Paddy field mapping and yield estimation by satellite imagery and in situ observations
Abstract
Since Asian countries are responsible for approximately 90% of the world rice production and consumptions, rice is the most significant cereal crop in Asia. In order to ensure food security and take mitigation strategies or policies to manage food shortages, timely and accurate statistics of rice production are essential. It is time and cost consuming work to create accurate statistics of rice production by ground-based measurements. Hence, satellite remote sensing is expected to contribute food security through the systematic collection of food security related information such as crop growth or yield estimation. In 2011, Japan Aerospace Exploration Agency (JAXA) is collaborating with GISTDA (Geo-Informatics and Space Technology Development Agency, Thailand) in research projects of rice yield estimation by integrating satellite imagery and in situ data. Thailand is one of the largest rice production countries and the largest rice exporting country, therefore rice related statistics are imperative for food security and economy in the country. However, satellite observation by optical sensor in tropics including Thailand is highly limited, because the area is frequently covered by cloud. In contrast, Japanese microwave sensor, namely Phased-Array L-Band Synthetic Aperture Radar (PALSAR) on board Advanced Land Observing Satellite (ALOS) is suitable for monitoring cloudy area such as Southeast Asia, because PALSAR can penetrate clouds and collect land-surface information even if the area is covered by cloud. In this study, rice crop yield over Khon Kaen, northeast part of Thailand was estimated by combining satellite imagery and in-situ observation. This study consists of mainly two parts, paddy field mapping and yield estimation by numerical crop model. First, paddy field areas were detected by integrating PALSAR and AVNIR-2 data. PALSAR imagery has much speckle noise and the border of each landcover is ambiguous compared to that of optical sensor. To overcome this problem, we used AVNIR-2 data for object-based image analysis and derived each object was linked with backscatter coefficient of PALSAR. Then, paddy field areas were detected by using seasonal changes of backscatter coefficients. Derived paddy field map over Kohn Kean area was validated with ground-based measurements and it showed high accuracy. Finally, in order to estimate rice crop yield, numerical crop model was run with model parameters related to physiological aspect of rice and meteorological data collected by Automatic Weather Station (AWS) placed at study area, field survey and satellite products. This processing was implemented all over the detected paddy filed areas and overall yield was estimated by counting up each result. Consequently, it was found that the yield estimation was reasonable validated with agricultural statistics in Thailand.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2011
- Bibcode:
- 2011AGUFM.B13B0564O
- Keywords:
-
- 0402 BIOGEOSCIENCES / Agricultural systems;
- 0430 BIOGEOSCIENCES / Computational methods and data processing;
- 0480 BIOGEOSCIENCES / Remote sensing;
- 9320 GEOGRAPHIC LOCATION / Asia