Spatial and Temporal Variation of Biomass Open Burning Emission Estimation by Using Remote Sensing Information

Name: Agapol Junpen

keyword: Biomass Open Burning

Abstract: Thailand is faced with air pollution from continuous biomass open burning especially forest fires during the dry season. This study aimed at studying on the spatial and temporal variations of air emissions from forest fires in Thailand. The study includes 4 main parts: (1) the characterization of forest fires using ground observations. (2) the relationship between the rate of fire spread and environmental factors.(3) the estimation of burned area using satellite information and ground observation, and (4) the assessment of spatial and temporal variation of emission and emission trends. The characterization of forest fires using ground observations covers (1) the study of the fire spread rate, the flame length, the fire intensity, and the burned area growth rate; (2) the study on factors related to forest fire behavior as fbel, weather, and topography; and (3) the study on the combustion efficiency of surface forest fire. In this regards, simulated fire experiments were set in actual dry dipterocarp and mixed deciduous forests, which are the most subject to fires annually. The fire simulation experiments site was located at Doi Suthep-Pui National Park, Chiangmai, Thailand. The results showed that most of the fires are surface fires, confirming the ground survey observations. The biomass fuel is the surface biomass composed of leaves, grass, twig, and undergrowth. The density of the surface biomass ranges from 2.7 to 4.6 tons/ha. Most of surface biomass is dead leaf, which represents 1.4 to 2.8 tonsha with moisture content about 5.0 to 16.0%. The flame length is within the range of 0.4 to 2.0 m, resulting in the rate of fire spread of 0.5 to 2.6 m/min, corresponding to a fire intensity of about 39 to 380 kW/m, which indicated that the fire is of low intensity level. The fire consumed almost all of the surface fuel, only the trunk of undergrowth is not burned. Regarding the fire behavior, it was found that the fire went from the bottom to the top of the topography. The fire intensity is higher for terrain slope than for flat surface since the angle between the flame and the biomass is narrower,and hence the heat can diffise more easily. Also, the dryer the biomass, the more complete combustion is. The second part is dedicatedto the relationship between fire behavior and surrounding environmental factors. It includes (1) the study on the factors related to fire spread; (2) the development of forest fire spread model; and (3) the application of the model to forecasting forest fire burned areas. The analysis of the related factors affecting the forest fire behavior was based on Pearson's correlation method. An empirical model of the rate of fire spread was then developed using the correlation between factors. The relationship between the rate of fire spread and the related factors is analyzed using non-linear regression method. The results indicated that the rate of fire spread, in m/min, depends on 3 factors, which are (1) the moisture content of the biomass fuel (%), (2) the slope of the terrain (%); and (3) the biomass fuel density (tondha). The moisture content of biomass has a negative exponential relationship with the rate of head fire spread, and describes the rate of head fire spread about 60 to 73% confidence. The slope has a positive exponential relationship, and describes the rate of head fire spread about 39 to 42% confidence. The biomass density has a positive exponential relationship, and describes the rate of head fire spread about 31 to 48% confidence. When taking into account the 3 factors together, it was found that the moisture content of fine biomass, the fine biomass density and the slope can describe the rate of head fire spread about 96 to 97% confidence. Therefore, only 3 to 4% of fire spread is caused by other factors such as wind, arrangement of fuel, etc. For the development of the forest fire burned area prediction model, the results showed that the forest fire burned area (m2) depends on the rate of fire spread (m2/min) and the burning time (min). Under assumption of constant rate of fire spread, the forest fire burned area has a positive polynomial relationship with the burning time. Additionally, in case of assuming constant the burning time, the forest fire burned area has a positive polynomial relationship with the rate of fire spread as well. This finding showed that the forest fire burned area increased rapidly with burning time, and that the risk of fast fire spread and of rapid augmentation of burned area is higher when the slope increases and the moisture content decreases. The third part is focused on the estimation of burned area using satellite information and ground observation. This part covers (1) the assessment of the spatial and temporal distribution of burned area using satellite information and ground observation; and (2) the assessment of the probability of forest fire detection by Moderate Resolution Imaging Spectroradiometer (MODIS). The assessment of spatial and temporal distributions of burned area was performed by using (1) fire hotspots detected by MODIS aboard on Terra and Aqua satellites that detect fires fiom temperature measurement, and (2) satellite images fiom LANDSAT-5 TM obtained in the year 2009 to evaluate the uncertainty of forest fire burned area assessment. The relationship between the size of burned area deduced fiom MODIS and that fiom LANDSAT-5 TM (Thematic Mapper) is analyzed based on logistic regression analysis. From the results, it was found that the burned area during 2005 to 2009 had been about 1.4 to 2.3 Mha. The comparison of MODIS data with satellite images fiom LANDSAT-5 TM showed that MODIS provided high uncertainty data. This is due to (1) using the size of pixel (1 km x 1 km) as a representative of burned area is not suitable because the actual size represents only 