Modelling burned area and emissions with deep learning

<p>Natural hazards often result from interacting physical processes across a wide range of spatial and temporal scales. Namely, the occurrence of heatwaves and droughts increases the risk of wildfires, with recurrent extensive human, ecological and economic losses being reported throughout the world. In Australia, several extreme bushfires have occurred following severe droughts and heatwaves, namely the Black Saturday bushfires in 2009 and the extreme bushfire season of 2019-2020.</p><p>In this work, we analyze the relation between fire occurrence, drought conditions, and temperature extremes in southeastern Australia for the period 1982-2018 and considering the months between December and February. Monthly burned area (BA) was retrieved from the FireCCILT11 dataset, with a spatial resolution of 0.25˚. The drought index SPEI was used to assess the drought conditions and was computed using monthly precipitation and temperature data from the CRU TS 4.05 database. The occurrence of temperature extremes was assessed using the index Number of Hot Days (NHD), which was computed using daily maximum temperature obtained from the ERA5 dataset.</p><p>The study area comprises pixels that have burned at least 25 times on these months. A correlation analysis was performed between BA and SPEI at time scales of 1, 3, and 6 months, and between BA and NHD. The influence of current and previous conditions on BA was assessed, by correlating BA with SPEI and NHD at the current month, and in the previous 1 to 3 months. The joint probability of BA, drought, and temperature extremes was also assessed, using copula functions.</p><p>The results show a negative correlation between BA and SPEI, and a positive correlation between BA and NHD. For previous months, the correlation is stronger between BA and SPEI, than for BA and NHD, pointing to an effect of the drought conditions on previous months, whereas the effect of temperature on BA is seen instantaneously. The probabilistic analysis shows a clear increase in the probability of BA exceeding the 80th percentile, given drought conditions, compared to non-drought conditions. A similar result was obtained for the case of extreme temperature.</p><p>Acknowledgements: This study was partially supported by national funds through FCT (Fundação para a Ciência e a Tecnologia, Portugal) under projects FIRECAST (PCIF/GRF/0204/2017) and Floresta Limpa (PCIF/MOG/0161/2019); and the 2021 FirEUrisk project, funded by the European Union’s Horizon 2020 research and innovation programme under the Grant Agreement no. 101003890.</p>

Connectivity among populations influences resilience, genetic diversity , adaptation and speciation, so understanding this process is fundamental for conservation and management. This chapter summarises the main mechanisms of gene flow within and among seagrass meadows, and what we know about the spatial patterns of gene flow around Australia’s coastline. Today a significant body of research on the demographic and genetic connectivity of Australian seagrass meadows has developed. Most studies have focused on the genera Posidonia, Zostera, Heterozostera and Thalassia, in tropical and temperate systems across a range of habitats. These studies have shown overwhelmingly, that sexual reproduction is important for meadow persistence, as in most cases Australian seagrass meadows are genotypically diverse, with moderate to high levels of genotypic diversity. This high diversity could be generated through demographic connectivity, recruitment of individuals sourced from within a meadow, or from dispersal between meadows. Attempts to understand the relative significance of these processes are limited, highlighting a major gap in our understanding. Genetic structure is apparent across a range of spatial scales, from m’s to 100’s to 1000’s km. At local and regional scales, particularly in confined systems such as estuaries and bays, it is not necessarily the dominant oceanographic currents influencing patterns of genetic connectivity, but local eddies, winds and tides. Over larger spatial scales, isolation by distance is consistently significant, with unique genetic clusters spreading over 100s of kilometres. This indicates that regional structure occurs at the limits of long distance dispersal for the species and this is particularly evident where meadows are highly fragmented. The number of genetic studies on Australian seagrasses has increased dramatically recently; however, there are still many opportunities to improve our understanding through focusing on species with different dispersal potentials, more detailed sampling across a range of spatial and temporal scales and combining ecological and modelling approaches.

