• Open Access

The spatial distribution of cereal bioenergy potential in China


Correspondence: Li Zhang, tel. + 44 0 207 848 2692, fax + 44 0 207 848 2287, e-mail: li.zhang@kcl.ac.uk


Biomass energy that exists in crop residues can be used for electricity generation and fuel production. However, its spatial distribution has formed a bottleneck in its utilization. This study introduces a data fusion method that uses the Net Primary Productivity (NPP) product of the Moderate-resolution Imaging Spectroradiometer (MODIS) data as a weighting factor to downscale crop statistics from a county scale to a 1 km2 spatial resolution using GIS to accurately map the spatial distribution of cereal bioenergy potential in China. The study demonstrates that the combination of remote sensing and statistical methods improves both spatial resolution and accuracy of the results, and resolves errors and uncertainties stemming from remote sensing processes. The results of the study will allow better decision making for siting biomass power plants, which will in turn reduce the cost of transportation of materials and increase the use of bioenergy.


Bioenergy, a renewable energy, has been described as a chemical form of solar radiation energy that is stored in matter derived mainly from wood, wood waste, straw, manure, sugar cane, etc. Due to its wide availability and renewability, biomass energy has been increasingly utilized in both developed and developing countries. Advanced technologies and an industrialization mechanism for effective utilization have been formed (Berndes et al., 2003; Oecd/Iea, 2007b). For example, combustion and combined heat and power (CHP) are key conversion technologies for the production of power and heat. Fermentation, gasification, digestion, and extraction are the main conversion routes for transportation fuels, synthetic fuels, and biodiesel fuel. However, energy from most biomass fuel crops has not been effectively utilized due to transportation costs, land use constraints, and a lack of policy incentives (Suurs, 2002; Oecd/Iea, 2007a; Ren et al., 2009).

China started to encourage and develop renewable energy in the late 1970s. Since then, China has become the world's second largest energy user and has announced and implemented a series of policies and legislation to facilitate the development of bioenergy in the last 3 decades. For example, the Renewable Energy Law began to take effect in China in January 2006. A new agricultural renewable energy policy ‘Opinions on Promoting Comprehensive Utilization of Agricultural Straws’ was issued by the General Office of the State Council in August 2008. A few additional incentives including preferential loans and tax benefits were also issued by the central government. For example, a policy document ‘Provisional Measures on Special Subsidies for Straw Utilization’ was issued by Ministry of Finance in October, 2008 and then the ‘Provisional Measures on Special Subsidies for Renewable Energy Electricity’ was published in April, 2012. Bioenergy targets have also been included in the National Economic and Social Development Five-Year Plans. According to the policy document ‘National Twelfth Five-Year Plan on Comprehensive Utilization of Agricultural Straws’, the government will increase support for bioenergy development by encouraging clean and efficient and utilization of bioenergy. The comprehensive utilization rate is expected to reach 80% by 2015. Within a top-down bioenergy policy framework, local governments establish their own practical incentives under the guidance of the objectives of the central government.

The Chinese economy is predominantly agricultural and produces a significant amount of agricultural residues that can be used for generating renewable energy. According to the National Bureau of Statistics of China, 521 million tons of crops were produced in 2004, which were associated with 590 million tons of residues (Nbs, 2005). To convert crop residues to renewable energy, transportation costs are the immediate major barrier that results in low profitability and instability of the material supply in the country. For this reason, an accurate knowledge of the spatial distribution of currently available biomass resources and their potential to produce bioenergy is vital for improving biomass energy utilization.

