NORMALISASI DATA UNTUK EFISIENSI K-MEANS PADA PENGELOMPOKAN WILAYAH BERPOTENSI KEBAKARAN HUTAN DAN LAHAN BERDASARKAN SEBARAN TITIK PANAS

Kawasan indonesia merupakan bagian dari daerah tropis yang memiliki potensi kebakaran sangat tinggi terlebih pada musim kemarau, sehingga perlunya sebuah langkah kongkrit untuk dilakukan mitigasi supaya potensi-potensi kebakaran hutan itu menjadi terminimalisir. Untuk melakukan itu dibutuhkan suatu metode teknologi yang lebih mumpuni dan terbaru untuk memetakan wilayah-wilayah yang mempunyai potensi besar terjadinya kebakaran hutan. Sistem pencitraan dan Informasi dari sistem satelit (MODIS) adalah salah satu informasi tentang kondisi permukaan bumi, yaitu parameter Latitude, Longitude, Brightness, FRP (Fire Radiative Power), dan Confidence dapat dijadikan dasar pengelompokan suatu wilayah memiliki potensi kebakaran atau tidak. K-Means adalah salah satu metode dalam machine learning yang bisa digunakan sebagai salah satu metode dalam pengelompokan wilayah-wilayah tersebut. Akurasi dalam menguji hasil pengelompokan K-Means dapat diuji dengan metode Davies Bouldin Index (DBI) dan Silhouette Coefficient.

Estimation of Byram’s Fire Intensity and Rate of Spread from Spaceborne Remote Sensing Data in a Savanna Landscape

Fire behavior is well described by a fire’s direction, rate of spread, and its energy release rate. Fire intensity as defined by Byram (1959) is the most commonly used term describing fire behavior in the wildfire community. It is, however, difficult to observe from space. Here, we assess fire spread and fire radiative power using infrared sensors with different spatial, spectral and temporal resolutions. The sensors used offer either high spatial resolution (Sentinel-2) for fire detection, but a low temporal resolution, moderate spatial resolution and daily observations (VIIRS), and high temporal resolution with low spatial resolution and fire radiative power retrievals (Meteosat SEVIRI). We extracted fire fronts from Sentinel-2 (using the shortwave infrared bands) and use the available fire products for S-NPP VIIRS and Meteosat SEVIRI. Rate of spread was analyzed by measuring the displacement of fire fronts between the mid-morning Sentinel-2 overpasses and the early afternoon VIIRS overpasses. We retrieved FRP from 15-min Meteosat SEVIRI observations and estimated total fire radiative energy release over the observed fire fronts. This was then converted to total fuel consumption, and, by making use of Sentinel-2-derived burned area, to fuel consumption per unit area. Using rate of spread and fuel consumption per unit area, Byram’s fire intensity could be derived. We tested this approach on a small number of fires in a frequently burning West African savanna landscape. Comparison to field experiments in the area showed similar numbers between field observations and remote-sensing-derived estimates. To the authors’ knowledge, this is the first direct estimate of Byram’s fire intensity from spaceborne remote sensing data. Shortcomings of the presented approach, foundations of an error budget, and potential further development, also considering upcoming sensor systems, are discussed.

Forecasting the regional fire radiative power for regularly ignited vegetation fires

This paper presents a phenomenological framework for forecasting the area-integrated fire radiative power from wildfires. In the method, a region of interest is covered with a regular grid, which cells are uniquely and independently parameterized with regard to the fire intensity according to (i) the fire incidence history, (ii) the retrospective meteorological information, and (iii) remotely-sensed high temporal resolution fire radiative power taken together with (iv) consistent cloud mask data. The parameterization is realized by fitting the predetermined functions for diurnal and annual profiles of fire radiative power to the remote-sensing observations. After the parametrization, the input for the fire radiative power forecast is the meteorological data alone, i.e., the weather forecast. The method is tested retrospectively for south-central African savannah areas with grid cell size of 1.5° × 1.5°. The input data included ECMWF ERA5 meteorological reanalysis and SEVIRI/MSG Fire Radiative Power and Cloud Mask. It has been found that in the areas with large numbers of wildfires regularly ignited on a daily basis during dry seasons from year to year, the temporal fire radiative power evolution is quite predictable, whereas the areas with irregular fire behaviour predictability was low. The predictive power of the method is demonstrated by comparing the predicted fire radiative power patterns and fire radiative energy values against the corresponding remote-sensing observations. The current method showed good skills for the considered African regions and was useful in understanding the challenges in predicting the wildfires in a more general case.