Intertidal landscapes are highly complex and dynamic habitats that exhibit variability over a range of spatial and temporal scales. The spatial arrangement of structure-forming biogenic features such as seagrasses and bivalves influences ecosystem function and the provision of important ecosystem services, though quantification and monitoring of intertidal landscape structure has been hindered by challenges collecting spatial data in the coastal zone. In this study, an intertidal landscape mosaic of eelgrass ( Zostera marina) and blue mussels ( Mytilus edulis) was observed using low-altitude aerial photography from a balloon-mounted digital camera platform. Imagery representing seagrass-bivalve landscape structure was classified and analysed using multiple metrics of landscape composition and configuration at the patch scale and the landscape scale. Patch-scale imagery was compared to a previously collected dataset in order to track temporal changes in seagrass patch metrics over a 26-month period. Seagrass and bivalve patches exhibited distinct spatial patterning at different spatial scales. At the patch scale, the change in seagrass metrics was consistent with patch border expansion at the expense of patch density and integrity. These methods demonstrate a novel approach for collecting high-resolution spatial data that could also be valuable to physical geographers dealing with similar fine-scale landscapes. The application of spatial metrics at multiple spatial scales quantified elements of the configuration and composition of a seagrass-bivalve habitat mosaic and allowed for the tracking of patch metrics through time to depict landscape change. Continued development of landscape metrics within intertidal habitats will increase understanding of the ecological function of these areas with benefits to management and monitoring of ecosystem health.

Quantifying the degradation of organic matter in marine sediments: A review and synthesis

In the context of ongoing climate change, extreme weather events are drawing increasing attention from the public and news media. A question often asked is how the likelihood of extremes might have changed by anthropogenic greenhouse-gas emissions. Answers to the question are strongly influenced by the model used, duration, spatial extent, and geographic location of the event—some of these factors often overlooked. Using output from four global climate models, we provide attribution statements characterised by a change in probability of occurrence due to anthropogenic greenhouse-gas emissions, for rainfall and temperature extremes occurring at seven discretised spatial scales and three temporal scales. An understanding of the sensitivity of attribution statements to a range of spatial and temporal scales of extremes allows for the scaling of attribution statements, rendering them relevant to other extremes having similar but non-identical characteristics. This is a procedure simple enough to approximate timely estimates of the anthropogenic contribution to the event probability. Furthermore, since real extremes do not have well-defined physical borders, scaling can help quantify uncertainty around attribution results due to uncertainty around the event definition. Results suggest that the sensitivity of attribution statements to spatial scale is similar across models and that the sensitivity of attribution statements to the model used is often greater than the sensitivity to a doubling or halving of the spatial scale of the event. The use of a range of spatial scales allows us to identify a nonlinear relationship between the spatial scale of the event studied and the attribution statement.

Advances in weather and climate research have demonstrated the role of the stratosphere in the Earth system across a wide range of temporal and spatial scales. Stratospheric ozone loss has been identified as a key driver of Southern Hemisphere tropospheric circulation trends, affecting ocean currents and carbon uptake, sea ice, and possibly even the Antarctic ice sheets. Stratospheric variability has also been shown to affect short-term and seasonal forecasts, connecting the tropics and midlatitudes and guiding storm-track dynamics. The two-way interactions between the stratosphere and the Earth system have motivated the World Climate Research Programme's (WCRP) Stratospheric Processes and their Role in Climate's (SPARC) activity on Modelling the Dynamics and Variability of the Stratosphere-Troposphere System (DynVar) to investigate the impact of stratospheric dynamics and variability on climate. This assessment will be made possible by two new multimodel datasets. First, roughly 10 models with a well-resolved stratosphere are participating in the Coupled Model Intercomparison Project phase 5 (CMIP5), providing the first multimodel ensemble of climate simulations coupled from the stratopause to the sea floor. Second, the Stratosphere Resolving Historical Forecast Project (Strat-HFP) of WCRP's Climate Variability and Predictability (CLIVAR) program is forming a multimodel set of seasonal hind-casts with stratosphere-resolving models, revealing the impact of both stratospheric initial conditions and dynamics on intraseasonal prediction. The CMIP5 and Strat-HFP model datasets will offer an unprecedented opportunity to understand the role of the stratosphere in the natural and forced variability of the Earth system and to determine whether incorporating knowledge of the middle atmosphere improves seasonal forecasts and climate projections.

