Resúmenes
| Estimating and mapping forest canopy fuel parameters from GEDI LiDAR data in Europe |
|---|
| Ponente |
| Elena Aragoneses de la Rubia |
| Resumen |
| Spatially-explicit information on canopy fuel parameters is key for wildfire propagation modelling, emission estimations and risk assessment. This work aims to develop easily-replicable methods to estimate critical fuel canopy parameters from spaceborne LiDAR observations acquired by the Global Ecosystem Dynamics Investigation (GEDI) sensor onboard the International Space Station. GEDI-like pseudo-waveforms were modelled from discrete Airborne Laser Scanning (ALS) data and used to select the best GEDI predictor metrics to derive European wall-to-wall forest height, canopy cover and canopy base height maps. Then, GEDI spaceborne footprints were used to generate continental maps of canopy parameters through a two-steps approach: 1) Spatial interpolation of GEDI footprints inside homogeneous forest fuel type polygons, and 2) Modelling machine learning algorithms for the forest fuel type polygons without GEDI footprints inside, using auxiliary multispectral and RADAR imagery and biophysical variables. Our results show the capabilities of remote sensing and GEDI to estimate and map the spatial patterns of critical forest canopy fuel parameters in fire risk prevention and contribute to generating the necessary tools to develop an integrated risk-wise strategy that reduces fire vulnerability of ecosystems across Europe. |
| Análisis de los usos de suelo urbano a nivel de parcela en diferentes municipios españoles |
|---|
| Ponente |
| Nikolai Shurupov |
| Resumen |
| Los bosques constituyen una de las fuentes más significativas de carbono en nuestro planeta, proporcionan refugio a innumerables especies y desempeñan un papel crucial en la lucha contra el cambio climático. La teledetección se ha convertido en una herramienta invaluable para monitorear y gestionar los bosques. Mediante su uso, podemos observar y analizar cambios en la cobertura forestal, así como evaluar la salud de los árboles. Esto no solo permite una toma de decisiones más fundamentada en cuanto a la conservación y el uso sostenible de estos recursos, sino que también contribuye en las políticas ambientales. El objetivo principal de este estudio es encontrar la combinación de datos más eficiente para la estimación de variables biofísicas a través de un algoritmo de machine learning utilizando como referencia los datos obtenidos de los vuelos LiDAR llevados a cabo a través del Plan de Ortofotografía Aérea (PNOA) para dos tipos de bosques: mediterráneo (Comunidad de Madrid) y atlántico (País Vasco). La altura y la cobertura del dosel se estimó para ambas regiones, a través de sensores ópticos (HLS, historial de perturbaciones: CCDC-SMA), radar (Sentinel-1 y TanDEM- X) e incluyendo predictores topográficos. Aunque más eficientes los modelos que incorporan TanDEM-X, no son viables a la hora de generar una serie temporal debido a la falta de datos. Para los bosques mediterráneos los modelos más eficientes fueron los que incorporaban datos ópticos mientras que para la zona atlántica destacó la incorporación de Sentinel-1 en combinación con las imágenes ópticas para la altura de la cubierta vegetal. |
| A European-scale analysis reveals the complex roles of anthropogenic and climatic factors in driving large fires initiation |
|---|
| Ponente |
| Clara Ochoa, Emilio Chuvieco, Avi Bar-Massda |
| Resumen |
| Analyses of wildfire initiation patterns and determination of their main drivers can inform more effective and efficient fire prevention strategies. Yet these analyses are typically carried out at local to national scales, thus preventing cross-border comparisons and broad-scale policy development. Here, we assessed how human and climatic factors drive the spatial variation in wildfire initiation that spread and created burned areas > 100 ha across Europe. Machine learning algorithms (Random Forest, RF) were trained to estimate sources of fire initiation in the European territory (“ET scale” hereafter) as a function of a wide set of explanatory variables, based on initiation locations derived from the FRY database of fires that occurred between 2001 and 2019. We generated six RF models: three that includes all fires with burned area > 100 ha and three that included only fires > 1000 ha (Climatic models, Human models and Mix models). This helped us to test different parameters and to learn which model is more suited for this scale and which variables help the model to reach better accuracy. The climatic and mix models had high predictive capacity (AUC between 79% and 81%), while the human models have an AUC of 60 %. Feature importance analysis with Shapley Additive Explanations (SHAP) helped us to assess the main initiation drivers across the ET. In terms of comparison, mixed and climate models show aridity and evapotranspiration as the most important variables. However, models