Author's personal copy ARTICLE IN PRESS BIOMASS AND BIOENERGY 32 (2008) 35 – 43 Available at www.sciencedirect.com http://www.elsevier.com/locate/biombioe Using spatial information technologies to select sites for biomass power plants: A case study in Guangdong Province, China Xun Shia,, Andrew Elmoreb, Xia Lic, Nathaniel J. Gorenced, Haiming Jine, Xiaohao Zhangf, Fang Wangc a Geography Department, Dartmouth College, 6017 Fairchild, Hanover, NH 03755, USA Environmental Studies Program, Dartmouth College, Hanover, NH 03755, USA c School of Geography and Planning, Sun Yat-sen University, Guangzhou 510275, China d Epic Systems Corporation, Madison, WI 53703, USA e TX Energy, LLC, Houston, TX 77056, USA f Guangzhou Institute of Geography, Guangzhou 510070, China b art i cle info ab st rac t Article history: Biomass is distributed over extensive areas. Therefore, transportation cost is a critical Received 8 September 2006 factor in planning new biomass power plants. This paper presents a case study of using Received in revised form remote sensing and geographical information systems (GIS) to evaluate the feasibility of 5 June 2007 setting up new biomass power plants and optimizing the locations of plants in Guangdong, Accepted 6 June 2007 China. In this study, the biologically available biomass was estimated from MODIS/Terra Available online 15 August 2007 remote sensing data. The amount of biomass that is usable for energy production was then Keywords: Biomass estimation Site selection Transportation cost Remote sensing Geographical information systems Supply area 1. derived using a model incorporating factors including vegetation type, ecological retaining, economical competition, and harvest cost. GIS was employed to define the supply area of each candidate site based on transportation distance along roads. The amount of usable biomass within the supply area was calculated and optimal sites were identified accordingly. This study presents a procedural framework for taking advantage of spatial information technologies to achieve more scientific planning in bioenergy power plant construction. Introduction Biomass energy has been considered a successful alternative to fossil fuels [1,2]. Unlike fossil fuels, however, biomass is distributed over extensive areas. Thus the transportation cost becomes a critical factor in planning new biomass power plants [1–4]. Transportation costs can be reduced by optimizing the locations of power plants. Spatial information technologies, particularly remote sensing and geographical information systems (GIS), can be highly helpful in evaluating & 2007 Elsevier Ltd. All rights reserved. the feasibility of setting up new biomass power plants in a given region and optimizing their locations. Remote sensing can be used to efficiently locate biomass over vast areas. For example, Baath et al. [5] used Landsat 5 TM and SPOT 2 images, with reference to the national forest inventory, to estimate the current and future forest fuels in two areas in Sweden. As a matter of fact, within the field of remote sensing a vast literature exists for land-use/landcover classification and net primary production (NPP) estimation, which can be integrated to estimate biomass energy Corresponding author. Tel.: +1 603 646 0884; fax: +1 603 646 1601. E-mail addresses: xun.shi@dartmouth.edu (X. Shi), lixia@mail.sysu.edu.cn (X. Li), IGCCenergy@gmail.com (H. Jin). 0961-9534/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.biombioe.2007.06.008 Author's personal copy ARTICLE IN PRESS 36 BIOMASS AND BIOENERGY potential for different land-cover classes. This can be done in two ways: (1) by applying pre-determined ‘‘available biomass’’ values to each land-cover unit [5] and/or (2) by applying standard NPP-to-available biomass ratios to each land-cover unit [6]. In the first case, the assumption is made that variability in available biomass occurs spatially but not temporally (i.e., is determined by the land cover and is not influenced temporally by climate). The second method incorporates temporal variability (such as that due to drought or shifts in the success of agricultural management) in addition to spatial variability. Note that for agricultural land supporting annual crops, annual NPP is equivalent to the standing biomass at the time of harvest. GIS, on the other hand, is a powerful tool to integrate data of various factors and to perform spatial analyses for feasibility evaluation and location optimization. Noon and Daly’s BRAVO (The Biomass Resource Assessment Version One) might be the first GIS-based decision support system in the biomass energy sector [1]. The authors of BRAVO found that transportation accounted for a major part of the overall cost, and thus power plant location is crucial to the feasibility of converting coal-fired power plants into co-firing plants (plants using both coal and biomass as the fuel source). Since then, quite a few studies have been done on site selection for biomass power plants [e.g., 2–4,7]. Generally, site selection in this field has taken two approaches: suitability analysis and optimality analysis. Suitability