Characterizing the 2015 Indonesia Fire Event Using Modified MODIS Aerosol Retrievals

Abstract

Abstract. The Indonesian fire and smoke event of 2015 was an extreme episode that affected public health and caused severe economic and environmental damage. The MODIS Dark Target (DT) aerosol algorithm, developed for global applications, significantly underestimated regional aerosol optical depth (AOD) during this episode. The larger-than-globalaveraged uncertainties in DT product over this event were due to both an overly zealous set of masks that mistook heavy smoke plumes for clouds and/or inland water, and also an aerosol model developed for generic global aerosol conditions. 15Using Aerosol Robotic Network (AERONET) Version 3 sky inversions of local AERONET stations, we created a specific aerosol model for the extreme event. Thus, using this new less-absorbing aerosol model, cloud masking based on results of the MODIS cloud optical properties algorithm, and relaxed thresholds on both inland water tests and upper limits of the AOD retrieval we created a research algorithm, and applied it to 80 appropriate MODIS granules during the event.Collocating and comparing with AERONET AOD shows that the research algorithm doubles the number of MODIS 20 retrievals greater than 1.0, while also significantly improving agreement with AERONET. The final results show that the operational DT algorithm had missed approximately 0.22 of the regional mean AOD, but as much as AOD = 3.0 for individual 0.5° grid boxes. This amount of missing AOD can skew the perception of the severity of the event, affect estimates of regional aerosol forcing, and alter aerosol modeling and forecasting that assimilate MODIS aerosol data products. These results will influence the future development of the global DT aerosol algorithm. 25

AERONET Remotely Sensed Measurements and Retrievals of Biomass Burning Aerosol Optical Properties During the 2015 Indonesian Burning Season

“…In Figure d, MODIS satellite C6 DT algorithm retrievals of AOD for Borneo on this date show that retrievals were not made in the densest smoke areas, only on the edges or periphery where the AOD is much lower. This suggests that the operational C6 DT algorithm screened these heavy smoke features due to both cloud and a conservative filter that was intended to mask out conditions that are not optimal for retrieving aerosols (Shi et al, ). This type of event led to the development of a new high AOD event DT algorithm that relies in part on the aerosol optical properties that are retrieved from AERONET retrievals during the same event (Shi et al, ).…”

“…This suggests that the operational C6 DT algorithm screened these heavy smoke features due to both cloud and a conservative filter that was intended to mask out conditions that are not optimal for retrieving aerosols (Shi et al, ). This type of event led to the development of a new high AOD event DT algorithm that relies in part on the aerosol optical properties that are retrieved from AERONET retrievals during the same event (Shi et al, ). Figure e is the same except for MODIS DB, indicating significantly greater coverage than the DT retrievals of AOD over the land areas in Borneo.…”

“…The MODIS DT algorithm does not retrieve any AOD from 5 September through 12 October 2015 (>1 month) corresponding to the period with the consistently highest AOD as measured from AERONET. This is due to multiple reasons: (1) The extremely high aerosol loadings are mistaken as cloud by the low cloud mask, (2) these features are filtered out by inland‐water mask as the retrieving conditions are not optimal, (3) the MODIS DT retrieved AOD are higher than the upper limit of AOD in the algorithm, which is caused by using a too absorbing aerosol model (Shi et al, ). Note that Shi et al () developed an event‐based MODIS DT research algorithm targeting this fire event over Indonesia.…”

“…These are the highest AOD ever estimated from AERONET data, and we have no knowledge of smoke AOD values this high having been previously reported in the scientific literature. Retrievals of midvisible AOD from the standard MODIS algorithms (DT, DB, and MAIAC algorithms) consistently underestimated the magnitude of the AOD in the peak burning region (Palangkaraya site) during this event and also did not make retrievals on many high AOD days. However, the special high AOD event algorithm recently introduced and described by Shi et al () overcomes many of these shortcomings in MODIS satellite AOD retrieval for this particular event. Likewise, the new VIIRS DB greatly increases coverage in these heavy‐smoke situations, although the retrievals are inherently more uncertain in such extreme environments. The new hybrid sky radiance scan available only in the new model Cimel Model‐T Sun‐sky radiometers allowed for the retrieval of aerosol refractive indices and size distributions at higher AOD levels due to retrieval capability at smaller SZAs when the 440‐nm signal is not completely attenuated.…”

“…Additionally, a new version of the Cimel Sun‐sky radiometer with advanced capability was deployed in two sites impacted by this event: Palangkaraya, Indonesia, (southern Borneo) in one of the major biomass burning source regions and also Singapore, a downwind site ~1,200 km from southern Borneo and ~500 km from southern Sumatra that was heavily impacted by these biomass burning emissions from Indonesia in 2015. We also present some of the standard Moderate Resolution Imaging Spectroradiometer (MODIS) satellite remote sensing retrievals of midvisible AOD during this event, noting that this topic, including recently developed algorithms to measure higher AOD events, are much more thoroughly investigated in other work (Shi et al, ).…”