with BA > 1000ha show a marked influence of human factors in their top five of importance. In the human models, the most important variables were grazing, forest- agriculture interfaces, distance to roads and population density age > 65. In addition to the global analysis of variable importance, SHAP performed a geographic analysis of importance to determine which variables helped the model to predict the initiations in each observation. In the climatic models we show, spatially, in which regions one variable is more important than another, aridity, evapotranspiration and warm climates the most important. Mixed models showed that human variables have a lower proportion of importance. However, we can highlight as influential the distance to roads and the forest-agricultural interface. In the human model, grazing is the variable with the highest spatial proportion of importance. These findings can help develop new forest fire management strategies at the European level, integrating them into a comprehensive strategy of fire risk assessment, reduction, and adaptation. This summary is part of an ongoing paper to be published in the coming months. |
| Global impacts of fire regimes on wildland bird diversity |
|---|
| Ponente |
| Fátima Arrogante Funes |
| Resumen |
| Fire is a natural disturbance that has a significant impact on ecosystems and plays a crucial role in the distribution and preservation of biota worldwide. The effects of fires on bird biodiversity can be both positive, as they can create new habitats, and negative, as they can reduce nesting success. To fully understand the ecological implications of fires, we need to understand the spatial distribution of birds and fire regimes, and how fire regimes affect wildland bird ecosystems. This paper examines the global effects of different fire regimes on the diversity of forest bird species. Initially, we used the MaxEnt algorithm to model the potential distribution of 1,115 forest bird species over a 20-year period. We also processed satellite observations of burned areas during the same period to estimate fire regime characteristics, including the average proportion of burnt vegetation, interannual variability in the burnt area, and fire intensity.
Our findings revealed the following :(I) The most affected wildland bird communities are those found in tropical forests, where the majority of fires occur; (II) high fire intensity values and a substantial proportion of burned vegetation have a positive impact on maintaining a diverse population of wildland birds in biomes characterized by savannah or grassland covers, as seen in temperate or tropical zones. Conversely, low fire intensity values and a smaller proportion of burned vegetation also promote greater diversity of wildland birds in forested areas in boreal or temperate zones and (III) in Mediterranean forests, a clear association between bird species diversity and wildfires could not be established. This research could help identify ecological areas that are vulnerable to wildfires. It could also be useful in the development of sustainable landscape management practices and the conservation of priority ecological zones in the tropical ecosystems. |
| Detección de patrones en la respuesta espectral según el nivel de decaimiento en masas de pino silvestre |
|---|
| Ponente |
| Marina Rodes Blanco |
| Resumen |
| El cambio climático ha producido un aumento en la frecuencia y en la severidad de las perturbaciones como sequias, plagas o incendios que, unido a otros factores como la propia sucesión natural y el cambio de uso de suelo por el abandono rural, está originando cambios en los bosques Mediterráneos.
En algunos casos, se están generando procesos de decaimiento que se traducen en áreas de bosques con altos índices de defoliación y de mortalidad. Sin embargo, todavía se desconoce la extensión y el alcance de estos procesos. Las afecciones más severas se esperan en el límite sur del rango de distribución de las especies donde, normalmente, las altas temperaturas y la disponibilidad hídrica son los factores limitantes de su crecimiento. En este trabajo se ha seleccionado la especie Pinus sylvestris, cuyo limite sur de su distribución se localiza en la Península Ibérica, para estudiar este fenómeno. Para ello, se ha realizado un extenso trabajo de campo durante el cual se han instalado parcelas del tamaño de un pixel de Landsat, en zonas con diferentes niveles de daño, y se han tomado medidas de inventario y de defoliación. El objetivo es entender cómo cambia la respuesta espectral entre zonas sanas y con decaimiento y modelizar la defoliación de acuerdo con las series extraídas para poder extrapolar los resultados fuera de las parcelas de estudio. |
| Modelling drivers of fire-related forest loss in sub-Saharan Africa |
|---|
| Ponente |
| Amin Khairoun |
| Resumen |
| Palabras clave: forest loss, small fires, seasonality, fragmentation, sub-Saharan Africa.
