AGRICULTURE DROUGHT AND FOREST FIRE MONITORING IN CHONGQING 
CITY WITH MODIS AND METEOROLOGICAL OBSERVATIONS* * 
HONGRUI ZHAO a ZHONGSHI TANG a , BIN YANG b AND MING ZHAO a 
a 3S Centre, Tsinghua University, Beijing, 100084, China 
b College of Resources and Safety Engineering, China University of Mining and Technology (Beijing), Beijing, 
100083, China 
KEY WORDS: Forest Fire, Remote Sensing, Monitoring, Agriculture drought 
ABSTRACT: 
In 2006, Chongqing suffered from agriculture drought that happens once a hundred years, and forest fire was caused at the same 
time. In order to provide accurate and timely data to Chongqing Government for their decision making, we took MODIS (Moderate 
Resolution Imaging Spectroradiometer) as remote sensing data resource, and meteorological observations as another important data. 
To simplify the problem and make the solution more practical, agriculture drought is expressed with soil moisture index (SI) which 
is generated from LST(land surface temperature) and NDVI. For LST retrieval, a fast statistical model based on meteorological 
observations is deduced. Forest fire monitoring is accomplished with a comprehensive threshold method, in which reflective bands, 
middle infrared bands and thermal infrared bands are used together. 
1. INTRODUCTION 
Chongqing city is one of the most important cities in the 
Southwestern of China. With fast developing economy and 
frequent human being activities, its ecological environment is 
becoming weak and disasters occur frequently in this area. In 
2006, Chongqing suffered from agriculture drought that 
happens once a hundred year, and forest fire was caused at the 
same time. In order to provide accurate and timely data to 
Chongqing Government for their decision making, we took 
MODIS (Moderate Resolution Imaging Spectroradiometer) as 
remote sensing data resource, and meteorological observations 
as another important data. To simplify the problem and make 
the solution more practical, a scheme that combines remote 
sensing data with meteorological data is proposed for 
monitoring agriculture drought and forest fire in Chongqing. 
This paper is divided into three parts. Part I is about land 
surface temperature (LST) retrieval, which is one of the most 
important parameter in environment remote sensing. It is the 
foundation of both agriculture drought and forest fire 
monitoring. A LST retrieval physical model is analyzed. 
Though the retrieved LST meets a higher precision, however, 
the method is not as efficient as needed. A fast statistical model 
based on meteorological observations is deduced. 
Following LST retrieval, part II describes agriculture drought 
monitoring. Agriculture drought is regarded as a complex 
natural event and one of the most damaging environmental 
phenomena. Combining with the operational meteorological 
observations, a soil moisture index (SI) was presented. SI was 
generated according to NDVI and LST for operational 
production. 
However, drought monitoring with remote sensing is difficult, 
since drought is affected by many factors and always lasting for 
longer time. With the operational meteorological observations 
included, we try to deduce a simple method which can be used 
in daily drought monitoring in Chongqing city. 
Based on the monitoring result, the character of agriculture 
drought in Chongqing is analyzed at the end of part II. 
Part III discusses about forest fire remote sensing monitoring. 
Forest fire broke out frequently in summer of 2006 in 
Chongqing city. It resulted from the lack of rain and a 
continuous hot weather which was rare in the century but 
happened in summer 2006 in Chongqing. As a result, forest fire 
occurred 6 times a day in average from August 1 st to 13 th in 
Chongqing city. 
For higher temperature forest fire points, channel 7 of MODIS 
is mainly used and the fire points are detected in a more direct 
way. 
Non-high temperature forest fire points are classified into 3 
types. They are the points before firing, the fired points and the 
firing points with a smaller area than that of the image spatial 
resolution. A comprehensive threshold method was proposed 
with reflective bands, middle infrared bands and thermal 
infrared bands used together. Reflective bands were used to 
eliminate pseudo-fire points that resulted from city hot island. 
Middle infrared bands and thermal infrared bands were used to 
detect 3 types of non-high temperature forest fire points. The 
forest fire monitoring result is analyzed at the end of this part. 
In comparison with the soil moisture observation, the 
operational meteorological data, the remote sensing inversion 
result of soil moisture index is coherent with the ground truth. 
However, our research reveals that the drought monitoring 
model will be more practical along with accumulation of 
drought monitoring data including remote sensing data and 
meteorological observations. In fact, both the inversion model 
• Supported by the National Natural Science Foundation of China (Grant No. 40771135) 
• Corresponding Author * HONGRUI ZHAO. Telephone: 8610-62794967; Email: zhr@tsinghua.edu.cn) 
421