Two types of methods are currently used to map bioenergy potential, these being remote sensing and statistical-based methods. Remote sensing is a widely recognized technology for mapping natural resources because of its efficiency and cost-effectiveness. It has been widely introduced to address ecological issues, such as carbon cycle and ecosystem health monitoring, at scales ranging from moderate resolution to global scale for terrestrial ecological research, but seldom has it been applied to energy planning and management. Traditionally, decision making for energy planning relies on statistical data and often lacks reliable spatial information. Several vegetation variables have been successfully extracted from remotely sensed data to estimate biomass, e.g. biomass, net primary productivity (NPP), absorbed photosynthetic active radiation (APAR), and light use efficiency (ε), among which NPP is the most typical variable for bioenergy potential estimation. The relationships between vegetation variables have been intensively studied by various researchers (Monteith, 1972; Sellers, 1987; Goward & Huemmrich, 1992; Field et al., 1995). Following their studies, a number of biomass estimation models were established, including the FOREST-BGC model, the BEPS model, the CASA model, the BIOME-BGC model, the GLO-PEM2 model, the PHYTOMASS model, and the 3PGS model and etc (Running & Coughlan, 1988; Potter et al., 1993; Hu et al., 1996; Hunt et al., 1996; Liu et al., 1997; White et al., 2000; Tao et al., 2005; Thanyapraneedkul et al., 2008). These ecosystem process models were used to simulate and test NPP. From the perspective of data quality, researchers have proven that the Moderate-resolution Imaging Spectroradiometer (MODIS) NPP product (MOD17) is a reliable data source that can be used to estimate biomass at 1 km2 spatial scale (Heinsch et al., 2003; Turner et al., 2006a). As an ecological concept, NPP can explain spatial variability well. It can be extended and incorporated into energy applications, helping decision-makers to enhance their ability of energy management if dealt with appropriately. In the energy applications field, bioenergy estimates are often based on statistical data at an administration scale using several empirically determined conversion factors. Related studies have been conducted on the grain-to-straw ratio discrepancy caused by regions and species, the loss coefficient, the calorific value, and the normalized coefficient (Koopmans & Koppejan, 1997; Cuiping et al., 2004a; Milbrandt, 2005; Wang et al., 2006; Liu & Shen, 2007; Zhang et al., 2008). Both methods have their advantages and disadvantages. Remote sensing provides detailed spatial information of the distribution of the estimated bioenergy potential, however, it may contain errors caused by various reasons such as atmospheric impacts to the spectral signature of the data. Statistical-based methods can present accurate data for different administration units yet lack the spatial detail that is essential for energy planning, e.g. potential site locations for bioenergy plants.

This study presents a comprehensive approach toward bioenergy potential mapping in China by describing a hybrid model which combines both statistical and remote sensing methods to produce a quantitative high quality and high-resolution bioenergy map. In this model, MODIS NPP (MOD17; NASA, Washington, DC, USA) satellite data were introduced into statistical data as a weighting factor to downscale the coarse spatial resolution at a county administration level to a fine spatial resolution of 1 km2 allowing realistic bioenergy planning and management.

Materials and methods

The climate and terrain of China's plain regions are favorable for agriculture, making it possible to exploit bioenergy utility on a large scale. The two major sources of biomass from agricultural activities in China are agricultural residues (e.g. stalks, straw, husks, and shells) on arable lands and purpose-grown energy crops (e.g. cassava and Jerusalem artichoke) on degraded lands (Milbrandt, 2005; Nijsen et al., 2012). Only agricultural residues are considered in this study because purpose-grown energy crops are subject to many strict land use limitations, such as forestation, biodiversity, food crops production, and urbanization. In China underutilized lands are predominantly used to secure food supply rather than for bioenergy. The cultivation of energy crops is widely scattered, and the locations of these crops are changeable so that the utilization efficiency of purpose-grown energy crops is lower. On the contrary, stable agricultural production is able to maintain a plentiful and persistent material supply for energy generation, thus agricultural residues are more suitable for large-scale centralized power generation. The Chinese 2006 Energy Statistics showed that agricultural straw energy supplements accounted for about 9% of the total national energy production (2321.67 million tce). In China, rice, wheat, and corn are the most widely cultivated high-yield food crops and the burning of their straw residues in field has become an urgent environmental problem in the country. These crops are mainly distributed in the flat arable areas across China. This makes it convenient to collect, transport and utilize the straw residues for bioenergy. Although other crops (such as rapeseed and peanut) can also produce straw their production and spatial distribution are very limited. Therefore, the three most important food crops in China, i.e. rice, wheat, and corn were included in this analysis.


