Overwintering Peat Fires in Russia’s Boreal Forests: Persistence, Detection, and Suppression

Peat fires are a major hazard to human and animal health and can negatively impact livelihoods. Once peat fires start to burn, they are difficult to extinguish and can continue to burn for months, destroying biomass and contributing to carbon emissions globally. In areas with limited accessibility and periods of thick haze and fog, these fires are difficult to detect, localize, and tackle. To address this problem, thermal infrared cameras mounted on drones can provide a potential solution since they allow large areas to be surveyed relatively quickly and can detect thermal radiation from fires above and below the peat surface. This paper describes a deep learning pipeline that detects and segments peat fires in thermal images. Controlled peat fires were constructed under varying environmental conditions and thermal images were taken to form a dataset for our pipeline. A semi-automated approach was adopted to label images using Otsu’s adaptive thresholding technique, which significantly reduces the required effort often needed to tag objects in images. The proposed method uses a pre-trained ResNet-50 model as a backbone (encoder) for feature extraction and is augmented with a set of up-sampling layers and skip connections, like the UNet architecture. The experimental results show that the model can achieve an IOU score of 87.6% on an unseen test set of thermal images containing peat fires. In comparison, a MobileNetV2 model trained under the same experimental conditions achieved an IOU score of 57.9%. In addition, the model is robust to false positives, which is indicated by a precision equal to 94.2%. To demonstrate its practical utility, the model was also tested on real peat wildfires, and the results are promising, as indicated by a high IOU score of 90%. Finally, a geolocation algorithm is presented to identify the GNSS location of these fires once they are detected in an image to aid fire-fighting responses. The proposed scheme was built using a web-based platform that performs offline detection and allows peat fires to be geolocated.

Desiccation of soil is the gradual process of moisture loss from a fully or partially saturated condition to a dry state. If the soil is constrained against shrinkage during desiccation, it will begin to crack when tensile stresses in the soil exceed the tensile strength. Fundamental to understanding desiccation crack development in clayey soils is the accurate determination of the tensile strength during desiccation. Although numerous published studies have focused on soil desiccation and soil tensile strength, very little has been focused on experimental determination of tensile strength during desiccation. This paper presents the first generation of a novel apparatus designed to determine tensile strength during desiccation. During drying the tensile forces in a thin compacted soil layer are measured while simultaneously monitoring moisture content changes and crack formation via time-lapse photography. Results are presented from desiccation tests performed on two clayey compacted soils of medium to high plasticity. Soil water characteristic curves (SWCCs) were determined to relate the tensile strength to suction. The results demonstrate that the tensile strength of compacted soil is influenced by initial water content, suction, dry unit weight, soil pore structure, and soil type. Measured tensile strengths compare favorably with results of a theoretical model that considers microstructural and macrostructural aspects of unsaturated compacted clayey soil.

Snow cover phenology plays an important role in vegetation dynamics over the boreal region, but the observed evidence of this interaction is limited. A comprehensive understanding of the changes in vegetation dynamics and snow cover phenology as well as the interactions between them is urgently needed. To investigate this, we calculated two indicators, the start of the growing season (SOS) and the annual maximum enhanced vegetation index (EVImax), as proxies of vegetation dynamics using the Moderate Resolution Imaging Spectroradiometer (MODIS) enhanced vegetation index (EVI). Snow cover duration (SCD) and snow cover end date (SCE) were also extracted from MODIS snow cover datasets. Then, we quantified the spatial-temporal changes in vegetation dynamics and snow cover phenology as well as the relationship between them over the boreal region. Our results showed that the EVImax generally demonstrated an increasing trend, but SOS varied in different regions and vegetation types from 2001 to 2014. The earlier onset of SOS was mainly concentrated in the Siberian boreal region. In the Eurasian boreal region, we observed an advance in the SCE and decrease in the SCD, while in the North American boreal region, the spatial distribution of the trends exhibited substantial heterogeneity. Our results also indicated that the snow cover phenology had significant impacts on the SOS and the EVImax, but the effects varied in different regions, vegetation types, and climate gradients. Our findings provide strong evidence of the interaction between vegetation dynamics and snow cover phenology, and snow cover should be considered when analyzing future vegetation dynamics in the boreal region.

