Estimating the probability of wildfire occurrence in Mediterranean landscapes using Artificial Neural Networks

Wildfires are a major disturbance in the Mediterranean Basin and an ecological factor that constantly alters the landscape. In this context, it is crucial to understand where wildfires are more likely to occur as well as the drivers guiding them in complex landscapes such as the Mediterranean area. The objectives of this study are to estimate wildfire probability occurrence as a function of biophysical and human-related drivers, to provide an assessment of the relative impact of each driver and analyze the performance of machine learning techniques compared to traditional regression modeling. By employing an Artificial Neural Network model and fire data (2004–2012), we estimated wildfire probability across two geographical regions covering most of the Italian territory: Alpine and subalpine region and Insular and peninsular region. The high classification accuracy (0.68 for the Alpine and subalpine region and 0.76 for the Insular and peninsular region) and good performances of the technique (AUC values of 0.82 and 0.76, respectively) suggest that our model can be used in the areas studied to assess wildfire probability occurrence. We compared our model with a logistic function, which showed a weaker predictive power (AUC values of 0.78 for the Alpine and subalpine region and 0.65 for the Insular and peninsular region) compared to the Artificial Neural Network. In addition, we assessed the importance of each variable by isolating it in the model. The importance of an individual variable differed between the two regions, underscoring the high diversity of wildfire occurrence drivers in Mediterranean landscapes. Results show that in the Alpine and subalpine region, the presence of forest is the most important variable, while climate resulted as being the most important variable in the Insular and peninsular region. The majority of areas recently affected by large wildfires in both regions have been correctly classified by the ANN model as ‘high fire probability’. Hence, the use of an Artificial Neural Network is efficient and robust for understanding the probability of wildfire occurrence in Italy and other similar complex landscapes.     

Use of Forest Ecosystem Classification systems in fire management

Forest Ecosystem Classification (FEC) systems have been used in the past mainly for forest management decision-making. FEC systems can also serve an important role for decision-making in other disciplines, such as fire management for both wildfire suppression and prescribed burning operations. FEC systems can provide an important means of identifying potential fuels that may be present on a forest site. This fuel information, in combination with current fire weather conditions, as determined by the Canadian Forest Fire Weather Index (FWI) system, can assist fire managers in determining potential fire behaviour if ignition should occur. FEC systems provide a means of identifying the possible presence of a live understory vegetation component, a fuel layer that has been largely ignored in the past due to a lack of information. Dense understory vegetation can produce a very moist microlimate that can effectively hinder fire spread. The use of FEC systems can help in setting priorities on which wildfires need to be attacked aggressively. For prescribed burning, FEC systems can assist in achieving burn objectives better and more safely.     

An integrated methodology for estimation of forest fire-loss using geospatial information

These days, wildfires are prevalent in almost all areas of the world. Researchers have been actively analyzing wildfire damage using a variety of satellite images and geospatial datasets. This paper presents a method for detailed estimation of wildfire losses using various geospatial datasets and an actual case of wildfire at Kang-Won-Do, Republic of Korea in 2005. A set of infrared (IR) aerial images acquired after the wildfire were used to visually delineate the damaged regions, and information on forest type, diameter class, age class, and canopy density within the damaged regions was retrieved from GIS layers of the Korean national forest inventory. Approximate tree heights were computed from airborne LIDAR and verified by ground LIDAR datasets. The corresponding stand volumes were computed using tree volume equations (TVE). The proposed algorithm can efficiently estimate fire loss using the geospatial information; in the present case, the total fire loss was estimated as $5.9 million, which is a more accurate estimate than $4.5 million based on conventional approach. The proposed method can be claimed as a powerful alternative for estimating damage caused by wildfires, because the aerial image interpretation can delineate and analyze damaged regions in a comprehensive and consistent manner; moreover, LIDAR datasets and national forest inventory data can significantly reduce field work.     