Remote sensing data

The 2006 MODIS NPP (MOD17; NASA) annual composite (at 1 km2 spatial resolution) has been chosen for the estimation of bioenergy potential. The Moderate-resolution Imaging Spectroradiometer (MODIS; NASA) is a key instrument aboard the Terra (EOS AM) and Aqua (EOS PM) satellites that cover the whole earth surface every 1–2 days for studying global dynamics and processes. MOD17 is a product of the combination of the MODIS LAI/FPAR and daily meteorological data that is updated every 6 h by the NASA Data Assimilation Office (DAO). The accuracy of it has been proven to satisfy ecological application by many studies (Running et al., 1999). It has been widely used for various applications since the launch of the TERRA (EOS AM) satellite by NASA in 1999 (Turner et al., 2005, 2006b; Heinsch et al., 2006). Both TERRA and AQUA provide the NPP dataset at the same resolution. Qu et al. (2006) demonstrated an excellent agreement between the two instruments.

In the MOD17 algorithm, the annual NPP was calculated as the sum of the cumulative daily net photosynthesis products (PSNnet) minus the costs associated with annual maintenance and growth respiration, such that

display math(1)

where Rmois the maintenance respiration minus the maintenance respiration from the leaves and the roots, and Rg is the growth respiration. In addition, the MOD17A2 (NASA) provided an 8 day summation of the GPP and an 8 day summation of the net photosynthesis products (PSNnet). The PSNnet_1 km is equal to the GPP_1 km minus the maintenance respiration from leaves and fine roots. The PSNnet was calculated using an 8 day composite period and defined as

display math(2)

where Rml and Rmr are, respectively, the leaf maintenance respiration and the maintenance respiration of the fine root mass. The growth respiration was not taken into account in the products of the 8 day PSNnet.

Land use data

The land use/cover data often provide some essential auxiliary information for mapping bioenergy (Edwards et al., 2005; Milbrandt, 2005; Stewart et al., 2008; Nijsen et al., 2012). The land use data used in this study were obtained from the Data Center for Resource and Environmental Sciences (RESDC) of the Chinese Academy of Sciences. The fractions of each land use category in 2000 in each 1 km2 grid were derived from the Landsat MSS, TM, and ETM images by manual interpretation. The boundaries of all objects were delineated according to experts' understanding of the spectral reflectance, texture, terrain, and other object information. This data have been validated by intensive field surveys, including an accumulated survey covering 75 271 km2 across China. The overall accuracy of the land use map was 95% for 25 land use classes. It is the most accurate data among the national-scale land use data products in China (Jiyuan et al., 2002). Apart from the intensive validation, many field surveys were also carried out to guarantee the accuracy of the land use classification. We believe that this land use map satisfies both spatial resolution and accuracy requirements of this study. Due to lack of data, land use data from the year 2000 was utilized instead of 2006. Proper approximation of this can be accepted because land use change in this period in China is primarily a reduction of arable land. In fact, arable lands were drastically reduced from 128 243.10 ×103 ha to 121 800.00 × 103 ha. This indicates that the NPP data in 2006 would not exceed the scope of the arable land boundary in 2000. When the data fusion approach is conducted in the hybrid model, the inconsistency problem can be resolved.

Statistical data

The county-level ground-based agricultural statistical data collected by the National Bureau of Statistics(NBS) in 2006 were used to estimate the bioenergy quantity. This study focused on cereal straw bioenergy potential; so forest and grassland were not included. Agricultural statistics are fundamental datasets for assessing the general conditions of agricultural production and rural economy in China and are proven to be reliable and useful by various applications. The crop production data in the datasets were the results of the Sample Survey of Farm Crops, a nationwide survey designed by NBS and implemented by survey teams throughout China following standard sample selection and estimation procedures. Having agriculture census data as a reference, 130 000 sample plots were selected from approximately 20 000 villages cross country through a comprehensive multistage and multiphase stratified systematic sampling program. Crops were cut and measured at these plots to estimate national crop production. The survey is characterized by a multipurpose probability that is proportional in size to a sample design, of which the sampling error is kept within ±2% and the confidence interval is at 95%.