analysis mainly uses geoprocessing procedures (e.g., buffer and overlay) to locate suitable sites based on a number of constraining and favoring factors (e.g., land use, distance to roads, distance to transmission line, etc.). Ma and Scott [7] used this analysis to locate the best sites for new power plants that utilize manure resources in New York State. The highlight of their work is the use of analytic hierarchy process for weighting different factors. Optimality analysis, on the other hand, considers the relationship between biomass and power plants similar to that between supply and demand in business and aims at finding the optimal power plant locations that minimize the transportation cost. In practice, the final decision should be based on both suitability and optimality. The work of Voivontas et al [2] for utilizing agricultural residues on the island of Crete attempted to achieve that combination. The potential of spatial information technologies in the biomass energy sector has received considerable attention, but there is a general deficit of exemplar case studies illustrating the comprehensive process beginning from a biomass survey using remote sensing and ending at a site selection using GIS. This paper presents such a case study in Guangdong Province, China. The case study started with mapping the biologically available biomass in Guangdong using the widely available MODIS/Terra and TM remote sensing data. It ended with identifying, using GIS, a few most optimal locations for setting up new biomass power plants in the province. In between the biomass estimation using remote sensing data and the site selection using GIS, a generic model was employed to estimate the amount of biomass that is eventually usable for energy production at a specific location. The construction of this model was inspired by the work of Voivontas et al [2] that classifies biomass into four sequential potential levels—theoretical, available, tech- 32 (2008) 35 – 43 nological, and economical. The model used in this study adjusts the biologically available biomass by ecologic, economic, and other factors to obtain an estimate of the usable biomass. Site selection performed by GIS was then based on the spatial distribution of the usable biomass. 2. Study area Adjacent to Hong Kong, the study area of this research, Guangdong Province, lies in the southernmost part of China. In the past 20 years, it has grown from a traditionally agricultural province into one of the most developed regions in the country. Its gross domestic production and export trade output have been significantly higher than those of other provinces in China since 1997. The province attracted over half of the nation’s foreign investment on processing industries and joint ventures. Not surprisingly, Guangdong is currently facing serious energy and environmental problems. On the other hand, the province has a subtropical or tropical monsoon climate, experiencing abundant sunshine, warmth, and rainfall. Its mean annual sunshine duration is 1745 h, mean annual temperature 22.3 1C, and mean annual precipitation 1777 mm. Its total area of land that is suitable for agriculture is 43,400 km2 and for forestry 110,000 km2 [8]. Therefore, its physical condition sets a solid basis for the development of the bioenergy industry. It is expected that the energy problem of Guangdong can be partly resolved by utilizing local biomass resources. 3. Estimation of usable biomass 3.1. A generic model for estimating usable biomass Not all biologically available biomass can be used for energy production because of a series of restrictions. These restrictions can be in ecological (e.g., soil carbon maintenance), economical (e.g., higher value uses), and other aspects. Here we propose a generic model for estimating the amount of biomass that is eventually usable for energy production at a specific location: P ¼ B  r  m  ð1  c  e  lÞ, (1) where all parameter values are specific to the vegetation type at the test location. B is the biologically available biomass, derived from the annual NPP. For crops, this means the above-ground biomass when the crops are ready for harvest. For forest, this refers to the net annual change of the above-ground biomass. It is the ultimate source of the biomass for energy production and thereby is the most basic variable in the model. Remote sensing can be used to estimate this value. The ratio r is in the fraction of B that is not primary yield. For crops, usually only the residues (e.g., straws and stems) are used for energy production, resulting in a relatively low value of r. For fast-growing grass/trees that are dedicated to energy production, r can be equal or close to 1 [9,10]. Author's personal copy ARTICLE IN PRESS BIOMASS AND BIOENERGY Parameter m converts the biomass of different vegetation types into a standard measurement of ‘‘energy content’’ [10], so as to be comparable. Fraction c is for excluding the biomass that should not be used due to ecological/environmental reasons. A major consideration here is that a certain amount of organic matter should return to the field to maintain soil