Advancing effects analysis for integrated, large-scale wildfire risk assessment

In this article, we describe the design and development of a quantitative, geospatial risk assessment tool intended to facilitate monitoring trends in wildfire risk over time and to provide information useful in prioritizing fuels treatments and mitigation measures. The research effort is designed to develop, from a strategic view, a first approximation of how both fire likelihood and intensity influence risk to social, economic, and ecological values at regional and national scales. Three main components are required to generate wildfire risk outputs: (1) burn probability maps generated from wildfire simulations, (2) spatially identified highly valued resources (HVRs), and (3) response functions that describe the effects of fire (beneficial or detrimental) on the HVR. Analyzing fire effects has to date presented a major challenge to integrated risk assessments, due to a limited understanding of the type and magnitude of changes wrought by wildfire to ecological and other nonmarket values. This work advances wildfire effects analysis, recognizing knowledge uncertainty and appropriately managing it through the use of an expert systems approach. Specifically, this work entailed consultation with 10 fire and fuels program management officials from federal agencies with fire management responsibilities in order to define quantitative resource response relationships as a function of fire intensity. Here, we demonstrate a proof-of-concept application of the wildland fire risk assessment tool, using the state of Oregon as a case study.     

Solar plants,environmental degradation and local socioeconomic contexts: A case study in a Mediterranean country

Photovoltaic plants developed on rural land are becoming a common infrastructure in the Mediterranean region and may contribute, at least indirectly, to various forms of environmental degradation including landscape deterioration, land take, soil degradation and loss in traditional cropland and biodiversity. Our study illustrates a procedure estimating (i) the extension of ground-mounted photovoltaic fields at the municipal scale in Italy and (ii) inferring the socioeconomic profile of the Italian municipalities experiencing different expansion rates of ground-mounted photovoltaic fields over the last years (2007-2014). The procedure was based on diachronic information derived from official data sources integrated into a geographical decision support system. Our results indicate that the surface area of ground-mounted photovoltaic fields into rural land grew continuously in Italy between 2007 and 2014 with positive and increasing growth rates observed during 2007-2011 and positive but slightly decreasing growth rates over 2012-2014, as a result of market saturation and policies containing the diffusion of solar plants on greenfields. We found important differences in the density of ground-mounted solar plants between northern and southern Italian municipalities. We identified accessible rural municipalities in southern Italy with intermediate population density and large availability of non-urban land as the most exposed to the diffusion of solar plants on greenfields in the last decade. Our approach is a promising tool to estimate changes in the use of land driven by the expansion of photovoltaic fields into rural land.     

Structural characteristics of post-wildfire and clearcut landscapes

A continuing discussion in the field of ecology and forest management concerns the implications of clearcutting as a functional replacement for wildfire in disturbance-driven ecosystems. At the landscape level, spatial pattern has been shown to influence many ecologically important processes. Satellite imagery allows the evaluation of structural patterns created by alternative forest management activities at broad scales. In Northwestern Ontario, both clearcutting and wildfire have occurred over large contiguous areas. Spatial characteristics including composition, patch size, patch shape, and interspersion were calculated from classified Landsat Thematic Mapper (TM) data at two thematic scales and used to compare post-wildfire and clearcut landscapes. Patches in the clearcut landscape were found to be larger in size, and had a more irregular shape than those in the wildfire landscape. Differences in landscape structure were much more pronounced at broad scales than at fine thematic scales.     

Spatial Distribution of Forest Fires and Controlling Factors in Andhra Pradesh, India Using Spot Satellite Datasets

