INTEGRATING REMOTELY SENSED DATA WITH AN ECOSYSTEM MODEL
TO ESTIMATE CROP YIELD IN NORTH CHINA
a, b, * a 11: "a
P. Yang ^5, G.X. Tan ^, Y. Zha®, R. Shibasaki*
“Institute of Industrial Science, Tokyo University, Tokyo 153-8505, Japan
yangpeng@iis.u-tokyo.ac.jp, (gxtan, shiba)@skl.iis.u-tokyo.ac.jp
° Institute of Natural Resources & Regional Planning, Chinese Academy of Agricultural Sciences
Beijing 100081, China - yzha@caas.ac.cn
Commission VII, WG VII/2
KEY WORDS: Remote Sensing, Agriculture, Crop, Estimation, Integration, Model, Accuracy, GIS
ABSTRACT:
This paper describes a method of integrating remotely sensed data (the MODIS LAI product) with an ecosystem model (the spatial
EPIC model) to estimate crop yield in North China. The traditional productivity simulations based on crop models are normally site-
specific. To simulate regional crop productivity, the spatial crop model is developed firstly in this study by integrating Geographical
Information System (GIS) with Environmental Policy Integrated Climate (EPIC) model. The integration applies a loose coupling
approach. Data are exchanged using the ASCII or binary data format between GIS and EPIC model without a common user interface.
It is crucial for the simulation accuracy of the spatial EPIC model to get the detailed initial conditions (sowing date, initial soil water
content, etc) and management information (irrigation schedule, fertilizer schedule, tillage schedule, etc). But when applied at a large
scale, the initial conditions and management information are most unlikely obtained through direct measurement. Therefore, the
spatial EPIC model is integrated secondly with the MODIS LAI product from the Earth Resources Observation System (EROS) Data
Center Distributed Active Archive Center. The integration of the MODIS LAI product makes the real time information taken into
account in the simulation of spatial EPIC model, such as the amount of solar radiation captured by plant canopies, soil-water or
nutrient effects on crop growth, and the effects of natural or man-made disturbances caused to crop yield. Finally, the method is
conducted to estimate the Winter Wheat yield in North China in the year of 2003.
1. INTRODUCTION
Monitoring agricultural crop conditions during the growing
season and estimating the potential crop yields are both
important for the assessment of seasonal production (Paul et al.
2003). The accurate and real-time estimation of crop yield in
provincial and national level are of great interest to the
Department of Agriculture in many countries. Integrating of
satellite data and crop productivity models is one of the most
important quantitative analysis methodologies for yields
estimation in regional level.
The traditional crop yield estimation based on satellite data is
using the empirical relationships between dry biomass of
various crops and Vegetation Indices, which are combinations
of visible and near infrared bands. Hamar et al. (1996)
established a linear regression model to estimate corn and wheat
yield at a regional scale based on vegetation spectral indices
computed with Landsat MSS data. Similar relationships are
obtained on various crops (for example: Rasmussen [1992] for
millet yield, Manjunath et al. [2002] for wheat yield). Although
the VI approach is simple, the relationships only have a local
value and are sensitive to soil and atmospheric conditions as
well as measurement geometries. To estimate crop yield in any
conditions, it is necessary to describe the physiological and
biological mechanisms, which control crop growth and
development (Moulin et al. 1998). Therefore various
mechanistic models are inevitable to be integrated with remote
sensing data for yield assessment of major crops in regional
scale.
Mechanistic models can simulate the time profiles of the crop
state variables (leaf area index, crop stress factor, potential
biomass increase etc.) and of energy, carbon, water and nutrient
fluxes at the crop-soil-atmosphere interfaces. For more than
three decade, mechanistic models have been developed for the
major crops in the world. Compared with many other crop
models around agro-ecosystems, the EPIC Model seems to be
more suitable to simulate crop yields for relative comparisons of
soils, crops, and management scenarios and has a good accuracy
to estimate field yields (Tan et al. 2003). It was originally
developed by United States Department of Agriculture to
examine the relationship between soil erosion and agricultural
productivity. The model integrates the major processes that
occur in the soil-crop—atmosphere—management system,
including: hydrology, weather, erosion, nutrients, plant growth,
soil temperature, tillage, plant environmental control, and
economics (Sharpley et al. 1995). Extensive tests of EPIC
simulations were conducted at over 150 sites and on more than
10 crop species and generally those tests concluded that EPIC
adequately simulated crop yields (Easterling et al. [1998],
Izaurralde et al. [2003 ]).
The different ways to integrate a crop model with the
radiometric observations were described initially by Maas
(1988). Delecolle et al. (1992) and Moulin et al. (1998)
* Corresponding author. Tel.: *81-3-5452-6417; Fax: *81-3-5452-6414; E-mail address: yangpeng@1is.u-tokyo.ac.jp
150
International
HC ea
identified the
driving varia
updating of
the model; «
resolution of
step requirer
direct use ol
the other thr
developed
introducing
Specific Wa
(1996) used
over sugar |
adjusted par
growth rate,
current resea
remote sensi
whole yield
Guerif et al.
with the ci
establishmer
crops. Paul
model with :
North Dako
the input pa
regional levi
coupling rad
LAI produc
Spectroradic
be suit for
operational
The objectis
yield model
USDA (Un
(Erosion Pri
MODIS LA
Geographica
The study :
Tianjin, the
Henan Prov
longitude by
Figure ]. '
km” area