High omission rates of global coarse-resolution burned area (BA) products entail significant underestimation of actual fire impacts, especially in the tropics. Here, we analyse fire impacts on forest cover in sub-Saharan Africa (SSA) for 2016 and 2019 based on medium-resolution satellite data (20-30 m). We found that fires contribute to more than 46% of total forest losses, more than twice the estimates of coarse-resolution BA products (≥ 250 m) and that burned forest areas had more than twice the chance to be lost than unburned ones. In moist tropical forest, the most vulnerable biome to fires, burning had even six times more chance to precede forest loss. We also found that medium-resolution BA detects more fires in late fire season, which tend to have higher impact on forests than early-season fires. Small fires associated with shifting agriculture represented the major driver of forest loss after fires subsequently medium-resolution BA products are needed to better assess fire and forest dynamics in tropical ecosystems. |
| Evaluación de los procesos superficiales de la Península de Byers (Islas Shetland del Sur) empleando técnicas de interferometría diferencial de radar de apertura sintética |
|---|
| Ponente |
| Claudia Giménez Poblador |
| Resumen |
| Introducción:
La progresiva retirada de los hielos, debido al cambio climático del planeta, afecta en especial a las zonas libres de hielo de la Antártida. Estos cambios son especialmente preocupantes en la región norte de la península antártica, donde está previsto que se produzca el mayor aumento de extensión de áreas libres de hielo en las próximas décadas (Lee et al., 2017) debido al ascenso de las temperaturas y a una mayor frecuencia de precipitaciones en forma de lluvia en lugar de nieve (Turner et al., 2014). Por este motivo, esta amenaza medioambiental surge como una oportunidad para el estudio de los diferentes procesos que afectan al permafrost de estas zonas prístinas. Objetivos: El objetivo de este trabajo es determinar los desplazamientos del terreno en la península de Byers, en la isla de Livingston, en las Islas Shetland del Sur utilizando imágenes de satélite SAR (Synthethic Aperture Radar) para hacer interferogramas con técnicas de DInSAR (Diferential Interferomtry Synthethic Aperture Radar), y medir los procesos de subsidencia o levantamientos en la península de Byers. Metodología: La metodología aplicada en este estudio se basa principalmente en la integración de técnicas de DInSAR para cumplir con el objetivo propuesto.
Por tanto, para cada verano o periodo de estudio se realizaron 11 y 12 interferogramas, respectivamente. Por último, el software utilizado ha sido SNAP (Sentinels Application Platform) versión 9.0, y SNAPhu plugin (version 2.0) de códigos abiertos perteneciente a la ESA. Resultados: Los valores medios de desplazamiento con respecto a la línea de visión o LOS (Line of Sight) son mínimos, en concreto de 1 a 2 cm, durante los periodos de tiempo indicados y muy probablemente están influidos por las variaciones estacionales cuando la cubierta de nieve está presente, así como por la fluctuación del contenido de humedad del suelo. Sin embargo, cuando la capa de nieve es mínima, se observan desplazamientos con respecto a la LOS de hasta 7 cm, alrededor del frente del glaciar y en determinadas zonas de las plataformas y laderas superiores, especialmente donde el permafrost y los rasgos periglaciares están más presentes. Conclusiones: Estos resultados proporcionan una primera aproximación de la subsidencia y el levantamiento que se producen en la zona libre de hielo y sirven para vigilar estas zonas cuando en los próximos años se disponga de nuevos datos adquiridos. Bibliografía: Diferencial SAR. Trabajo Fin de Máster. Universidad de Salamanca. Disponible en: https://gredos.usal.es/bitstream/handle/10366/138637/TFM_GonzalezCalvoL_Seguimientoycontrol.pdf?sequence=1yisAllowed=y. (Consultado 15/5/2022). Instrument Payload. The European Space Agency. Disponible en: https://sentinels.copernicus.eu/web/sentinel/missions/sentinel-1/instrument-payload. Consultado 20/10/2023. Interferometric Wide Swath. The European Space Agency. Disponible en: https://sentinels.copernicus.eu/web/sentinel/user-guides/sentinel-1-sar/acquisition-modes/interferometric-wide-swath. Consultado 20/10/2023. Sentinel-1. The European Space Agency. Disponible en: https://sentinels.copernicus.eu/web/sentinel/missions/sentinel-1. Consultado 20/10/2023. SNAP. The European Space Agency. Disponible en: https://earth.esa.int/eogateway/tools/snap. (Consultado 22/10/2023). |
| Monitoring forest loss and degradation in Spain (1988-2023) using Landsat time series |
|---|
| Ponente |
| Sofía Miguel Romero and Mihai Tanase |
| Resumen |
| Palabras clave: forest disturbance, Landsat, temporal analysis.