quality [10]. In a forest ecological system, c may also mean the percentage of biomass that should stay in the system to preserve the microenvironments for certain species. Accurate estimation of c requires deep ecological understanding. Fraction e is for excluding the biomass used for economic purposes other than energy production, such as forage and papermaking. Energy production is among the various possible uses that compete for the available biomass. The value of e is determined by the market values of these uses, and the market values are in turn highly related to the energy policy of a region or a country. Economic modeling is needed to obtain an accurate estimate of e. Fraction l represents the loss during the whole process, e.g., the loss occurring during the harvest. 3.2. Estimations of usable biomass in Guangdong The vegetation-type information is from a detailed land-use dataset provided by The Institute of Geography in Guangzhou, Chinese Academy of Sciences. The dataset is based on the year 1998 Landsat TM images and also incorporates a great amount of ground truth data. In this study, seven major different vegetation types were identified based on the landuse data (Table 1). The last type in Table 1, ‘‘Other’’, includes all the land-use types that are not likely to contribute usable biomass for energy production, including water, industrial land, barren land, urban areas, etc. We did not consider the production of municipal wastes in this study. B in Eq. (1) was calculated from the NPP data produced by the University of Montana based on annual time series of MODIS/Terra data [11,12] processed at 1 km resolution. The process of modeling NPP from remote sensing time series is well published [13], and the uncertainties are becoming better understood [14–17]. NPP includes both above- and belowground production and is measured as the amount of carbon per unit area (g C per m2). Usually only the above-ground biomass is used for energy production, so a conversion is 37 32 (2008) 35 – 43 needed to obtain the above-ground NPP value from the overall value. Also, m in Eq. (1), which makes biomass from different vegetation types comparable based on their ‘‘energy contents’’, is usually given by the literature based on the amount of dry biomass. Thus, to use the m values in literature, NPP must also be converted to dry biomass. The following equation was used to convert NPP from MODIS/Terra data to above-ground dry biomass: B ¼ NPP a=b, (2) where a is the proportion of above-ground biomass in the total biomass and b is carbon concentration in dry biomass. For crops and grass, we used 0.8 for a [6,18] and 0.45 for b [6,19]. For forests and other woody vegetations, we used 0.5 for a [20] and 0.50 for b [21]. The residual/production ratios (r) for different crop types were derived from the ‘‘ratio of grain to grass’’ data in the MOA/DOE report [22]. r for different wood types were derived from the ‘‘coefficient of forest residual/production’’ data in [9]. Since the land-use types in the land-use dataset do not exactly match the crop and forest classes in the two references, some conversion work was done. Specifically, while ‘‘dry farm land’’ is a single type in the land-use dataset, the MOA/DOE report [22] provides ratio values for a few different crops on the ‘‘dry farm land’’, including wheat, corn, soybeans, sugarcane, etc. Using the production of each crop in 1998 as the weight, a weighted average of the ratio values of those crops on ‘‘dry farm land’’ was calculated and used as the composite ratio for this land-use type. The equation is as follows: , n n X X pi =ð1  ri Þ, (3) ri pi =ð1  ri Þ R¼ i¼1 i¼1 where R is the composite ratio for the dry farm land; ri is the ratio for crop i; pi is the production of i in 1998; and n is the number of crops. The crop production data are from [23]. The ratio for ‘‘forest’’ in the land-use data was derived in the same way, as ‘‘forest’’ is also a composite type that may include economic forest, sparse forest, protection forest, timber stand, firewood forest, and fast-growing forest. Since there are no statistic data for forest productions available, the production of each forest type was calculated by multiplying the production per unit area by the total area of that type. The values of production per unit area for each forest type are from [9]. The forest area data are from [23]. Table 1 – Parameter values of different land-use types for calculating usable biomass in Guangdong Land-use type Aboveground/ total (a) C/biomass (b) Residual/ total (r) Return to soil (c) Other economic use (e) Loss (l) Energy content (m) Paddy land Dry land Orchard Forest Shrubbery Grassland Other 0.8 0.8 0.5 0.5 0.5 0.8 0 0.45 0.45 0.5 0.5 0.5 0.45 1 0.384 See Table 2 0.091 See Table 2 0.50 1.0 0 0.5 0.5 0.5 0.5 0.5 0.5 0 0.263 0.263 0 0 0 0.263 0 0.05 0.05 0.05 0.05 0.05 0.05 0 13.97 See Table 2 16 16 16 16 0 Author's personal copy ARTICLE IN PRESS 38 BIOMASS AND BIOENERGY The composite ratios for the dry farm land and forest landuse types were calculated per prefecture, the subdivision of the province. That is, the crop production