As one of the core instruments of the Fengyun-3 (FY-3) satellites, the updated Medium Resolution Spectral Imager (MERSI) has become one of the most advanced remote sensing instruments based on wide-swath imaging. The FY-3D satellite was launched in November 2017 with 10 sets of remote sensing instruments, including the Medium Resolution Spectral Imager II (MERSI-II). MERSI-II integrates the functions of the two original imaging instruments (MERSI-I and VIRR) of FY-3B and FY-3C, with a total of 25 channels, including visible light, near infrared, medium infrared and far infrared (as in Table 1). The infrared imaging, detection sensitivity and calibration accuracy of MERSI-II are improved greatly. It is the first imaging instrument that can access the 250 m resolution infrared split-window area globally and capture seamless 250 m resolution true-color global images on a daily basis. MERSI-II also enables the high-quality retrieval of atmospheric, land and marine parameters such as clouds, aerosols, vapor, land surface features and ocean color, supporting global support for environment and climate issues.

There are two middle-infrared bands (3.8 and 4.05 µm), and both are sensitive to strong heat signals. Their differences lie in their performance under different temperature and radiation conditions. The 3.8 µm band is closer to the wavelength of solar radiation and has better reflection under solar radiation. As a comparison, the 4.05 µm band more easily misses weak fires. Therefore, current FY-3D fire products are mainly produced based on the 3.8 µm band for better fire identification. According to the calculation, the emissivity of forest and grassland fires in the mid-infrared band can be hundreds of times higher than that of the surface at normal temperature, making the radiance and brightness temperature of the fire spot significantly higher than surrounding pixels. For rapid monitoring of global wildfires, it is necessary to develop an algorithm for the automatic identification of fire spots.

MERSI-II fire monitoring products from the FY-3D satellite can provide fire spot location, sub-pixel fire spot area, temperature and fire spot intensity in inland areas around the world and generate global fire spot pixel information (including day and night) in HDF files. FY-3D fire products are produced following a projection with equal latitude and longitude (0.01∘). Fire spot intensity is classified according to the sub-pixel fire spot area and temperature, with an overall accuracy above 85 %. Based on daily monitoring products, the SMART (Satellite Monitoring Analyzing and Remote sensing Tools) system can generate the images of global monthly fire spot distribution, with a resolution of 0.25∘.

The algorithm for fire spot identification depends on the sensitivity of mid-infrared channels to high-temperature heat sources. The radiance and brightness temperature of the pixels in the mid-infrared channels with sub-pixel fire spots are higher than those of the surrounding non-fire pixels and those of the pixels in the far-infrared channels. Therefore, the pixels with fire spots can be identified by setting an appropriate threshold, and the estimation of background temperature is the key to high detection accuracy and sensitivity.

Sub-pixel fire spot estimation relies on the brightness temperature in mid-infrared channels, and the far-infrared channels are employed when the mid-infrared channels have saturated brightness temperature. In the single-channel estimation formula, the temperature of the open-flame spot is set to 750 K.

Fire spot intensity, namely fire radiation power (FRP), is obtained by substituting the area and temperature of sub-pixel fire spots into the Stefan–Boltzmann formula of full-band blackbody radiation. 1J∗=εσT4 The radiant emittance J∗ has dimensions of energy flux, and the SI units of measure are joules per second per square meter. The SI unit for absolute temperature T is the kelvin. ε is the emissivity for the grey body; if it is a blackbody, ε=1.0 is the Stefan–Boltzmann constant.

FRP is divided into 10 levels, indicating different ranges of radiation intensity and the fire behavior at fire spot pixels. Fire spots are classified into four groups with regard to credibility, namely the real fire spots, possible fire spots, fire spots affected by the cloud and noisy fire spots (disturbed by clouds and noise).