An integrated ice tracking and mitigation system on the stagnation line of a cylindrical surface based on thermal imaging and electro-thermal elements

Abstract. Detailed information on seasonal snow cover and depth is essential to the understanding of snow processes, to operational forecasting, and as input for hydrological models. Recent advances in uncrewed or unmanned aircraft systems (UASs) and structure from motion (SfM) techniques have enabled low-cost monitoring of spatial snow depth distribution in resolutions of up to a few centimeters. Here, we study the spatiotemporal variability in snow depth and interactions between snow and vegetation in different subarctic landscapes consisting of a mosaic of conifer forest, mixed forest, transitional woodland/shrub, and peatland areas. To determine the spatiotemporal variability in snow depth, we used high-resolution (50 cm) snow depth maps generated from repeated UAS–SfM surveys in the winter of 2018/2019 and a snow-free bare-ground survey after snowmelt. Due to poor subcanopy penetration with the UAS–SfM method, tree masks were utilized to remove canopy areas and the area (36 cm) immediately next to the canopy before analysis. Snow depth maps were compared to the in situ snow course and a single-point continuous ultrasonic snow depth measurement. Based on the results, the difference between the UAS–SfM survey median snow depth and single-point measurement increased for all land cover types during the snow season, from +5 cm at the beginning of the accumulation to −16 cm in coniferous forests and −32 cm in peatland during the melt period. This highlights the poor representation of point measurements in selected locations even on the subcatchment scale. The high-resolution snow depth maps agreed well with the snow course measurement, but the spatial extent and resolution of maps were substantially higher. The snow depth range (5th–95th percentiles) within different land cover types increased from 17 to 42 cm in peatlands and from 33 to 49 cm in the coniferous forest from the beginning of the snow accumulation to the melt period. Both the median snow depth and its range were found to increase with canopy density; this increase was greatest in the conifer forest area, followed by mixed forest, transitional woodland/shrub, and open peatlands. Using the high-spatial-resolution data, we found a systematic increase (2–20 cm) and then a decline in snow depth near the canopy with increasing distance (from 1 to 2.5 m) of the peak value through the snow season. This study highlights the applicability of the UAS–SfM in high-resolution monitoring of snow depth in multiple land cover types and snow–vegetation interactions in subarctic and remote areas where field data are not available or where the available data are collected using classic point measurements or snow courses.

<strong class="journal-contentHeaderColor">Abstract.</strong> Detailed information on seasonal snow cover and depth is essential to the understanding of snow processes, operational forecasting, and as input for hydrological models. Recent advances in unmanned aircraft systems (UASs) and structure from motion (SfM) techniques have enabled low-cost monitoring of spatial snow depth distribution in resolutions up to a few centimeters. Here, we study the spatiotemporal variability of snow depth and interactions between snow and vegetation in different subarctic landscapes consisting of a mosaic of conifer forest, mixed forest, transitional woodland/shrub, and peatland areas. To determine the spatiotemporal variability of snow depth, we used high-resolution (50 cm) snow depth maps generated from repeated UAS-SfM surveys in the winter of 2018/2019 and a snow-free bare ground survey after snowmelt. Due to poor sub-canopy penetration with the UAS-SfM method, tree masks were utilized to remove canopy areas and the area (36 cm) immediately next to the canopy before analysis. Snow depth maps were compared to the in-situ snow course and a single-point continuous ultrasonic snow depth measurement. Based on the results, the difference between the UAS-SfM survey median snow depth and single-point measurement increased for all land cover types during the snow season, from +5 cm at the beginning of the accumulation to -16 cm in coniferous forests and -32 cm in peatland during the melt period. This highlights the poor representation of point measurements even on the sub-catchment scale. The high-resolution snow depth maps agreed well with the snow course measurement, but the spatial extent and resolution of maps were substantially higher. The snow depth variability (5&ndash;95 percentiles) within different land cover types increased from 17 cm to 42 cm in peatlands and from 33 cm to 49 cm in the coniferous forest from the beginning of the snow accumulation to the melt period. Both the median snow depth and its variability were found to increase with canopy density; this increase was greatest in the conifer forest area, followed by mixed forest, transitional woodland/shrub, and open peatlands. Using the high spatial resolution data, we found a systematic increase (2&ndash;20 cm), then a decline of snow depth near the canopy with increasing distance (from 1 m to 2.5 m) of the peak value through the snow season. This study highlights the potential of the UAS-SfM in high-resolution monitoring of snow depth in multiple land cover types and snow-vegetation interactions in subarctic and remote areas where field data is not available.