Fires are one of the major causes of forest disturbance and destruction in several dry deciduous forests of southern India. In this study, we use remote sensing data sets in conjunction with topographic, vegetation, climate and socioeconomic factors for determining the potential causes of forest fires in Andhra Pradesh, India. Spatial patterns in fire characteristics were analyzed using SPOT satellite remote sensing datasets. We then used nineteen different metrics in concurrence with fire count datasets in a robust statistical framework to arrive at a predictive model that best explained the variation in fire counts across diverse geographical and climatic gradients. Results suggested that, of all the states in India, fires in Andhra Pradesh constituted nearly 13.53% of total fires. District wise estimates of fire counts for Andhra Pradesh suggested that, Adilabad, Cuddapah, Kurnool, Prakasham and Mehbubnagar had relatively highest number of fires compared to others. Results from statistical analysis suggested that of the nineteen parameters, population density, demand of metabolic energy (DME), compound topographic index, slope, aspect, average temperature of the warmest quarter (ATWQ) along with literacy rate explained 61.1% of total variation in fire datasets. Among these, DME and literacy rate were found to be negative predictors of forest fires. In overall, this study represents the first statewide effort that evaluated the causative factors of fire at district level using biophysical and socioeconomic datasets. Results from this study identify important biophysical and socioeconomic factors for assessing ‘forest fire danger’ in the study area. Our results also identify potential ‘hotspots’ of fire risk, where fire protection measures can be taken in advance. Further this study also demonstrate the usefulness of best-subset regression approach integrated with GIS, as an effective method to assess ‘where and when’ forest fires will most likely occur.     

Trends in land use and land cover change in the protected and communal areas of the Zambezi Region,Namibia

Land management decisions have extensively modified land use and land cover in the Zambezi Region. These decisions are influenced by land tenure classifications, legislation, and livelihoods. Land use and land cover change is an important indicator for quantifying the effectiveness of different land management strategies. However, there has been no evidence on whether protected or communal land tenure is more affected by land use and land cover changes in southern Africa and particularly Namibia. Our study attempted to fill this gap by analyzing the relationship between land use and land cover change and land tenure regimes stratified according to protected and communal area in the Zambezi Region. Multi-temporal Landsat TM and ETM+ imagery were used to determine the temporal dynamics of land use and land cover change from 1984 to 2010. The landscape showed distinctive modifications over the study period; broad trends include the increase in forest land after 1991. However, changes were not uniform across the study areas. Two landscape development stages were deduced: (1) 1984–1991 represented high deforestation and gradual increase in shrub land; (2) 1991–2000 and 2000–2010 represented lower deforestation and slower agropastoral expansion. The results further show clear patterns of the dynamics, magnitude, and direction of land use and land cover change by tenure regime. The study concluded that land tenure has a direct impact on land use and land cover, since it may restrict some activities carried out on the land in the Zambezi Region.     

An Ecosystem Approach to Combat Desertification on the Colorado Plateau

Desertification of shrub and grassland into pinyon-juniper woodland is occurring over much of the Colorado Plateau in the southwestern United States. As trees invade, they out-compete shrubs and grasses, increasing erosion rates and reducing infiltration of moisture into the soil. This has caused habitat problems for wildlife, and reduced forage for livestock. These impacts also affect the human communities that rely on ranching and tourism related to hunting. Past land use and management practices including heavy livestock grazing, fire suppression and introduction of exotic annual plants are believed to have led to current conditions. The Montrose office of the Bureau of Land Management has implemented an ecosystem-based program to reverse the desertification process on public land. The program is centered on detailed landscape objectives describing the desired vegetation mosaic on 360 000 ha of public land. The objectives outline proportions of plant seral stages and arrays of patch sizes for each planning unit. These objectives are based on priority management issues and the need to replicate a natural vegetation mosaic. Where the existing mosaic does not meet objectives, mechanical vegetation treatments and prescribed fire are used to create early and mid-seral patches on the ground. This restored vegetation pattern and type should be sustained over time through a natural fire regime and improved livestock management. Because many uncertainties exist, an adaptive management process is being used that allows mosaic objectives to be changed or processes modified where monitoring or scientific research indicate a need.     

Analysing land cover changes for understanding of forest dynamics using temporal forest management plans