A Bi-Spectral Microbolometer Sensor for Wildfire Measurement

This study describes the development of a prototype bi-spectral microbolometer sensor system designed explicitly for radiometric measurement and characterization of wildfire mid- and long-wave infrared radiances. The system is tested experimentally over moderate-scale experimental burns coincident with FLIR reference imagery. Statistical comparison of the fire radiative power (FRP; W) retrievals suggest that this novel system is highly reliable for use in collecting radiometric measurements of biomass burning. As such, this study provides clear experimental evidence that mid-wave infrared microbolometers are capable of collecting FRP measurements. Furthermore, given the low resource nature of this detector type, it presents a suitable option for monitoring wildfire behaviour from low resource platforms such as unmanned aerial vehicles (UAVs) or nanosats.

Geocorrection of Airborne Mid-Wave Infrared Imagery for Mapping Wildfires without GPS or IMU

The increase in annual wildfires in many areas of the world has triggered international efforts to deploy sensors on airborne and space platforms to map these events and understand their behaviour. During the summer of 2017, an airborne flight campaign acquired mid-wave infrared imagery over active wildfires in Northern Ontario, Canada. However, it suffered multiple position-based equipment issues, thus requiring a non-standard geocorrection methodology. This study presents the approach, which utilizes a two-step semi-automatic geocorrection process that outputs image mosaics from airborne infrared video input. The first step extracts individual video frames that are combined into orthoimages using an automatic image registration method. The second step involves the georeferencing of the imagery using pseudo-ground control points to a fixed coordinate systems. The output geocorrected datasets in units of radiance can then be used to derive fire products such as fire radiative power density (FRPD). Prior to the georeferencing process, the Root Mean Square Error (RMSE) associated with the imagery was greater than 200 m. After the georeferencing process was applied, an RMSE below 30 m was reported, and the computed FRPD estimations are within expected values across the literature. As such, this alternative geocorrection methodology successfully salvages an otherwise unusable dataset and can be adapted by other researchers that do not have access to accurate positional information for airborne infrared flight campaigns over wildfires.

Current Trend of Carbon Emissions from Wildfires in Siberia

Smoke from wildfires in Siberia often affects air quality over vast territories of the Northern hemisphere during the summer. Increasing fire emissions also affect regional and global carbon balance. To estimate annual carbon emissions from wildfires in Siberia from 2002–2020, we categorized levels of fire intensity for individual active fire pixels based on fire radiative power data from the standard MODIS product (MOD14/MYD14). For the last two decades, estimated annual direct carbon emissions from wildfires varied greatly, ranging from 20–220 Tg C per year. Sporadic maxima were observed in 2003 (>150 Tg C/year), in 2012 (>220 Tg C/year), in 2019 (~180 Tg C/year). However, the 2020 fire season was extraordinary in terms of fire emissions (~350 Tg C/year). The estimated average annual level of fire emissions was 80 ± 20 Tg C/year when extreme years were excluded from the analysis. For the next decade the average level of fire emissions might increase to 250 ± 30 Tg C/year for extreme fire seasons, and to 110 ± 20 Tg C/year for moderate fire seasons. However, under the extreme IPCC RPC 8.5 scenario for Siberia, wildfire emissions might increase to 1200–1500 Tg C/year by 2050 if there were no significant changes in patterns of vegetation distribution and fuel loadings.