FY-3D MERSI-II daily global fire monitoring products are illustrated in Fig. 1. The major processing of daily fire spot products is the generation of 5 min fire spot lists, which include such information as the observation time of fire spot pixels, latitude and longitude, the sub-pixel fire spot area and temperature, and FRP. Next, all the 5 min fire spot information for each day is merged into the daily global fire information list.

FY-3D MERSI-II monthly global fire monitoring products consist of the information list of global fire spot pixels and the density map of global fire spots. The information list of monthly global fire spots covers all global fire spot pixels in the particular month. Concerning the multi-time monitoring information of the same pixel, the maximum fire spot area is taken as the current-month fire spot information for the pixel. Figure 2 is an illustration of the density map of global fire spots based on FY-3D MERSI-II, in which different colors indicate the number of fire spot pixels on a 0.25∘ × 0.25∘ spatial grid. Compared with daily FY-3D fire products, monthly FY-3D fire products were advantageous in revealing the global patterns of fire spots. As shown in Fig. 2, the global fire spots were mainly distributed in southern Africa, central South America, southern North America, north-central Asia and northern Australia in June 2019.

Channel 20 of FY-3D MERSI-II is mid-infrared, with a wavelength of 3.55–3.95 µm, while Channels 24 and 25 are far-infrared, with a wavelength of 10.3–11.3 and 11.5–12.5 µm, respectively. According to Wien's displacement law, 2λ⋅T=b, where λ is the peaks at the wavelength; T is the absolute temperature; and b is a constant of proportionality called Wien's displacement constant, equal to about 2898 µm K. Blackbody temperature T is inversely proportional to peak radiation wavelength λmax⁡, as the higher temperature can lead to the smaller peak radiation wavelength. The peak radiation wavelength of the surface at normal temperature (about 300 K) is close to that of Channels 24 and 25; the combustion temperature of forest fires is generally 500–1200 K, and the peak wavelength of thermal radiation is close to that of Channel 20. When a fire spot appears in the observed pixel, the radiance increment in Channel 20 caused by the high temperature in the small sub-region of the pixel, where the fire spot is located (since the pixel resolution of the scanning radiometer is 1.1 km, usually, in such a large area, all open-flame areas will not appear simultaneously), is much higher than surrounding pixels without an open flame and also greater than that in Channels 24 and 25. In this case, the weighted averages of the radiance increase and brightness temperature increase of each channel differ notably in this pixel, based on which the fire information can be extracted and analyzed.

As indicated by Fig. 4a, when the fire spot temperature grows, the brightness temperature of Channel 20 (CH20) pixels increases rapidly. Even if the fire spot only accounts for 0.1 % the pixel area, the brightness temperature increment can reach 10 K (44 K) when the fire spot is 500 K (900 K). Although the brightness temperature increase of CH24 also rises with the higher fire spot temperature, it is far lower than that of CH20. Figure 4b illustrates that as the fire spot area becomes larger, the brightness temperature of CH20 mixed pixels grows rapidly. It reaches 12 K when the fire spot is 900 K, even if the fire spot only accounts for 0.01 % of the pixel area. Similarly, the brightness temperature increment of CH24 grows at a much lower rate than that of CH20.

FRP can be calculated using the Stefan–Boltzmann formula (Matson and Schneider, 1984) through the estimation of the sub-pixel fire spot area and temperature.

Wildfires are characterized by random and rapid changes, so it is difficult to verify the product accuracy of GFR (global fire) according to actual ground information. In this paper, the accuracy of FY-3 fire products is tested through visual interpretation and cross-verification of other products. Specifically, due to the extremely large size of GFR datasets, we set the different strategies for accuracy assessment. For visual interpretation, several 5 min data segments with regional representation were selected for verification using manually identified fire spots. For cross-verification with other fire products, global fire spot data throughout 2019 were employed.