<strong class="journal-contentHeaderColor">Abstract.</strong> Detailed information on seasonal snow cover and depth is essential to the understanding of snow processes, operational forecasting, and as input for hydrological models. Recent advances in unmanned aircraft systems (UASs) and structure from motion (SfM) techniques have enabled low-cost monitoring of spatial snow depth distribution in resolutions up to a few centimeters. Here, we study the spatiotemporal variability of snow depth and interactions between snow and vegetation in different subarctic landscapes consisting of a mosaic of conifer forest, mixed forest, transitional woodland/shrub, and peatland areas. To determine the spatiotemporal variability of snow depth, we used high-resolution (50 cm) snow depth maps generated from repeated UAS-SfM surveys in the winter of 2018/2019 and a snow-free bare ground survey after snowmelt. Due to poor sub-canopy penetration with the UAS-SfM method, tree masks were utilized to remove canopy areas and the area (36 cm) immediately next to the canopy before analysis. Snow depth maps were compared to the in-situ snow course and a single-point continuous ultrasonic snow depth measurement. Based on the results, the difference between the UAS-SfM survey median snow depth and single-point measurement increased for all land cover types during the snow season, from +5 cm at the beginning of the accumulation to -16 cm in coniferous forests and -32 cm in peatland during the melt period. This highlights the poor representation of point measurements even on the sub-catchment scale. The high-resolution snow depth maps agreed well with the snow course measurement, but the spatial extent and resolution of maps were substantially higher. The snow depth variability (5&ndash;95 percentiles) within different land cover types increased from 17 cm to 42 cm in peatlands and from 33 cm to 49 cm in the coniferous forest from the beginning of the snow accumulation to the melt period. Both the median snow depth and its variability were found to increase with canopy density; this increase was greatest in the conifer forest area, followed by mixed forest, transitional woodland/shrub, and open peatlands. Using the high spatial resolution data, we found a systematic increase (2&ndash;20 cm), then a decline of snow depth near the canopy with increasing distance (from 1 m to 2.5 m) of the peak value through the snow season. This study highlights the potential of the UAS-SfM in high-resolution monitoring of snow depth in multiple land cover types and snow-vegetation interactions in subarctic and remote areas where field data is not available.