This study analyses forest dynamics and land use/land cover change over a 43-year period using spatial-stand-type maps of temporal forest management plans of Karaisal? Forest Enterprise in the Eastern Mediterranean Region of Turkey. Stand parameters (tree species, crown closures and developmental stages) of the dynamics and changes caused by natural or artificial intervention were introduced and mapped in a Geographic Information System (GIS) and subjected to fragmentation analysis using FRAGSTATS. The Karaisal? Forest Enterprise was first planned in 1969 and then the study area was planned under the Mediterranean Forest Use project in 1991 and five-term forest management plans were made. In this study, we analysed only four periods (excluding 1982 revision plans): 1969, 1991, 2002 and 2012. Between 1969 and 2012, overall changes included a net increase of 3,026 ha in forested areas. Cumulative forest improvement accounted for 2.12 % and the annual rate of total forest improvement averaged 0.08 %. In addition, productive forest areas increased from 36,174 to 70,205 ha between 1969 and 2012. This translates into an average annual productive forest improvement rate of 1.54 %. At the same time, fully covered forest areas with crown closure of “3” (>70 %) increased about 21,321 ha, and young forest areas in developmental stage of “a” (diameter at breast height (dbh)?<?8 cm) increased from 716 to 13,305 ha over the 43-year study period. Overall changes show that productive and fully covered forest areas have increased egregiously with a focus on regenerated and young developmental stages. A spatial analysis of metrics over the 43-year study period indicated a more fragmented landscape resulting in a susceptible forest to harsh disturbances.     

Effects of Forest Management Practices on Mid-Atlantic Streams

Agricultural and urban land use activities have affected stream ecosystems throughout the mid-Atlantic region. However, over 60% of the mid-Atlantic region is forested. A study was conducted to investigate the effects of management practices on forested stream ecosystems throughout the mid-Atlantic region. The study consisted of two phases: Phase 1 was a literature synthesis of information available on the effects of forest management practices on stream hydrology, erosion and sedimentation, riparian habitat alteration, chemical addition, and change in biotic diversity in the mid-Atlantic region. In Phase 2, data from mid-Atlantic streams were analyzed to assess the effects of forest land use on stream quality at the regional scale. Typically, it is the larger order streams in which monitoring and assessment occurs—3^rd order or higher streams. The impacts of forest management practices, particularly hydrologic modifications and riparian buffer zone alteration, occur predominantly in first and second order streams with cumulative impacts translating to higher order streams. Based on the literature review and mid-Atlantic Highland streams analysis, there are short-term (e.g., 2 to 5 years) effects of forest management practices on stream quality at local scales. However, signatures of cumulative effects from forest management practices are not apparent at regional scales in the Highlands. In general, forested land use is associated with good stream quality in the region compared with other land use practices.     

Assessing the impact of multiple drivers of land sensitivity to desertification in a Mediterranean country

Experiencing climate changes and increased human pressure, Mediterranean regions are considered representative hotspots of desertification. However, relatively few studies have been devoted to quantify the individual impact of different factors shaping land sensitivity to desertification in these contexts. Our study contributes to this deserving (positive and normative) issue with a diachronic analysis of the impact of multiple drivers of desertification risk on six indicators of land sensitivity based on the Environmentally Sensitive Area (ESA) approach. Indicators (average and maximum ESA score, coefficient of variation and normalized range in the ESA scores, share of ‘fragile’ and ‘critical’ areas in total landscape) were calculated in 777 rural districts of Italy at three time points (early-1960s, early-1990s, and early-2010s). Multivariate models were used to determine the impact of 12 predictors (climate, soil, vegetation, and land management quality) on each indicator of land sensitivity. Results of the analysis identified two non-redundant dimensions respectively associated with the average level of land sensitivity and its intrinsic variability across space. Impacts of climate and vegetation qualities on the level of land sensitivity were high, decreasing over time, and more intense respectively in Northern and Southern Italy. Impacts of soil and land management qualities were moderate, increasing over time, and involving almost all the country's area. Our study emphasizes the role of context-based measures promoting sustainable land management. The ‘local’ dimension proved to be crucial in any strategy of risk mitigation undertaken at disaggregated spatial scales.     