The error was defined as the distance from the positions (longitude and latitude) of automatically identified fire spot pixels to corresponding manually identified ones. When the difference in latitude and longitude was less than or equal to 0.02∘, the automatically identified pixel was regarded as a successful identification. (lat1-lat2)2+long1-long22≤0.02∘, where lat1 and lat2 are the latitude of PGS (product generation system) fire spot pixels and manually identified pixels (reference pixels) and long1 and long2 are the longitude of PGS fire spot pixels and manually identified pixels (reference pixels), respectively.

In addition to the visual-check-based accuracy assessment at the global scale, we also employed a set of field-collected reference data to verify the suitability of FY-3D in China, which is further explained in the following sections.

In this research, 5 min segments of FY-3D fire products in different continents, including Africa, South America, the Indochinese Peninsula, Siberia and Australia, were collected at 12:15 (UTC) on 13 June 2018, 17:05 (UTC) on 21 August 2019, 06:15 (UTC) on 13 March 2019, 03:40 (UTC) on 13 November 2019 and 17:40 (UTC) on 29 May 2018, respectively, for visual interpretation. The specific observation positions are shown in Fig. 5 with five corresponding fire detection pictures of FY-3D.

These regions were selected for evaluating the global reliability of FY-3D fire products for the following reasons. Firstly, Africa, South America, the Indochinese Peninsula, Siberia and Australia are the regions with the most frequent fire events across the globe. Secondly, there is rich vegetation in these regions, which provides the foundation for stable combustion across a year. Thirdly, these regions cover large areas with generally unified underlying surfaces. Fourthly, these areas are of regional representation: Siberia represents typical regions with frequent forest fires in the Northern Hemisphere. Africa represents typical tropical grasslands and forests in the Equator regions. South America represents virgin tropical rain forests.

Figure 5 presents the spatial distribution of GFR spots and manually identified fire pixels in the 5 min segment of the above regions. According to Fig. 5b, most fire spots in FY-3D products and manually extracted fire spots in South America were in the same positions. In Fig. 5c, most FY-3D and manually extracted fire spots in Africa coincided or were in a close position. In Fig. 5d, despite a few mismatched fire spots, the positions of FY-3D and manually extracted fire spots in the Indochinese Peninsula were consistent. Figure 5e and f also show that most fire spots are matched in Russia and Australia. Table 4 shows accuracy of GFR spots in the five typical regions. The accuracy of automatically identified fire spot in all regions was generally consistent and all exceeded 90 %. Since these selected regions represented distinct vegetation types and are located in different hemispheres, the verification of FY-3D fire products based on 0.24 SMART proved its stability and reliable high-accuracy at the global scale.

It is worth mentioning that the visual-check-based accuracy assessment mainly considered the commission error, while omission error cannot be effectively revealed for the following reason. The omitted fires were mainly caused by the requirement of a minimum burning area. Since the spatial resolution of FY-3D and MODIS active fire products is 1 km, small fires (less than 100 m2) could not be captured by sensors and recognized through visual check. Meanwhile, thermal abnormalities were seen at the edge of cloud and water bodies, which could be recognized through visual check. In this case, the visual-check-based accuracy assessment mainly considered the commission errors.