<strong class="journal-contentHeaderColor">Abstract.</strong> Detailed information on seasonal snow cover and depth is essential to the understanding of snow processes, operational forecasting, and as input for hydrological models. Recent advances in unmanned aircraft systems (UASs) and structure from motion (SfM) techniques have enabled low-cost monitoring of spatial snow depth distribution in resolutions up to a few centimeters. Here, we study the spatiotemporal variability of snow depth and interactions between snow and vegetation in different subarctic landscapes consisting of a mosaic of conifer forest, mixed forest, transitional woodland/shrub, and peatland areas. To determine the spatiotemporal variability of snow depth, we used high-resolution (50 cm) snow depth maps generated from repeated UAS-SfM surveys in the winter of 2018/2019 and a snow-free bare ground survey after snowmelt. Due to poor sub-canopy penetration with the UAS-SfM method, tree masks were utilized to remove canopy areas and the area (36 cm) immediately next to the canopy before analysis. Snow depth maps were compared to the in-situ snow course and a single-point continuous ultrasonic snow depth measurement. Based on the results, the difference between the UAS-SfM survey median snow depth and single-point measurement increased for all land cover types during the snow season, from +5 cm at the beginning of the accumulation to -16 cm in coniferous forests and -32 cm in peatland during the melt period. This highlights the poor representation of point measurements even on the sub-catchment scale. The high-resolution snow depth maps agreed well with the snow course measurement, but the spatial extent and resolution of maps were substantially higher. The snow depth variability (5&ndash;95 percentiles) within different land cover types increased from 17 cm to 42 cm in peatlands and from 33 cm to 49 cm in the coniferous forest from the beginning of the snow accumulation to the melt period. Both the median snow depth and its variability were found to increase with canopy density; this increase was greatest in the conifer forest area, followed by mixed forest, transitional woodland/shrub, and open peatlands. Using the high spatial resolution data, we found a systematic increase (2&ndash;20 cm), then a decline of snow depth near the canopy with increasing distance (from 1 m to 2.5 m) of the peak value through the snow season. This study highlights the potential of the UAS-SfM in high-resolution monitoring of snow depth in multiple land cover types and snow-vegetation interactions in subarctic and remote areas where field data is not available.

Detailed information on seasonal snow cover and depth is essential to the understanding of snow processes, operational forecasting, and as input for hydrological models. Recent advances in unmanned aircraft systems (UASs) and structure from motion (SfM) techniques have enabled low-cost monitoring of spatial snow depth distribution in resolutions up to a few centimeters. Here, we study the spatiotemporal variability of snow depth and interactions between snow and vegetation in different subarctic landscapes consisting of a mosaic of conifer forest, mixed forest, transitional woodland/shrub, and peatland areas. To determine the spatiotemporal variability of snow depth, we used high-resolution (50 cm) snow depth maps generated from repeated UAS-SfM surveys in the winter of 2018/2019 and a snow-free bare ground survey after snowmelt. Due to poor sub-canopy penetration with the UAS-SfM method, tree masks were utilized to remove canopy areas and the area (36 cm) immediately next to the canopy before analysis. Snow depth maps were compared to the in-situ snow course and a single-point continuous ultrasonic snow depth measurement. Based on the results, the difference between the UAS-SfM survey median snow depth and single-point measurement increased for all land cover types during the snow season, from +5 cm at the beginning of the accumulation to -16 cm in coniferous forests and -32 cm in peatland during the melt period. This highlights the poor representation of point measurements even on the sub-catchment scale. The high-resolution snow depth maps agreed well with the snow course measurement, but the spatial extent and resolution of maps were substantially higher. The snow depth variability (5–95 percentiles) within different land cover types increased from 17 cm to 42 cm in peatlands and from 33 cm to 49 cm in the coniferous forest from the beginning of the snow accumulation to the melt period. Both the median snow depth and its variability were found to increase with canopy density; this increase was greatest in the conifer forest area, followed by mixed forest, transitional woodland/shrub, and open peatlands. Using the high spatial resolution data, we found a systematic increase (2–20 cm), then a decline of snow depth near the canopy with increasing distance (from 1 m to 2.5 m) of the peak value through the snow season. This study highlights the potential of the UAS-SfM in high-resolution monitoring of snow depth in multiple land cover types and snow-vegetation interactions in subarctic and remote areas where field data is not available.

Snow depth (SD) is an essential climate variable widely used for flood forecasting, water quantity assessments, road and building safety assessment, habitat assessment, and climate studies. Currently, SD is systematically monitored using manual ruler measurements or dedicated instrumentation with a limited spatial footprint (~1m2). Here, an approach for automated SD estimation using images of narrow stakes remotely acquired by a 3Mpixel trail camera is described. The approach relies on the automated application of image processing and machine learning algorithms packaged within a single freely available application deployed on a personal computer or cloud computing environment. The application requires minimal user input to define an initial template image and provides two independent estimates of SD at each stake that can be used to produce an estimate of total uncertainty and sources of error. The system is compared to both manual ruler and ultrasonic instrument SD estimates at an open site and in a deciduous forest. Initial results over a melt period indicate the automated method agreed to within 1cm (RMSE) of manual ruler estimates and to within 3.6cm of ultrasonic estimates; with the latter comparison including spatial variability between measurement locations. The automated system should be considered for further deployment and evaluation over a range of surface and climate conditions.

