio (RBR) from Sentinel-1 synthetic aperture radar. Fire intensity and its impact on the ground were found to be lower during the controlled burn than in the wildfire, in which fire intensity was recorded to be 13 times higher. Likewise, in the measurement of smoldering values, wildfire recorded a high value of 95 %, as opposed to controlled burn which reported a lower level of 17 %. The different fire ignition patterns (heading-fire, backing-fire and flanking-fire behaviours) resulted in differences regarding fire intensity and severity. During the controlled burn, flanking-fire behaviour had a lower impact on the soil than heading-fire. However, it achieved a similar reduction of wood fuels. Hence, controlled burns constitute both a tool for firefighting training and an implement for forest fuel control, including the development of more resistant and resilient landscapes in the face of large wildfires. Chapter II evaluates fire severity both in the forest fire and the controlled burn, assessing them by using an unmanned aerial vehicle (UAV) with a Parrot SEQUOIA multispectral camera. Post-fire surface reflectance images were used: green (550 nm), red (660 nm), red-edge (735 nm) and near-infrared (790 nm) bands were used at a high spatial resolution (GSD 20 cm). Furthermore, as a means to estimate fire severity in both soil and vegetation, 153 field plots were introduced to the study, where severity patterns were explored through probabilistic neural network (PNN) algorithms, based on field inventory data and UAV imagery derived products. PNNs accurately classified 84.3 % of control marks or field severities in vegetation, and 77.8 % in soil. Chapter III identifies the ecosystem’s regeneration capacity after the passage of fire, comparing a controlled burn and a forest fire. For this purpose, pre and post fire inventories were carried out both 9 and 19 months after the fire, collecting data in each of the controlled burn sample land plots. The collected data registered the existing species, the percentage land covered by wood and herbaceous species, percentage of bare and stonecovered soil, the richness of herbaceous species and the overall vegetation height. In the area affected by the wildfire, which was homogenous in vegetation, slope and orientation characteristics, it was impossible to collect pre-fire data, although a post-fire inventory was carried out for a comparative analysis with the controlled burn data. During the controlled burn, vegetation speed recovery proved to be faster than in a wildfire. Additionally, a decrease in bare soil was also found to be more pronounced after a controlled burn. However, the richness of vegetable species after a fire in both modalities (wildfire and controlled burn) was found to be similar to the controlled situation, proving there is no record of biodiversity loss in the summer-autumn seasonality in this type of schrubland and its recurrence. Thus, this scientific knowledge constitutes a helpful tool in the decisionmaking process regarding a comprehensive management of fire in Atlantic shrublands, both from a preventive point of view and concerning its extinction and restoration.