The cross-verification between FY-3D fire products and the mainstream MODIS fire products, MYD14A1 V6 (https://firms.modaps.eosdis.nasa.gov/map/, last access: 10 January 2021) with a daily temporal resolution and 1 km spatial resolution, was conducted using all the 2019 datasets. The datasets with observation times of less than 1 h were selected; the underlying surfaces were visually checked to remove areas covered by non-vegetation such as water, ice and snow, and bare land. According to the criterion that the distance matching between the two fire spot pixels was less than 0.03∘, cross-verification was conducted with different months, underlying surfaces, regions and fire intensities. In 2019, there were 2 237 714 fire spot pixels in MODIS fire products, 1 866 920 of which were matched with FY-3D fire products, with an overall consistence of 84.4 % (as shown in Fig. 6). As shown in Fig. 6, global fire spots were mainly distributed in America, south-central Africa, East Asia and Southeast Asia, Australia, and parts of Europe, and there were notable spatiotemporal variations in identified fire spots. Specifically, given the overall data volume and spatial distribution, the total number of fire spot pixels from MODIS fire products was larger than that of FY-3D products. For individual regions, the more fire spots, the higher consistence between FY and MODIS fire products. Africa is the region with the most fire spots across the globe. From May to October, a majority of fire spots was located in southern Africa, while a majority of fire spots from November to the next April was located in the middle and western coastal regions of Africa. The consistence between MODIS and FY-3D products was higher than in other regions. The distribution of fire spots in South America also presented seasonal characteristics. From July to October, fire spots were mainly concentrated in the middle parts of South America. For other seasons, fire spots in South America were mainly concentrated in the north and other parts. The consistence between MODIS and FY-3D fire products also demonstrated seasonal differences, with a high consistence from August to November and a relatively low consistence in other seasons. For Eurasia, there were notable seasonal variations in spatial patterns of fire spots. During March to August, there were relatively many fire spots and the consistence between MODIS and FY-3D fire products was relatively high in this region.

In addition to the overall consistence between MODIS and FY-3D fire products, we also conducted cross-verification between the two global fire products in different months, underlying surfaces, regions and fire intensities as follows.

In addition to visual check and consistence check, we also referred to a large-scale field experiment to comprehensively assess the suitability of FY-3D fire products in China. The State Grid Corporation of China and China Meteorological Administration jointly conducted a fire detection experiment throughout 2020 in five provinces in China: Guangdong, Guangxi, Yunnan, Guizhou and Hainan. This experiment was conducted in the following steps. A large number of drones were employed to check the occurrence of fires. According to the local passing time of FY-3D, these drones reported the coordinates of actual fires for verifying the accuracy of FY-3D-identified fires. The temporal difference between the passing time of FY-3D and reported time was controlled to within 1 h. In this way, both omitted and misidentified fires could be effectively recognized (as shown in Fig. 9). Based on the field-collected reference of fires, we evaluated the suitability of FY-3D fire products in China (Table 8).

As shown in Fig. 9 and Table 8, FY-3D products achieved a good accuracy of 79.43 % in China. Meanwhile, MODIS also achieved a good accuracy of 74.23 %. As introduced above, the omission error in FY-3D and MODIS fire products was mainly attributed to a small fire area, which failed to meet the minimum fire area recognizable by sensors. When simply considering the commission error, FY-3D fire products achieved an accuracy of 88.50 %, notably higher than that of MODIS (79.69 %). This result proved that with the consideration of local underlying surfaces, FY-3D fire products are more suitable for fire monitoring in China.

As satellite instruments keep aging in the harsh space environment, the degradation of sensors is inevitable (Tian et al., 2015). Theoretically, sensor degradation can be corrected through atmospheric calibration. However, during the mission life, the solar diffuser and stability monitor required for atmospheric calibration also change across time (Wang et al., 2012). Since the MODIS instrument has been working for more than 20 years, its performance for fire detection will degrade, if it has not already, in the future. Furthermore, similarly to VIIRS and other algorithms, MODIS fire products may have large uncertainties in such regions as China (Fu et al., 2020; Ying et al., 2019). As major products of the FY-3D meteorological satellite, FY-3D fire products boast a high resolution and accuracy in China by specifically including the underlying surface parameters collected in China. Compared with MODIS and VIIRS, MERSI-II shows a resolution of 250 m in the far-infrared channel, which is the highest among meteorological satellites of the same type. The FY-3D fire identification algorithm learns from the advantages and technical ideas of MODIS and VIIRS fire identification algorithms. Furthermore, FY-3D fire products have been optimized in terms of auxiliary parameters, fire identification and re-identification as follows.