Carbon dioxide (CO2) emissions from Southeast Asia peatlands are contributing substantially to global anthropogenic emissions to the atmosphere. Peatland emissions associated with land-use change, and fires are closely related to changes in the water table level. Remote sensing is a powerful tool that is potentially useful for estimating peat CO2 emissions over large spatial and temporal scales. We related ground measurements of total soil respiration and water table depth collected over 19 months in an Indonesian peatland to remotely sensed gravity recovery and climate experiment (GRACE) terrestrial water storage anomoly (TWSA) data. GRACE TWSA can be used to predict changes in water storage on land. We combined ground observations from undrained forest and drained smallholder oil palm plantations on peat in Central Kalimantan to produce a representation of the peatland landscape in one 0.5° × 0.5° GRACE grid cell. In both ecosystem types, total soil respiration increased with increasing water table depth. Across the landscape grid, monthly changes in water table depth were significantly related to fluctuations in GRACE TWSA. GRACE TWSA explained 76% of variation in water table depth and 75% of variation in total soil respiration measured on the ground. By facilitating regular sampling across broad spatial scales that captures essential variation in a major driver of soil respiration and peat fires, our approach could improve information available to decision makers to monitor changes in water table depth and peat CO2 emissions. This would enable measures better targeted in space and time to more effectively mitigate CO2 emissions from tropical peat drainage and fires. Testing over larger regions is needed to operationalize this exploratory approach.

Snowline change and snow cover distribution patterns are still poorly understood in steep alpine basins of the Qilian Mountainous region because fast changes in snow cover cannot be observed by current sensing methods due to their short time scale. To address this issue of daily snowline and snow cover observations, a ground-based EOS 7D camera and four infrared digital hunting video cameras (LTL5210A) were installed around the Hulugou river basin (HRB) in the Qilian Mountains along northeastern margin of the Tibetan Plateau (3815'54"N, 9952'53"E) in September 2011. Pictures taken with the EOS 7D camera were georeferenced and the data from four LIL5210A cameras and snow depth sensors were used to assist snow cover estimation. The results showed that the time-lapse photography can be very useful and precise for monitoring snowline and snow cover in mountainous regions. The snowline and snow cover evolution at this basin can be precisely captured at daily scale. In HRB snow cover is mainly established after October, and the maximum snow cover appeared during February and March. The consistent rise of the snowline and decrease in snow cover appeared after middle part of March. This melt process is strongly associated with air temperature increase.