Decadal time-scale monitoring of forest fires in Similipal Biosphere Reserve,India using remote sensing and GIS

Analyzing the spatial extent and distribution of forest fires is essential for sustainable forest resource management. There is no comprehensive data existing on forest fires on a regular basis in Biosphere Reserves of India. The present work have been carried out to locate and estimate the spatial extent of forest burnt areas using Resourcesat-1 data and fire frequency covering decadal fire events (2004–2013) in Similipal Biosphere Reserve. The anomalous quantity of forest burnt area was recorded during 2009 as 1,014.7 km². There was inconsistency in the fire susceptibility across the different vegetation types. The spatial analysis of burnt area shows that an area of 34.2 % of dry deciduous forests, followed by tree savannah, shrub savannah, and grasslands affected by fires in 2013. The analysis based on decadal time scale satellite data reveals that an area of 2,175.9 km² (59.6 % of total vegetation cover) has been affected by varied rate of frequency of forest fires. Fire density pattern indicates low count of burnt area patches in 2013 estimated at 1,017 and high count at 1,916 in 2004. An estimate of fire risk area over a decade identifies 12.2 km² is experiencing an annual fire damage. Summing the fire frequency data across the grids (each 1 km²) indicates 1,211 (26 %) grids are having very high disturbance regimes due to repeated fires in all the 10 years, followed by 711 grids in 9 years and 418 in 8 years and 382 in 7 years. The spatial database offers excellent opportunities to understand the ecological impact of fires on biodiversity and is helpful in formulating conservation action plans.     

A comparative assessment of land cover dynamics of three protected forest areas in tropical eastern Africa

Processes of deforestation, known to threaten tropical forest biodiversity, have not yet been studied sufficiently in East Africa. To shed light on the patterns and causes of human influences on protected forest ecosystems, comparisons of different study areas regarding land cover dynamics and potential drivers are needed. We analyze the development of land cover since the early 1970s for three protected East African rainforests and their surrounding farmlands and assess the relationship between the observed changes in the context of the protection status of the forests. Processing of Landsat satellite imagery of eight or seven time steps in regular intervals results in 12 land cover classes for the Kakamega–Nandi forests (Kenya) and Budongo Forest (Uganda) whereas ten are distinguished for Mabira Forest (Uganda). The overall classification accuracy assessed for the year 2001 or 2003 is similarly high for all three study areas (81% to 85%). The time series reveal that, despite their protection status, Kakamega–Nandi forests and Mabira Forest experienced major forest decrease, the first a continuous forest loss of 31% between 1972/1973 and 2001, the latter an abrupt loss of 24% in the late 1970s/early 1980s. For both forests, the temporally dense time series show short-term fluctuations in forest classes (e.g., areas of forest regrowth since the 1980s or exotic secondary bushland species from the 1990s onwards). Although selectively logged, Budongo Forest shows a much more stable forest cover extent. A visual overlay with population distribution for all three regions clearly indicates a relationship between forest loss and areas of high population density, suggesting population pressure as a main driver of deforestation. The revealed forest losses due to local and commercial exploitation further demonstrate that weak management impedes effective forest protection in East Africa.     

Restoring the Great Basin Desert, U.S.A.: Integrating Science, Management, and People

The Great Basin Desert lies between the Sierra Nevada Mountains to the west and the Rocky Mountains to the east. Nearly 60% of the area’s deserts and mountains (roughly 30 million ha) are managed by the U. S. Department of Interior’s Bureau of Land Management. This area is characterized by low annual precipitation, diverse desert plant communities, and local economies that depend on the products (livestock grazing, recreation, mining, etc.) produced by these lands. The ecological and economic stability of the Great Basin is increasingly at risk due to the expansion of fire-prone invasive species and increase in wildfires. To stem this loss of productivity and diversity in the Great Basin, the BLM initiated the “Great Basin Restoration Initiative” in 1999 after nearly 0.7 million ha of the Great Basin burned in wildfires. The objective of the Great Basin Restoration Initiative is to restore plant community diversity and structure by improving resiliency to disturbance and resistance to invasive species over the long-term. To accomplish this objective, a strategic plan has been developed that emphasizes local participation and reliance on appropriate science to ensure that restoration is accomplished in an economical and ecologically appropriate manner. If restoration in the Great Basin is not successful, desertification and the associated loss of economic stability and ecological integrity will continue to threaten the sustainability of natural resources and people in the Great Basin.     