Forests are a key element for carbon sequestration and a major component of rural development providing habitat, protective functions, as well as contributing with goods and services. Biotic and abiotic disturbances are natural processes driving forest dynamics. However, many forest ecosystems have experienced increased disturbance rates with recent trends reaching an unprecedented level. In Europe alone, 17% of forests were affected by some type of disturbance over the past 20 years. Similar tendencies have been observed over the peninsular Spain where large forest areas are affected by fires every year while smaller areas are affected by insect outbreaks and pathogens. Such disturbances may be identified and monitored by means of remote sensing including from historic archives such as those used to derive the Spanish Atlas of Forest Disturbance (SAFoD). SAFoD uses the entire Landsat archive to generate a comprehensive database of disturbance events (1985-2022) as well as the major agents driving them. SAFoD combines the latest advances in remote sensing for forest monitoring including the Continuous Change Detection and Classification algorithm in conjunction with Spectral Mixture Analysis and the Continuous Degradation Detection algorithms to provide, among other, information on disturbance year and month, its magnitude as well as its probability (Chen et al. 2021). Using time series of Landsat images, the algorithm measures the proportion of different types of cover within a single pixel and detects those significant spectral responses on the earth’s surface, which makes it possible to locate changes in forest cover. This contribution presents the methodological developments behind SAFoD, the cartographic results over the entire Iberian Peninsula, as well as the preliminary validation of this database which reaches accuracies between 80 and 95% depending on the reference data used. |
| Fuel type classification using multi-seasonal Sentinel data, topographic and biophysical models |
|---|
| Ponente |
| Pegah Mohammadpour, Domingos Xavier Viegas, Emilio Chuvieco, Alcides Pereira |
| Resumen |
| Palabras clave: Biophysical models, fuel type mapping, feature selection, Random Forest, Sentinel data, spectral-temporal feature.