Growing conditions for street tree roots are generally harsh with restricted space and soils compacted from streetscape infrastructure.Allocasuarina littoralis, Corymbia maculata, Cupressus sempervirensvar.stricta, Eucalyptus polyanthemos, Lophostemon confertus, Olea europaea, Quercus palustris, andWaterhousea floribundawere grown in compacted and uncompacted soils for 20 months in experimental blocks. The bulk density and penetrative resistance of the soils and height, canopy spread, trunk diameter, leaf area, and chlorophyll fluorescence were measured regularly. Root and shoot biomass were determined after harvesting. Since the bulk density of compacted compared to uncompacted soil was root growth limiting, it was hypothesised that species would have reduced growth in compacted soils. However,C. maculataandE. polyanthemosgrew better,C. sempervirens, Q. palustris, andW. floribundagrew well, andA. littoralis, L. confertus, andO. europaeawere the worst performing in compacted soil.E. polyanthemos, L. confertus, andQ. palustrishad higher canopy:root ratios in compacted soil.Q. palustrishad greater mass below ground than above, which has implications for its use in confined sites. In a field study,C. maculata, E. polyanthemos, L. confertus, O. europaea, andQ. palustrisgrowing as street trees were surveyed to determine their rates of establishment and growth under urban conditions. In addition to the soil and tree parameters mentioned above, a Visual Tree Assessment (VTA) was undertaken.E. polyanthemoshad the largest trunk diameter, height and canopy spread, indicating its potential for rapid establishment in streets. It was the only species with a larger mean leaf area in compacted soil.E. polyanthemosandO. europaeawere the only species classed as healthy from chlorophyll fluorescence but there was no significant difference in fluorescence between compacted and uncompacted soils. VTA showed thatC. maculataandO. europaeaperformed best and thatE. polyanthemos, L. confertus, andQ. palustrishad reduced but acceptable growth in compacted soil. Soils ranged from non-saline to moderately saline and were slightly to strongly acidic. All soils were compacted to some degree and penetrative resistance was at root limiting levels. The results suggest that careful species selection and soil amelioration for species prone to the effects of compaction would facilitate street tree establishment.

Accurate temperature measurements are critical in manufacturing, affecting both product quality and energy consumption. At elevated temperatures, non-contact thermometers are often the only option. However, such instruments require prior knowledge of the surface emissivity, which is often unknown or difficult to determine, leading to large errors. Here we present a novel imaging luminescence thermometer based on the intensity ratio technique using magnesium fluorogermanate phosphor, with the potential to overcome this limitation. We describe measurements performed on a number of engineering alloys undergoing heat treatment at temperatures of up to 750 °C and compare these measurements against a traditional contact thermocouple and thermal imager system. Agreement between the luminescence and embedded thermocouple temperatures was found to be better than 45 °C at all temperatures. However, the thermal imager measurement on the bare metal samples, with the instrument emissivity set to 1.0, showed differences of up to 500 °C at 750 °C, a factor of 10 larger. In an effort to improve the thermal imager accuracy, its instrument emissivity was adjusted until its temperature agreed with that of the thermocouple. When measuring on the bare metal, the effective emissivity was strongly sample dependent, with mean values ranging from 0.205 to 0.784. Since the phosphor derived temperatures exhibited substantially smaller errors compared to the thermal imager, it is suggested that this method can be used to compliment the thermal imaging technique, by providing a robust mechanism for adjustment of the instrument emissivity until agreement between the thermal imager and phosphor thermometer is obtained.

The Global Fire Emissions Database (GFED3) and the FAOSTAT Emissions database, containing estimates of greenhouse gas (GHG) emissions from biomass burning and peat fires, are compared. The two datasets formed the basis for several analyses in the fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR5), and thus represent a critical source of information for emissions inventories at national, regional and global level. The two databases differ in their level of computational complexity in estimating emissions. While both use the same burned area information from remote sensing, estimates of available biomass are computed in GFED3 at tier 3 using a complex dynamic vegetation model, while they are computed in FAOSTAT using default, tier 1 parameters from the Intergovernmental Panel on Climate Change (IPCC). Over the analysis period 1997–2011, the two methods were found to produce very similar global GHG emissions estimates for each of the five GFED aggregated biomass fire classes: i) Savanna; ii) Woodland; iii) Forest; iv) Deforestation; v) Peatlands; with total emissions ranging 6–8 Gt CO2eq yr-1. The main differences between the two datasets were found with respect to peat fires, with FAOSTAT showing a lower 1997–1998 peak in emissions compared with GFED3, within an otherwise good agreement for the rest of the study period, when limited to the three tropical countries covered by GFED. Conversely, FAOSTAT global emissions from peat fires, including both boreal and tropical regions, were several times larger than those currently estimated by GFED3. Results show that FAOSTAT activity data and emission estimates for biomass fires offer a robust alternative to the more sophisticated GFED data, representing a valuable resource for national GHG inventory experts, especially in countries where technical and institutional constraints may limit access, generation and maintenance of more complex methodologies and data.