Abstract. Understanding how fire regimes change over time is of major importance for understanding their future impact on the Earth system, including society. Large differences in simulated burned area between fire models show that there is substantial uncertainty associated with modelling global change impacts on fire regimes. We draw here on sensitivity simulations made by seven global dynamic vegetation models participating in the Fire Model Intercomparison Project (FireMIP) to understand how differences in models translate into differences in fire regime projections. The sensitivity experiments isolate the impact of the individual drivers on simulated burned area, which are prescribed in the simulations. Specifically these drivers are atmospheric CO2 concentration, population density, land-use change, lightning and climate. The seven models capture spatial patterns in burned area. However, they show considerable differences in the burned area trends since 1921. We analyse the trajectories of differences between the sensitivity and reference simulation to improve our understanding of what drives the global trends in burned area. Where it is possible, we link the inter-model differences to model assumptions. Overall, these analyses reveal that the largest uncertainties in simulating global historical burned area are related to the representation of anthropogenic ignitions and suppression and effects of land use on vegetation and fire. In line with previous studies this highlights the need to improve our understanding and model representation of the relationship between human activities and fire to improve our abilities to model fire within Earth system model applications. Only two models show a strong response to atmospheric CO2 concentration. The effects of changes in atmospheric CO2 concentration on fire are complex and quantitative information of how fuel loads and how flammability changes due to this factor is missing. The response to lightning on global scale is low. The response of burned area to climate is spatially heterogeneous and has a strong inter-annual variation. Climate is therefore likely more important than the other factors for short-term variations and extremes in burned area. This study provides a basis to understand the uncertainties in global fire modelling. Both improvements in process understanding and observational constraints reduce uncertainties in modelling burned area trends.

<strong class="journal-contentHeaderColor">Abstract.</strong> Land biosphere processes are of central importance to the climate system. Specifically, ecosystems interact with the atmosphere through a variety of feedback loops that modulate energy, water, and <span class="inline-formula">CO₂</span> fluxes between the land surface and the atmosphere across a wide range of temporal and spatial scales. Human land use and land cover modification add a further level of complexity to land–atmosphere interactions. Dynamic global vegetation models (DGVMs) attempt to capture land ecosystem processes and are increasingly incorporated into Earth system models (ESMs), which makes it possible to study the coupled dynamics of the land biosphere and the climate. In this work we describe a number of modifications to the LPJ-GUESS DGVM, aimed at enabling direct integration into an ESM. These include energy balance closure, the introduction of a sub-daily time step, a new radiative transfer scheme, and improved soil physics. The implemented modifications allow the model (LPJ-GUESS/LSM) to simulate the diurnal exchange of energy, water, and <span class="inline-formula">CO₂</span> between the land ecosystem and the atmosphere and thus provide surface boundary conditions to an atmospheric model over land. A site-based evaluation against FLUXNET2015 data shows reasonable agreement between observed and modelled sensible and latent heat fluxes. Differences in predicted ecosystem function between standard LPJ-GUESS and LPJ-GUESS/LSM vary across land cover types. We find that the emerging ecosystem composition and carbon fluxes are sensitive to both the choice of stomatal conductance model and the response of plant water uptake to soil moisture. The new implementation described in this work lays the foundation for using the well-established LPJ-GUESS DGVM as an alternative land surface model (LSM) in coupled land–biosphere–climate studies, where an accurate representation of ecosystem processes is essential.

Land biosphere processes are of central importance to the climate system. Specifically, biological processes interact with the atmosphere through a variety of feedback loops that modulate energy, water and CO2 fluxes between the land surface and the atmosphere across a wide range of temporal and spatial scales. Human land use and land cover modification add a further level of complexity to land-atmosphere interactions. Dynamic Global Vegetation Models (DGVMs) attempt to capture these land surface processes, and are increasingly incorporated into Earth System Models (ESMs), which makes it possible to study the coupled dynamics of the land-biosphere and the climate. In this work we describe a number of modifications to the LPJ-GUESS DGVM, aimed at enabling direct integration into an ESM. These include energy balance closure, the introduction of a sub-daily time step, a new radiative transfer scheme, and improved soil physics. The implemented modifications allow the model (LPJ-GUESS/LSM) to simulate the diurnal exchange of energy, water and CO2 between the land-ecosystem and the atmosphere. A site-based evaluation against FLUXNET2015 data shows reasonable agreement between observed and modeled sensible and latent heat fluxes. Differences in predicted ecosystem function between standard LPJ-GUESS and LPJ-GUESS/LSM vary across land cover types, but the emergent ecosystem composition and structure are consistent between the two versions. We find that the choice of stomatal conductance model has a major impact on the model's predictions. The new LSM implementation described in this work lays the foundation for using the well established LPJ-GUESS DGVM as an alternative LSM in coupled land-biosphere-climate studies, where an accurate representation of ecosystem processes is essential.