Wildfires significantly reshape the Mediterranean basin’s landscape, leading to alterations in forest composition, structure, and diversity. Hence, fuel mapping is crucial for improving fire risk assessment and enhancing fire behavior modeling. This study aims to generate a fuel type map by considering forest structure and phenology, shrubland, and grassland height in central Portugal, FirEUrisk pilot site 3(PS3).The vegetation fuel type map has been developed based on the FirEUrisk hierarchical fuel classification system (EHFCS) using spectral and radar data (Sentinel-1, Sentinel-2, and vegetation indices(VIs)), Gray-level co-occurrence matrix (GLCM) texture features, topographic variables, and biophysical models. Multi-seasonal images were used to provide spectral–temporal features on phenological changes in vegetation cover, which is expected to facilitate the fuel type classification. Feature selection was performed based on Recursive Feature Extraction (RFE) and Random Forest built-in feature importance measures using Mean Decrease Gini (MDG) and Mean Decrease Accuracy (MDA). Random Forest classifier was applied to different feature sets, and classification results were assessed by cross-validation of overall accuracy (OA), out-of-bag error (OOB), and F1-score. The proposed methodology achieved an OA of 83.62% by assessing with an independent validation set for the fusion of VIs, GLCM textures, VH (vertical transmit–horizontal receive) of Sentinel-1, and elevation with integration of biophysical models to estimate shrubland and grassland height. Our results suggest that incorporating different spectral-temporal features with texture and radar data can improve both prediction accuracy and interclass separability of fuel type classification. Moreover, biophysical models based on vegetation normalized vegetation index (NDVI) and precipitation effectively estimated grassland and shrubland height. |
| Deriving annual forest biophysical variables by combining Airborne Laser scanning with active and passive satellite sensors |
|---|
| Ponente |
| M. Cristina Mihai, Sofía Miguel and Mihai A. Tanase |
| Resumen |
| Comprehending forest dynamics is one of the main goals for forest management. Remote sensing data has been a crucial tool in the observation of forests in recent years. Given the high temporal and spatial resolution, as well as free access policies, a range of possibilities has been opened to understand forest growth patterns and monitor its dynamics. In this study, we analyze different combinations of active and passive satellite sensors to estimate forest biophysical variables (i.e., canopy height -CH, forest canopy cover -FCC, and above ground biomass -AGB) derived from Light Detection And Ranging (LiDAR). A series of Random Forest (RF) regression models were calibrated to analyze the sensor combination that provides the most efficient retrieval over two study areas located in Spain (Community of Madrid -CAM, characterized by a Mediterranean-Continental climate and the Basque Country -PV, characterized by its Oceanic climate). Passive (i.e., Landsat 7/8 and Sentinel-2 A/B), and active sensors (i.e., Sentinel-1 and PALSAR-2) were used to surpass limitations of consistent cloud cover over some regions. Results show that, overall, most combinations perform well, with an R2 ranging from 0.77 to 0.81 for CH; from 0.73 to 0.80 for FCC; and 0.75 to 0.77 for biomass in semi-arid forests (CAM). Models that inlcude Sentinel-2 data in combination with another sensor (Sentienl-2 and Landsat), show the lowest errors with a RMSE of 1.69 (m), 14.12 (%) and 28.59 (t/ha) (CH, FCC and AGB). The models were used to derive yearly mapping products that complement the quinquennial LiDAR derived biophysical variables. Nevertheless, over higher biomass atlantic forests (PV) no model provided satisafactory results. Best results were obtained for the model that combines Sentine-2 and Sentinel-1 with a R2 of 0.60 (CH), 0.61 (FCC) and 0.53 (AGB). |
| Análisis de patrones y modelización de las dinámicas del paisaje en el centro sur de Chile: El impacto de las plantaciones forestales |
|---|
| Ponente |
| Cristian Vergara |
| Resumen |
| Palabras claves: Plantaciones forestales, uso/cobertura del suelo, dinámicas del paisaje.