In this study, the crop production statistics were first converted to cereal straw residues and then to bioenergy potential. The cereal production was determined using statistics assembled at a county administration scale, which restricted the spatial resolution of the data to an administrative area. Crops grow only in arable land. This makes it possible to allocate the production statistics into the arable patches in each county therefore the spatial resolution of the data can be improved.

Straw and residues cannot be completely utilized for energy purposes because of other competitive uses such as land fertilizer, papermaking, and foraging. The spatial distribution of these alternative uses is variable and very difficult to quantify. In this study, we did not take these competitive uses into account in the spatial downscaling method. We assumed that straw residues are entirely used for bioenergy generation to assess a theoretical distribution of bioenergy potential in China. The uncertainty caused by this assumption will be discussed in the Limitations section.

Remotely sensed MODIS NPP data product (MOD17) was used to characterize the spatial discrepancies of each arable patch, as it reflected local precipitation, irradiation, temperature, and soil type. The NPP value at each pixel indicated the amount of photosynthetically converted mass by vegetation type, therefore the utilization of NPP to weigh the spatial distribution of the bioenergy estimated from statistical data at a higher spatial resolution could enable both the consideration of the regional growth variations and the enhancement of their spatial accuracy in the analysis.

Figure 1 illustrates the procedure of spatial downscaling of the crop production statistics dataset using MOD17 data. Presume that B units of potential bioenergy existed in a 16 km2 study area and in total 32 units of NPP were recorded by the MOD17 imagery in this area at 1 km2 spatial resolution, the NPP data can be used to downscale the potential bioenergy data using a data assimilation technique. The detail of the technique is described as three steps in the following context.

Figure 1.

Spatial downscaling using statistics and remote sensing data.

Step 1: Calculation of the bioenergy potential at a county level

Bioenergy potential was calculated from crop production statistics collected at a county level using an empirical index, i.e. crop straw/grain ratio that is widely used to estimate crop straw yield. First of all, the statistical grain production data were converted into straw residue using the following equation Eqn (3):

display math(3)

where Si is the straw residue quantity of crop i, Gi is the grain production of crop i, ri is the straw/grain ratio of crop i, and ηi is the collectable coefficient of the straw residue of crop i.

In reality, not all straw are collected from the field for energy use due to the practice of soil conservation (i.e. soil structure and nutrients) and lost through field collection and transportation. The collectable coefficient ηi in Eqn (3) is to consider the sustainable removal rate of straw residues, lost due to different collection methods from the fields and lost through transportation process. It is calculated using the following equation (4):

display math(4)

Where Li is the average height of crop i, Li,mc is the average height of crop i that collected using machinery, Li,mp is the average height of crop i that collected using manpower, Ji is the percentage of area under machinery collected to the total area in the county, Zi is the lose rate during collecting and transporting process.

The crop straw/grain ratio (ri) is defined as the ratio of the straw dry weight to the grain yield of the crops at maturity (Koopmans & Koppejan, 1997; Milbrandt, 2005; Cn-Ny, 2009; Ren et al., 2009), shown in the following Eq.(5):

display math(5)

where mi,S is the straw weight of crop i, mi,G is the grain weight of crop i, Ai,S is the moisture content of straw i (percentage), Ai,G is the moisture and impurity of grain i (percentage), the moisture percentage in dry straw equals to 15%, and the standard ratio of moisture and impurity is 12.5%.

The straw residue was then converted to an energy measure by another empirically determined conversion factor, i.e. calorific value using the following Eqn (6):

display math(6)

where BEi is the bioenergy potential of crop i, CV is the calorific value, also called heating value, that is defined as the amount of heat energy produced during the combustion of a unit of crop residue (Callé, 2007).