60% of the pixel; (2) MODIS is not designed for detecting small forest fires, and so not suitable for the case of surface fires occurred in Thailand, while in this study it was observed that 98% of forest fire occurrence in Thailand is of a size smaller than 10 ha, representing about 35% of total burned area, and only 2% of forest is a large fire (larger than 10 ha) which is about 65% of the forest fire area; and (3) the satellite didn't overpass the area when the fire occurred: most of forest fires in Thailand are small fires which have a short burning time, and which mostly occurred in the afternoon (local time), when MODIS didn't pass over the country. By correcting MODIS data with LANDSAT-5 TM information, it was found that the burned area during 2005 to 2009 was about 2.5 to 3.0 Mha. The annual burned area is actually related to the number of fire hotspots detected by MODIS, but has no clear trend. The analysis of spatial distribution vs. temporal distribution of forest fire showed that at the beginning of forest fire season, i.e. around December or January, forest fires occurred in the outskirt of the forest. In the middle to the end of the season, they took place at the inner part of the forest. The results from satellite information indicated that settlements and croplands, which are directly connected to a forest, have higher risk to get damaged by fires. The interpretation of satellite images in this study did not display any fire in the inner part of forest area On the other hand, it was found that forest fires in Thailand occur mostly in the afternoon, and that the month when fires are the most fiequent is March. The analysis of the number of detected fires indicated that it has a strong relationship with burned area, which is of logistic model type. From this result, it was found that a 1 km2 or 100 ha of burned area has a probability of 49% to be detected by MODIS. Therefore, when MODIS information is used with 95% confidence, the corresponding burned area should be larger or equal to 3.7 km2. The fourth part of this study is focused on the assessment of the spatial and temporal variation of emission and emission trends including the development of future scenarios of emissions. The forest fire emission is estimated using equations established by Seiler and Cmtzen in 1982. The spatial and temporal variation of forest fire emissions is analyzed and displayed in the form of grid density map. The emissions maps represented the cumulative of emission during forest fire season. The focus was put only on the emission which emitted from fbel combusted in the grid area. The size of the grid was set at 10 krn x 10 km, which was selected because of its sufficient resolution for implementation for forest fire control. It was found that the emissions depend on fire intensity. In this regard, the hture fire intensity can be estimated using the model developed in this study on the relationship between burned area and rate of fire spread in the mixed deciduous forest. The future burned areas were classified according to 3 climate types: (1) normal climate, (2) El Nino climate, and (3) La Nina climate. The results showed that the burned areas obtained from combination of MODIS and LANDSAT-5 TM information covered almost all actual forest fires in Thailand. The estimation on the emissions from forest fires during 2005 to 2009 indicated that the amount of CO, C02, CH4, and N2O was about 5,373,677 tons, 81,638,551 tons, 351, 356 tons, and 10,334 tons, respectively, or about 10,582,210 tons of CO2eq. They also emitted 439,915 tons of PM10. The year 2007 was found to be the year when the emissions were the largest. Annually, March is the period that has the maximum amount of forest fire emissions. The spatial and temporal distribution of emissions is similar to the spatial and temporal distribution of the burned area, which specifies that in case of small forest fire area, i.e. 1 to 25 km2 per grid, there will emit 1 to 500 tons of CO2eq per grid and 1 to 50 tons of PM 10 per grid. In case of large forest fire area, i.e. over 50 km2 per grid, there will be an emission of over 500 tons of C02eq per grid and over 50 tons of PM10 per grid. The areas with high density of forest fire emission were the forests situated in the northern, the western, and the upper northeastern parts of the country. Taking into account the annual forest fire emissions, it was found that the emissions highly fluctuated year by year and so can hardly be predictable. From the prediction of the amount of burned area in the future usirig the satellite information and the rate of burned area growth, it resulted that in the El Nino year has the highest fire intensity due to biomass has the lowest moisture content and the ambient temperature is higher than usual, the forest fire area was predicted to be about 0.55 to 1.1 Mha, which would emit 0.34 to 0.74 Mt of CO2eq and 14 to 30 thousand tons of PM10. In the normal climate year, the intensity of the fire would be of medium range, and the forest fire area would cover 0.32 to 0.85 Mha corresponding to an emission of 0.21 to 0.55 Mt of CO2eq and 9 to 23 thousand tons of PM10. In the La Nina year when the fire intensity would be the lowest due to higher amount of rainfall than usual, the burned area would be about 0.23 to 0.66 Mha, which would lead to an emission of 0.15 to 0.43 Mt of CO2eq and 6 to 18 thousand tons of PMlo.

King Mongkut's University of Technology Thonburi. KMUTT Library

Address: BANGKOK

Email: info.lib@mail.kmutt.ac.th

Name: Savitri Garivait

Role: Advisor

Created: 2011

Modified: 2555-11-08

Issued: 2012-10-30

วิทยานิพนธ์/Thesis

application/pdf

CallNumber: JGSEE314

eng

DegreeName: Doctor of Philosophy

Level: Doctoral Degree

Descipline: Energy Technology

Grantor: King Mongkut's University of Technology Thonburi

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	JGSEE314.pdf	6.36 MB	83	2025-04-05 09:39:13
2	JGSEE314ab.pdf	295.32 KB	53	2020-09-17 21:33:07

ใช้เวลา

0.03119 วินาที