Land biosphere processes are of central importance to the climate system. Specifically, biological processes interact with the atmosphere through a variety of feedback loops that modulate energy, water and CO2 fluxes between the land surface and the atmosphere across a wide range of temporal and spatial scales. Human land use and land cover modification add a further level of complexity to land-atmosphere interactions. Dynamic Global Vegetation Models (DGVMs) attempt to capture these land surface processes, and are increasingly incorporated into Earth System Models (ESMs), which makes it possible to study the coupled dynamics of the land-biosphere and the climate. In this work we describe a number of modifications to the LPJ-GUESS DGVM, aimed at enabling direct integration into an ESM. These include energy balance closure, the introduction of a sub-daily time step, a new radiative transfer scheme, and improved soil physics. The implemented modifications allow the model (LPJ-GUESS/LSM) to simulate the diurnal exchange of energy, water and CO2 between the land-ecosystem and the atmosphere. A site-based evaluation against FLUXNET2015 data shows reasonable agreement between observed and modeled sensible and latent heat fluxes. Differences in predicted ecosystem function between standard LPJ-GUESS and LPJ-GUESS/LSM vary across land cover types, but the emergent ecosystem composition and structure are consistent between the two versions. We find that the choice of stomatal conductance model has a major impact on the model's predictions. The new LSM implementation described in this work lays the foundation for using the well established LPJ-GUESS DGVM as an alternative LSM in coupled land-biosphere-climate studies, where an accurate representation of ecosystem processes is essential.

Abstract. The simulation of current and projected wildfires is essential for predicting crucial aspects of vegetation patterns, biogeochemical cycling as well as pyrogenic emissions across the African continent. This study uses a data-driven approach to parameterize two burned area models applicable to dynamic vegetation models (DVMs) and Earth system models (ESMs). We restricted our analysis to variables for which either projections based on climate scenarios are available, or that are calculated by DVMs, and we consider a spatial scale of one degree as the scale typical for DVMs and ESMs. By using the African continent here as an example, an analogue approach could in principle be adopted for other regions, for global scale dynamic burned area modelling. We used 9 years of data (2000–2008) for the variables: precipitation over the last dry season, the last wet season and averaged over the last 2 years, a fire-danger index (the Nesterov index), population density, and annual proportion of area burned derived from the MODIS MCD45A1 product. Two further variables, tree and herb cover were only available for 2001 as a remote sensing product. Since the effect of fires on vegetation depends strongly on burning conditions, the timing of wildfires is of high interest too, and we were able to relate the seasonal occurrence of wildfires to the daily Nesterov index. We parameterized two generalized linear models (GLMs), one with the full variable set (model VC) and one considering only climate variables (model C). All introduced variables resulted in an increase in model performance. Model VC correctly predicts the spatial distribution and extent of fire prone areas though the total variability is underrepresented. Model VC has a much lower performance in both aspects (correlation coefficient of predicted and observed ratio of burned area: 0.71 for model VC and 0.58 for model C). We expect the remaining variability to be attributed to additional variables which are not available at a global scale and thus not incorporated in this study as well as its coarse resolution. An application of the models using climate hindcasts and projections ranging from 1980 to 2060 resulted in a strong decrease of burned area of ca. 20–25%. Since wildfires are an integral part of land use practices in Africa, their occurrence is an indicator of areas favourable for food production. In absence of other compensating land use changes, their projected decrease can hence be interpreted as a indicator for future loss of such areas.