The straw to grain ratio varies between crops and is dependent of climate conditions, farming practices, and cultivar properties. It highly depends on the spatial distributions and geographical conditions of these factors and is very difficult to estimate. Ratios estimated by different studies around the world vary dramatically. Research by Scarlat et al.(2010) using data from 27 European countries has showed a correlation between the straw to grain ratios and crop yield and they found that the correlation can be used for better estimation of crop residue production. Similarly, a range of straw to grain ratios of crops has been found in different studies due to various experiment conditions in China. To standardize the approaches, an official standard technical code for crop straw surveying and evaluating has been published in China following a nationwide survey that used the standard methods. Ming et al. (2008) calculated the ratios based on this nationwide sampling and experiment that are believed to well reflect the characteristics of straw/grain ratios in Mainland China. The straw/grain ratios of rice, wheat and corn used in our study are taken from the above literature, which are 0.68, 0.73, and 1.25, respectively, and the collectable coefficients of these crops are 0.78, 0.76, and 0.95 (Ming et al., 2008). Studies by Ming et al. (2008) and Cuiping et al. (2004a) have determined the straw calorific values of the abovementioned crops as 14.66, 16.56, and 16.64 MJ/kg, respectively (Table 1). These values were used as coefficients for the calculation of bioenergy potential. Affected by factors such as local climate condition, crop management, and cultivar properties, slight variations among these coefficients were seen. However, the variation was confined to such a small, finite range that researchers (Cuiping et al., 2004b; Edwards et al., 2005; Ming et al., 2008) believed that the variations were negligible and thus the coefficients could be approximately assumed spatially and temporally constant.

Table 1. Energy conversion coefficients
CropRatio of Residue to Crop (Percent)Collectable Coefficient (Percent)Moisture Content (Percent)Straw Calorific Value (MJ kg−1)

Step 2: Generation of weighting factor

NPP is defined as the net flux of carbon from the atmosphere into green plants per unit of time. As NPP represents the biomass increment, it is reasonable to use it as a weighting factor to downscale the spatial distribution of bioenergy potential. To further improve the accuracy of the NPP values used in this study, a map of arable land for three selected cereal plants was integrated into the remotely sensed MODIS NPP data as an arable contribution. This arable map was combined by two subtypes (paddy land, dry land) of RESDC land use dataset. The integration led to a new NPP map with arable contribution. This was done by multiplying the values of the two rasters, i.e. MODIS NPP and land use fraction map, on a cell-by-cell basis Eqn (7) (class 3 in Table 2).

Table 2. The seven aggregated classes for the three 1 km2 classifications in China
ClassLand use data (2000; Unit: percent)MODIS NPP (2006; Unit: kg C km−2)NPPn (2006; Unit: kg C km−2)
1Forest, woods, other forestAll forest, woody savannasForest/ woods
2Shrubland, all grasslandAll shrubland, savannas, grasslandGrass/ shrubland
3Paddy land, dry landCroplandsArable
4Permanent ice and snow, sandy land, gobi, salina, bare soil, bare rock, othersSnow and ice, barren or sparsely vegetatedBarren/ice
5All built-upUrban and built-upBuilt-up
6Swampland, beach and shorePermanent wetlandsWetlands
7Stream and rivers, lakes, reservoir and ponds, bottomlandWater bodiesWater bodies
display math(7)

The data used in this study, i.e. MODIS MOD17, land use map and agricultural statistics had different characteristics and were collected from very different sources, it was necessary to ensure their spatial and temporal consistency. Although the classification system of the Chinese land use fraction map and agricultural statistics were very similar, they differed from that of the MODIS data. Based on the research by Ran et al. (2010), a strategy was developed to convert the classifications of these data to a consistent system that included seven land use types (Table 2). Ran et al. (2010) had demonstrated that the cropland areas in the land use map and MODIS NPP data were spatially consistent with a arable accuracy of 65.09%. They also found a strong correlation between the MODIS land cover and the land use map, with a correlation coefficient of 0.98. Temporally, the land use maps were from year 2000, with a 6 year gap from the other two datasets. A comparison study was carried out to calculate the changes of arable land use in China in the year 2000 and 2006 using the MODIS NPP data. Due to urbanization between the year 2000 and 2006 in China, a few cropland patches (NPP cells) were found reduced and mainly transformed into built-up patches (non-NPP cells, NPP = 0). This indicated that the temporal inconsistency of the data was negligible for the purpose of this study.