Auxiliary parameters. Since the sole use of the vegetation index is limited to reflecting combustible materials, climatic boundaries and geographical environment data, which have a strong influence on vegetation types and growth, are added to FY-3D fire identification.

Fire identification. FY-3D adopts the adaptive threshold and reduces the limitations caused by fixed thresholds of MODIS and VIIRS algorithms. Meanwhile, FY employs a re-identification index according to geographical latitude and underlying surface types, as well as the influence by cloud, water bodies and bare land, and the comprehensive consideration of multiple influencing factors increases the accuracy of fire identification. Thirdly, since the far-infrared channel plays an important role in fire identification and FY-3D has a high resolution of 250 m in that channel, the accuracy of fire identification is improved.

Fire re-identification. FY-3D fire products can be used for both global climate change research and such practical implementations as forest and grassland fire prevention with a higher requirement for accuracy. Based on the initially identified fire spots, FY-3D employed the re-identification index to further remove false fire spots at cloud edges, water body edges and other high-reflection underlying surfaces.

The MODIS fire product is one of the most significant and frequently employed fire products with mature algorithms. Compared with MODIS, FY-3D receives limited emphasis for its capability for fire monitoring, which is mainly attributed to its short service periods. On one hand, due to its long time series and general reliability, MODIS fire products have remained a popular choice for monitoring long-term variations in fire spots across the world. However, the long-term running of MODIS sensors has led to growing uncertainties about the quality of recent and future MODIS fire products. In this regard, thanks to its similar spatiotemporal resolution, high consistence and visiting time difference of less than 1 h, FY-3D fire products have the potential to be widely employed as a potential alternative to and continuity for global MODIS fire products. Meanwhile, FY-3D fire products have a higher reliability in China and its surrounding regions than other fire products. Therefore, FY-3D fire products are an ideal selection for fire monitoring in China.

The main implementation of FY-3D fire products is fire monitoring. For vast forest and grassland areas, it is inefficient and time-consuming for manual and aircraft patrols to monitor wildfires. Satellite remote sensing can work for a continuous space with a wide monitoring range, providing massive amounts of information in fire detection, disaster relief and post-disaster assessment. In addition to fire spot identification and real-time fire tracking, the impact of pollutants produced by biomass combustion on the environment is another important topic. In China and Southeast Asia, air pollution caused by biomass burning has intensified in recent years. Agricultural activities such as crop residue burning and wildfires (e.g., forest fires and grassland fires) emit airborne pollutants (e.g., PM2.5, PM10, CO). In this regard, FY-3D fire products can be used as the emission sources for estimating their environmental effects.

China recently launched the FY-3E and FY-4B satellites in June and July 2021. Amid the launch and operation of a new generation of Fengyun meteorological satellites, the accuracy and timeliness of fire monitoring by meteorological satellites have been largely enhanced. Thanks to improved meteorological data, which provide a useful reference to understand the current status of combustibles and potential fire risk, the FY-3D satellite will be taken as a better data source to produce various secondary products for fire monitoring and prediction. Based on traditional fire spot identification, further research should concentrate on the assessment of the fire area, estimation of biomass carbon emissions, prediction of smoke impact, and early warning of forest and grassland fire using the series of Fengyun meteorological satellites. For instance, the water content of combustibles is closely related to temperature, light and cloud cover, which are important indicators in forest and grassland fire forecasts. However, this variable has rarely been considered in previous fire products. Based on a series of products from Fengyun meteorological satellites, such as the surface temperature, vegetation index, surface evapotranspiration, solar radiance and cloud cover, FY-3D fire products can be improved by establishing an estimation model for the water content of combustibles. Meanwhile, with fire products such as fire spots and smoke and meteorological products such as wind field data from Fengyun series satellites, we can predict the impact of smoke caused by forest and grassland fires on the atmospheric environment in the surrounding areas. In future implementations, Fengyun meteorological satellites will play a greater role in monitoring, forecasting and early warning of global fires and their ecological impacts.