Abstract. Fires play an important role in the Earth system but remain complex phenomena that are challenging to model numerically. Here, we present the first version of BuRNN, a data-driven model simulating burned area on a global 0.5° × 0.5° grid with a monthly time resolution. We trained Long Short-Term Memory networks to predict satellite-based burned area (GFED5) from a range of climatic, vegetation and socio-economic parameters. We employed a region-based cross-validation strategy to account for the high spatial autocorrelation in our data. BuRNN outperforms the process-based fire models participating in ISIMIP3a on a global scale across a wide range of metrics. Regionally, BuRNN outperforms almost all models across a set of benchmarking metrics in all regions. Through explainable AI we unravel the difference in regional drivers of burned area in our models, showing that the presence/absence of bare ground and C4 grasses along with the fire weather index have the largest effects on our predictions of burned area. Lastly, we used BuRNN to reconstruct global burned area for 1901–2019 and compare the simulations against independent long-term historical fire observation databases in five countries and the EU. Our approach highlights the potential of machine learning to improve burned area simulations and our understanding of past fire behaviour.

Defining the appropriate scale over which to conduct a study in the marine environment is critical to achieving appropriate scientific, management, mitigation and conservation objectives. This paper focuses on applications of passive acoustic technologies over a range of spatial and temporal scales. It is divided into sections dealing with archival and real-time passive acoustic sensor applications. Each section assesses the principles behind using the respective technology and pro- vides recent examples of research and management applications for marine mammals and fish. The section on archival sensors highlights the need for continued development of automated acoustic detectors to assess large data sets. Case studies are presented of detectors developed for determining seasonal occurrence and distribution of haddock sounds and humpback whale vocalizations. Also presented are studies of other applications using archival sensors: tracking singing humpback whales in Brazil, using vocalizations to assess the reproductive strategies of Arctic bearded seals and assessing regional variability in call patterns for North Atlantic right whales. The section on real-time passive acoustic sensors focuses on real-time buoys and towed arrays. Case studies presented include a real-time buoy system used for monitoring endangered North Atlantic right whales and a station- ary autonomous array providing real-time access to Antarctic acoustic data. The value of using towed arrays for real-time applications is also assessed, and a case study is provided on the use of towed arrays to improve abundance estimates of North Pacific cetaceans and to better understand vocaliza- tion behaviors.

Abstract. Fire interacts with many parts of the Earth system. However, its drivers are myriad and complex, interacting differently in different regions depending on prevailing climate regimes, vegetation types, socioeconomic development, and land use and management. Europe is facing strong increases in projected fire weather danger as a consequence of climate change and has experienced extreme fire seasons and events in recent years. Here, we focus on understanding and simulating burnt area across a European study domain using remote sensing data and generalised linear models (GLMs). We first examined fire occurrence across land cover types and found that all non-cropland vegetation (NCV) types (comprising 26 % of burnt area) burnt with similar spatial and temporal patterns, which were very distinct from those in croplands (74 % of burnt area). We then used GLMs to predict cropland and NCV burnt area at ∼9×9 km and monthly spatial and temporal resolution, respectively, which together we termed BASE (Burnt Area Simulator for Europe). Compared to satellite burnt area products, BASE effectively captured the general spatial and temporal patterns of burning, explaining 32 % (NCV) and 36 % (cropland) of the deviance, and performed similarly to state-of-the-art global fire models. The most important drivers were fire weather and monthly indices derived from gross primary productivity followed by coarse socioeconomic indicators and vegetation properties. Crucially, we found that the drivers of cropland and NCV burning were very different, highlighting the importance of simulating burning in different land cover types separately. Through the choice of predictor variables, BASE was designed for coupling with dynamic vegetation and Earth system models and thus enabling future projections. The strong model skill of BASE when reproducing seasonal and interannual dynamics of NCV burning and the novel inclusion of cropland burning indicate that BASE is well suited for integration in land surface models. In addition to this, the BASE framework may serve as a basis for further studies using additional predictors to further elucidate drivers of fire in Europe. Through these applications, we suggest BASE may be a useful tool for understanding, and therefore adapting to, the increasing fire risk in Europe.

Settlement-driven, multiscale demographic patterns of large benthic decapods in the Gulf of Maine