Step 3: Spatial downscaling

The spatial downscaling of the bioenergy potential, demonstrated in Fig. 1, was conducted using the following Eqn (8):

display math(8)

Where BEi,j,k indicates the bioenergy potential of pixel i, j within county k in China; NPPi,j,k is the net primary productivity of a 1 km2 spatial resolution pixel i, j in county k; math formula is the average net primary productivity of arable land cover for each pixel in county k; BEk is the total bioenergy potential in county k, calculated by Eqn (6); and nkwas the amount of pixels in county k.


Cereal bioenergy potential maps of China at a county scale

The distribution of the estimated potential bioenergy of the three main cereal crops (rice, wheat, and corn) in China at a county level is demonstrated in Fig. 2. The bioenergy potential by rice is mainly distributed in the south of the Huaihe River and Qinling Mountains including Yunnan-Guizhou Plateau, with a high concentration along the Yangtze River catchment (Fig. 2a). A fair amount lies in Northwestern China and some are scattered in the west of Xinjiang region (Fig. 2a). The bioenergy potential by wheat is widely distributed in Central, Southwestern, and the majority of Northern China, with the highest production in the Huang-Huai-Hai Plain, a large alluvial plain embracing Shangdong, Henan, Hebei, Anhui, and Jiangsu province (Fig. 2b). The bioenergy potential of corn stretches across the country except the Qinghai-Tibet Plateau, with the highest production in Northeastern China (Fig. 2c). Figure 2d demonstrates the spatial pattern of the total bioenergy potential of all three cereal crops, of which high values cluster in Northeastern China, North China Plain, Huang-Huai-Hai Plain, and the catchment of Yangtze River (Fig. 2d).

Figure 2.

Spatial distribution of bioenergy potential in 2006 in China at a county administration scale: (a) rice bioenergy; (b) wheat bioenergy; (c) corn bioenergy; and (d) total bioenergy.

2006 NPP map of China arable contribution

Figure 3 shows the distribution of the NPP with arable contribution at 1 km2 resolution in 2006. Each 1 km2 cell value was determined by both ecological conditions (NPP) and agricultural activities (arable land). High NPP values mainly exist in the Huang-Huai-Hai Plain, Hetao Plain (including the south of Inner Mongolia province and most of Ningxia province), Sichuan Basin, Northeastern China, and Shandong Peninsula. These regions were characterized by flat landscape, sufficient precipitation, and higher radiation that lead to high productivity. High NPP values were also geographically dispersed in large parts of Yunnan and some parts of Xinjiang province, where light and precipitation are abundant. However, due to the impact of mountainous or eremium environments, the yield of bioenergy in these regions was low in reality.

Figure 3.

2006 NPP map of China with arable contribution.

Cereal bioenergy potential map of China at 1 km2 scale

Figure 4 shows the cereal bioenergy potential distribution in China in 2006 at a 1 km2 spatial resolution, where high values were evident in areas in the Northeast Plain, Huang-Huai-Hai Plain (North China Plain), and Yangtze River Middle-downstream Basin that have high solar radiation and abundant precipitation. The bioenergy potential was scarce and sparsely distributed in Northwestern China because these areas are mostly high-altitude plateaus or deserts. The total amount of potential cereal bioenergy in China in 2006 was estimated to be at 4.39 × 1012 MJ.

Figure 4.

Cereal bioenergy potential map of China at 1 km2 resolution in 2006.