La expansión de las plantaciones forestales es uno de los principales cambios de uso cobertura del suelo que se han producido transversalmente en paises de bajo y alto ingreso como por ejemplo Chile, Australia, China, Portugal. Investigaciones previas indican que la expansión de las plantaciones forestales produce efectos tanto positivos como negativos. Entre los efectos positivos comúnmente se menciona la contribución al crecimiento económico, y la protección frente al deslizamiento de tierra, y la erosión del suelo. Por otro lado, en muchos casos la expansión de las plantaciones forestales ha producido una pérdida de diversidad, la sustitución de bosque nativo, aumento en el riesgo de incendios forestales, y la disminución en la provisión de agua. Dada esta controversia es necesario contar con regulación y procesos de planificación que puedan indicar la cantidad y localización adecuada para el desarrollo de esta actividad y su contribución a los objetivos de desarrollo regional y nacional, basado en promover los principios de la sostenibilidad del paisaje. En esta línea, este trabajo tiene como objetivo estudiar la configuración espacial que presentan las plantaciones forestales, los factores de localización que favorecen o disminuyen su expansión, y la simulación de la expansión en el futuro. A este respecto, Chile es un excelente caso de estudio, siendo uno de los paises que ha incrementado en mayor cantidad esta cobertura en los últimos 40 años, manifestando tanto los efectos positivos como negativos de este fenómeno. Estos resultados serán de gran importancia para contribuir a los procesos de formulación de planes de ordenamiento territorial dentro del país, elaborados a partir de recientes cambios en la legislación, y que entrarán en vigor en los próximos años. |
| Configuración y zonificación geológica de un modelo SWAT+ para simular de forma realista el sector de cabecera del río Tajo |
|---|
| Ponente |
| José Manuel Rodríguez Castellano |
| Resumen |
| Palabras clave: Modelización hidrológica, río Tajo, subcuencas, SWAT+, zonificación geológica
El sector de cabecera del río Tajo que vierte a los embalses de Entrepeñas y Buendía es sumamente relevante desde el punto de vista de la gestión del agua en España: además de la propia presencia de estos hiperembalses, de allí parte el trasvase Tajo-Segura, que envía un no desdeñable volumen de agua al sureste español. La modelización de este sector se configura por tanto como una herramienta útil para, entre otras cuestiones, poder evaluar futuros escenarios de disponibilidad de agua. Con el objetivo de simular las aportaciones al sistema de hiperembalses, se ha construido un modelo hidrológico de su cuenca vertiente con SWAT+ a través de su interfaz en QGIS (QSWAT+), que facilita tanto la delineación de la cuenca, la red hidrográfica y las unidades del paisaje, como la incorporación de la información relativa a usos del suelo, tipos de suelo y pendiente. Además, también mediante procesamiento SIG, se ha realizado una zonificación geológica de la cuenca de estudio, definiéndose tres regiones principales: carbonatada, detrítica de alta permeabilidad y detrítica de baja permeabilidad. Teniendo en cuenta esta zonificación, se ha diseñado una metodología de calibración regionalizada, teniendo como fin último la reproducción de forma realista de los procesos hidrológicos de las diferentes subcuencas. Como resultado final, se ha obtenido un modelo hidrológico estadísticamente satisfactorio, que reproduce con fidelidad los caudales observados y que simula de manera realista los componentes del balance hidrológico y de la escorrentía, habiendo jugado los SIG un papel fundamental en el proceso. |
| Forest canopy fuel load mapping using UAV high resolution RGB and multispectral imagery |
|---|
| Ponente |
| Álvaro Chávez Durán |
| Resumen |
| Palabras clave: Fuel loads; photogrammetry; segmentation; multispectral; Random Forest.
The development of appropriate strategies for fire management and fire risk reduction requires an accurate description of forest fuels. Canopy fuel loads determine the characteristics of the entire complex of forest fuels, due to their constant changes triggered by the environment. This paper presents a method for mapping spatial distribution of canopy fuel loads, consistent with its natural variability and three-dimensional spatial distribution, using geospatial data, UAV-based digital photogrammetry and multispectral data, segmentation analysis and the Random Forest algorithm. The proposed method was developed in the mixed forest of the natural protected area of “Sierra de Quila,” Jalisco, Mexico. Input data encompassed field data and Unmanned Aerial Vehicles (UAV) mounted RGB and multispectral imagery. UAV-RGB imagery allowed the generation of three-dimensional models from groups of trees that, were delimited through segmentation analysis and was estimated their canopy volume. UAV- multispectral imagery allowed textures and vegetation indices estimation. Canopy fuel load spatial distribution was processed using Machine Learning techniques, through Random Forest algorithm. Random Forest model reached R2=0.75, RMSE=1.78 Mg and average Biasrel = 18.62%. Canopy volume was identified as the most significant explanatory variable, achieving a mean decrease in impurity values greater than 80%, while texture and vegetation indices, jointly, presented importance values close to 20%. Our modeling approach allows to achieve accurate estimates of forest canopy fuel loads, according to the ecological context that governs its dynamics and spatial variability. The high precision achieved at the lowest cost, encourages constant updating of forest fuels maps to enable researchers, and forest managers to streamline decision-making on fuel and forest fire management. |