Real-Time Detection of Daytime and Night-Time Fire Hotspots from Geostationary Satellites

This paper introduces an enhanced version of the Biogeographical Region and Individual Geostationary HHMMSS Threshold (BRIGHT) algorithm. The algorithm runs in real-time and operates over 24 h to include both daytime and night-time detections. The algorithm was executed and tested on 12 months of Himawari-8 data from 1 April 2019 to 31 March 2020, for every valid 10-min observation. The resulting hotspots were compared to those from the Visible Infrared Imaging Radiometer Suite (VIIRS) and the Moderate Resolution Imaging Spectroradiometer (MODIS). The modified BRIGHT hotspots matched with fire detections in VIIRS 96% and MODIS 95% of the time. The number of VIIRS and MODIS hotspots with matches in the coincident modified BRIGHT dataset was lower (at 33% and 46%, respectively). This paper demonstrates a clear link between the number of VIIRS and MODIS hotspots with matches and the minimum fire radiative power considered.

Top-Down Estimation of Particulate Matter Emissions from Extreme Tropical Peatland Fires Using Geostationary Satellite Fire Radiative Power Observations

Extreme fires in the peatlands of South East (SE) Asia are arguably the world’s greatest biomass burning events, resulting in some of the worst ambient air pollution ever recorded (PM10 > 3000 µg·m−3). The worst of these fires coincide with El Niño related droughts, and include huge areas of smouldering combustion that can persist for months. However, areas of flaming surface vegetation combustion atop peat are also seen, and we show that the largest of these latter fires appear to be the most radiant and intensely smoke-emitting areas of combustion present in such extreme fire episodes. Fire emissions inventories and early warning of the air quality impacts of landscape fire are increasingly based on the fire radiative power (FRP) approach to fire emissions estimation, including for these SE Asia peatland fires. “Top-down” methods estimate total particulate matter emissions directly from FRP observations using so-called “smoke emission coefficients” [Ce; g·MJ−1], but currently no discrimination is made between fire types during such calculations. We show that for a subset of some of the most thermally radiant peatland fires seen during the 2015 El Niño, the most appropriate Ce is around a factor of three lower than currently assumed (~16.8 ± 1.6 g·MJ−1 vs. 52.4 g·MJ−1). Analysis indicates that this difference stems from these highly radiant fires containing areas of substantial flaming combustion, which changes the amount of particulate matter emitted per unit of observable fire radiative heat release in comparison to more smouldering dominated events. We also show that even a single one of these most radiant fires is responsible for almost 10% of the overall particulate matter released during the 2015 fire event, highlighting the importance of this fire type to overall emission totals. Discriminating these different fires types in ways demonstrated herein should thus ultimately improve the accuracy of SE Asian fire emissions estimates derived using the FRP approach, and the air quality modelling which they support.

Estimation of Direct Fire Emissions from Forests Burning in Siberia

Using a database on wildfires recorded by remote sensing for 1996–2020, we assessed the seasonal variation of direct carbon emissions from the burning in Siberian forests. We have implemented an approach that takes into account the combustion parameters and the changing intensity of the fire (in terms of Fire Radiative Power (FRP)), which affects the accuracy of the emission estimate. For the last two decades, the range of direct carbon emissions from wildfires was 20–250 Тg С per year. Sporadic maxima were fixed in 2003 (>150 Тg С/year), in 2012 (>220 Тg С/year), and in 2019 (>190 Тg С/year). Preliminary estimation of emissions for 2020 (on 30th of September) was ~180 Tg С/year. Fires in the larch forests of the flat-mountainous taiga region (Central Siberia) made the greatest contribution (>50%) to the budget of direct fire emission, affecting the quality of the atmosphere in vast territories during the summer period. According to the temperature rising and forest burning trend in Siberia, the fire emissions of carbon may double (220 Тg С/year) or even increase by an order of magnitude (>2000 Тg С/year) at the end of the 21st century, which was evaluated depending on IPCC scenario.