Accurate spatial distribution of available resources is crucial for the establishment of a bioenergy industry as it is dependent on the evaluation of raw material supply and the calculation of costs on transportation. As stated previously, acquisition of this information is always a challenge. This study used a unique data fusion method that combined two widely used methods, remote sensing and statistics, to produce more accurate and applicable information of the quantity and distribution of the potential bioenergy from three main crops in China. MODIS NPP data were used as a weighting factor to disaggregate the bioenergy quantity from statistical data and to produce a more detailed distribution map for sustainable bioenergy planning. In this method, the input of remotely sensed data has advanced the understanding of regional differences within an administrative county, meanwhile the utilization of statistical data has removed the uncertainty of remotely sensed data.

With the aid of spatial analysis, the results of this study can be used for not only locating biomass resources in China but also identifying the most suitable locations for a proposed bioenergy plant using parameters such as annual biomass demand and collection radius. Studies by Suurs (2002) and Edwards et al. (2005) demonstrated that a bioenergy plant with a capacity of 30 MWel, which may consume 200 000 tons straw per year, is capable of collecting biomass resource within a radius from 50 to 100 km. In our study, a 50 km radius of neighborhood was applied to each 1 km2 input cell to calculate the sum of available straw supply and potential regions for developing bioenergy industry in China were identified (Fig. 5).

Figure 5.

The distribution of the potential capacity of straw supply for (million tons) bioenergy within a radius of 50 km at each 1 km2 grid cell in China.


The total estimated cereal bioenergy for China in 2006 using the data fusion method is approximately 4.39 × 1012 MJ energy equivalent, equivalent to 392 million tons of straw. This estimation is theoretical as it did not consider factors that could constrain the application of bioenergy generation. For example, if straw could only be used in the 30 MWel power plants identified in Fig. 5, the available straw resource outside the 50 km radius coverage of the plants would not be able to be used due to transportation constraints. Therefore, the estimated bioenergy potential will be 2.78 × 1012 MJ. Previous estimations of bioenergy potential in China vary dramatically because of the differences in terms of conversion coefficients, data sources, and crop types. For example, Cuiping et al. (2004b) estimated the main sources of Chinese agricultural biomass to be 533 million tons, including 255.8 million tons of corn stalks, 138.6 million tons of rice straw and 138.6 million tons of wheat, using provincial statistical data. Ming et al. (2008) investigated the ratio of straw to grain at a national level and modified the method in the former example with a new technical standard, i.e., NY/T 1701–2009 Technical code of crop straw surveying and evaluating (2009). Their results showed that the theoretical resource amount of five main crop straws in China was 433 million tons.

To assess the reliability of this data fusion method, it has been compared with methods used in other studies. Parameters such as type of method, error source and controllability, spatial scale, and study crops (Table 3) are used for comparative analysis. Within these parameters, the errors and spatial scale are mainly focused on in this study. A better method should be the one that contains fewer errors (spatial and/or numerical errors) but retains more detailed spatial information. Specific comparison is discussed below.

Table 3. A comparison between the data fusion method used in this study and other methods
Authors/timeMethodsErrors source (mainly)Errors controllabilityScaleStudy crop(s)
Wang et al. (This study)Data Fusion Method (Combined NPP and statistics)StatisticsYes1 km gridMultiple crops
Cuiping et al., 2004b; StatisticsStatisticsYesProvince-levelMultiple crops
Edwards et al., 2005; Combined statistics and land useStatisticsYes5 km gridAverage straw
Wang et al., 2006; Remote sensingRemote sensing, uncertaintyHard1 km gridAverage biomass
Elmore et al., 2008; Remote sensingRemote sensing, uncertaintyHard1 km gridSingle crop
Ming et al., 2008StatisticsStatisticsYesProvince-levelMultiple crops

The data fusion method used in this study is very similar to the statistics-based methods (Cuiping et al., 2004b; Ming et al., 2008). The bioenergy quantities of these studies were all based on statistics at an administrative level. In China, the statistics were normally collected by a national government agency using various standard methods such as surveying, sampling, census and summarized under strict standards. Therefore, it provides sufficiently reliable and accurate information on the quantity of the results. Numerical errors in these data were generally very small and easy to control. These data are normally sufficiently reliable and accurate. However, the county-level administration division in China is usually used as a basic statistical unit. The statistics in this kind of data can only represent an average condition of the entire administrative unit which leads to a poor spatial accuracy and fails to show the high heterogeneity within each county. With the contribution of remote sensing data, the data fusion method has a much higher spatial resolution (1 km2) than the statistics-based methods.