Compositing climate change vulnerability of a Mediterranean region using spatiotemporally dynamic proxies for ecological and socioeconomic impacts and stabilities

The study presents a new methodology to quantify spatiotemporal dynamics of climate change vulnerability at a regional scale adopting a new conceptual model of vulnerability as a function of climate change impacts, ecological stability, and socioeconomic stability. Spatiotemporal trends of equally weighted proxy variables for the three vulnerability components were generated to develop a composite climate change vulnerability index (CCVI) for a Mediterranean region of Turkey combining Landsat time series data, digital elevation model (DEM)-derived data, ordinary kriging, and geographical information system. Climate change impact was based on spatiotemporal trends of August land surface temperature (LST) between 1987 and 2016. Ecological stability was based on DEM, slope, aspect, and spatiotemporal trends of normalized difference vegetation index (NDVI), while socioeconomic stability was quantified as a function of spatiotemporal trends of land cover, population density, per capita gross domestic product, and illiteracy. The zones ranked on the five classes of no-to-extreme vulnerability were identified where highly and moderately vulnerable lands covered 0.02% (12 km²) and 11.8% (6374 km²) of the study region, respectively, mostly occurring in the interior central part. The adoption of this composite CCVI approach is expected to lead to spatiotemporally dynamic policy recommendations towards sustainability and tailor preventive and mitigative measures to locally specific characteristics of coupled ecological–socioeconomic systems.     

Monitoring and analysis of natural vegetation in a Special Protected Area of Mountain Antichasia—Meteora, central Greece

Natural ecosystems are renewable resources with special environmental, social, and economical attributes and characteristics. The increasing need of human beings for a better environment results in the use of new technologies that offer many advantages in detecting changes in the ecosystems. Remote sensing tools, technology, and the spatial analysis of the Geographic Information System were used in determining any changes in this study which attempts to classify land cover over a 10-year period. The study area is in Thessaly, central Greece, and has been classified as a Special Protection Area, because of its important wild fauna. The results have shown that current technologies can be used for modeling environmental parameters which improve our knowledge of the attributes, characteristics, situation, trends, and changes of natural ecosystems. The changes over time that have been observed result from the development of the vegetation or to anthropogenic and socioeconomic reasons. Rational range management will be a very comprehensive tool for farmers. This action will have a positive impact on flora in the rangelands. The core strategy is to combine forest, pasture, and livestock so that each component produces usable products.     

A data-mining approach to predict influent quality

In wastewater treatment plants, predicting influent water quality is important for energy management. The influent water quality is measured by metrics such as carbonaceous biochemical oxygen demand (CBOD), potential of hydrogen, and total suspended solid. In this paper, a data-driven approach for time-ahead prediction of CBOD is presented. Due to limitations in the industrial data acquisition system, CBOD is not recorded at regular time intervals, which causes gaps in the time–series data. Numerous experiments have been performed to approximate the functional relationship between the input and output parameters and thereby fill in the missing CBOD data. Models incorporating seasonality effects are investigated. Four data-mining algorithms—multilayered perceptron, classification and regression tree, multivariate adaptive regression spline, and random forest—are employed to construct prediction models with the maximum prediction horizon of 5 days.     

Land mismatches,urban growth and spatial planning: A contribution to metropolitan sustainability

The unequal growth of population and buildings in metropolitan regions reflects dispersed urban expansion. This study illustrates an operational framework grounded on a diachronic analysis of urbanization processes in advanced economies that provides a comprehensive evaluation of the mismatch between resident population and building stock. Studying the urban cycle of a European city (Athens, Greece), a mismatch indicator was derived at the municipal level as the elasticity rate of resident population and total building stock changes over 7 time intervals between 1920 and 2010. Results indicate that divergences in population and building stock growth rates increased since the early 1980s. The population-buildings mismatch displays an increasingly asymmetric spatial distribution, evidencing more or less accelerated paths toward dispersed settlements that may outline unsustainable forms of land management. Municipalities with a compact morphology at the beginning of the study period showed a higher rate of self-contained urban expansion than municipalities with more dispersed settlements. A comparative analysis of the impact of town planning on enlarging population-settlement mismatches was finally proposed as a basic knowledge to sustainable land management in (rapidly expanding) metropolitan regions.