This method is also similar to the remote sensing methods (Wang et al., 2006; Elmore et al., 2008) in terms of high spatial resolution that exhibits comprehensive regional variations with detailed spatial information, and coefficients used to convert the NPP (kg C km−2) value into straw quantity (kg km−2) in each 1 km2 grid. However, using remotely sensed data as a weighting factor, it has avoided severe limitations of these methods, such as the differences between simulated imagery signatures and true surface variables that are caused by the distortion of the vegetation structure, the land surface complexity and the atmospheric conditions, and the uncertainty caused by various processing parameters.

To sum up, the data fusion method has taken a new approach in mapping bioenergy potential by incorporating accurate statistical data into the physical process of remote sensing to improve the quantitative accuracy and add spatial details to the results. It has increased the spatial resolution of the distribution map from a county scale to a detailed 1 km2 grid. The remotely sensed data MODIS NPP were only used as a weighting factor for the calculation, and thus the uncertainty problem of remote sensing data was minimized. In future studies, the controllable errors listed in the Table 3 will be tested using field measurements. For example, the quantity of straw can be subsequently controlled by ameliorating statistical survey techniques.


The lack of the latest ancillary information on crop distribution was a major limitation in this approach. Firstly, the method assumed an arable land distribution from the land use data for all crops, which may not always be true. More spatial data on different crops would be useful in the future. Secondly, crop residues are normally not only used for bioenergy generation but also used for various other purposes, including fertilizer, industrial material for papermaking, and foraging. Not all bioenergy potential will be converted into energy (Banowetz et al., 2008; Scarlat et al., 2010). However, these competitive utilizations of the crop residues are limited in China due to various reasons. For example, frequent crop rotation results in low rates of straw returns. Livestock is mainly raised by cereals (e.g. corn). The primary raw material in the Chinese paper industry is wood biomass rather than straw. A considerable amount of straw in China ends up being abandoned as an agricultural waste. Approximately 58.7% of crop residues could be available for energy according to the MOA/DOE report (Jinming & Overend, 1998; Banowetz et al., 2008). These limitations should be considered in the future studies.

Future work

The research presented in this study evaluated the available cereal resource for bioenergy and its spatial distribution in China using an innovative data fusion method that combined traditional statistics with remote sensed data to pursue analysis results with higher accuracy and detailed spatial information for planning and management. This method has overcome the disadvantages (such as single value, homogeneity, and coarse scale) of statistical data and the disadvantages (such as uncertainty) of remote sensing data and is capable of producing more reliable and practical information for the bioenergy industry and the central/local government(s). It can also be used as a benchmark for quantifying the spatial distribution of bioenergy. The results showed that there were 4.39 × 1012 MJ a−1 estimated bioenergy from cereal residues (approximately 8% of the primary energy production) in 2006 in China, mainly concentrated in three core regions of China. The results could be used for biomass source investigation, optimal site selection and renewable energy planning. The 1 km2 spatial resolution of the results has made the calculation of transportation costs possible. The limitations of this study were primarily due to the quality of MODIS NPP and the land use assessment. Therefore, our future work should concentrate on identifying the specific habitat of the crop species and analyzing the economy for bioenergy utilization.


This study was supported by Major Program of National Social Science Foundation of China (11&ZD157) and the Joint Construction Program of Beijing Education Commission of China. The authors wish to acknowledge the contribution of Christopher Ewing whose constructive suggestions helped to improve the manuscript.