values and the Table 2 – Prefecture-specific parameter values in Guangdong Prefecture Guangzhou Shenzhen Zhuhai Shantou Shaoguan Heyuan Meizhou Huizhou Shanwei Dongguan Zhongshan Jiangmen Foshan Yangjiang Zhanjiang Maoming Zhaoqing Qingyuan Chaozhou Jieyang Yunfu Dry land Forest Residual/ total (r) Energy value (m) Residual/ total (r) 0.15 0.58 0.10 0.38 0.42 0.50 0.53 0.29 0.32 0.14 0.15 0.20 0.30 0.33 0.12 0.41 0.30 0.35 0.36 0.36 0.52 15.02 15.31 15.00 15.05 15.14 15.28 15.31 15.07 15.05 15.03 15.04 15.04 15.09 15.13 15.01 15.16 15.10 15.14 15.12 15.06 15.29 0.27 0.18 0.21 0.36 0.32 0.30 0.31 0.29 0.29 0.26 0.28 0.29 0.31 0.28 0.25 0.28 0.30 0.29 0.21 0.26 0.28 32 (2008) 35 – 43 forest area values used in the calculations were at the prefecture level, and each prefecture received its own composite ratios for the dry farm land and forest types. The m values for rice, wheat, corn, and soybean are from [24]. To be consistent, we selected the low heating value for all the crops. For crops whose energy values are unknown, we adopted 15 MJ kg1, a fairly conservative value. The composite m value for the dry farm lands in each prefecture was calculated using the same weighted average method represented by Eq. (3). The value of 16 MJ kg1 was used for grass [10]. We also simplistically use 16 MJ kg1 for all the woody vegetations, as this value was used by Lal [10] for those ‘‘energy crops’’, including willow and poplar. The MOE/DOE report [22] states that in China the percentage of crop residuals returned to or left in the field is about 15%. Li et al. [24] and Liao et al. [9] also used this number in their estimations of available biomass for energy in China. However, according to our field survey in Gaoyao, a major rice production area in Guangdong, this percentage, particularly for the paddy field, can be as high as 70%. Without the knowledge to determine whether such a high percentage is applicable to a broader area and to other crop types, and whether it is absolutely necessary in terms of maintaining soil quality, in this study we subjectively specified 50% as the value for c in Eq. (1) for all crop types. For all the other vegetated land covers we also subjectively specified 50% for c, as no sufficient information is available for us to make a more accurate estimation. In the MOE/DOE report [22], the percentage of forage use of the total crop residuals in China is 24% and for papermaking Fig. 1 – Biomass usable for energy production in Guangdong, China (MJ 900 m2). Author's personal copy ARTICLE IN PRESS BIOMASS AND BIOENERGY it is 2.3%, which lead to an estimate of e in Eq. (1) for crops to be 26.3%. For all the other types except ‘‘Other’’, we assumed that the term ‘‘residual’’ already indicated the biomass to be the ‘‘leftover’’ from other economic uses, so we assign 0 as the value for e for those types. Parameter values used in this research for Eqs. (1) and (2) are summarized in Tables 1 and 2. A map of the usable biomass in Guangdong calculated through this process is given by Fig. 1. 4. Site selection for biomass energy power plants Between the suitability and optimality analyses, this study focused on the latter, because we consider that transportation cost is a primary concern in planning new biomass power plants. The techniques that have been used to optimize the locations for biomass power plants include locationallocation modeling and supply-area modeling. Locationallocation modeling optimizes plant locations based on all usable biomass in the area, even if some of the biomass locations are beyond the reasonable transportation distance to the plant locations. Location-allocation modeling is suitable when the planner intents to send all usable biomass in the region into power plants, e.g., when the power plants are especially for utilizing the outputs from energy plantations. Recently, Ranta [3] used location-allocation modeling to perform resource-side analysis for finding optimal power plant locations for utilizing logging residuals in Finland. Differently, supply-area modeling (or service-area modeling) puts power plants at those locations surrounded by high local biomass densities. Its philosophy is that biomass too distant from a power plant is useless to that plant and thus should not be taken into account when choosing the location for the plant. Supply-area modeling is appropriate when there is a threshold for transportation cost, which is usually the case in power plant planning. There are two approaches to implementing this modeling. The first is to set the criterion about the scale of the power plant and expand the supply area of a candidate site until the biomass in the supply area meets the scale of the plant. The cost for transporting the biomass in the supply area to the plant is then calculated to evaluate whether a plant at that location is profitable. This methodology was used by Zhan et al. [4] to evaluate pricing strategies for potential switchgrass-to-ethanol conversion facilities in Alabama . The second approach directly sets a transportation cost threshold and finds sites whose supply areas (defined by the threshold) contain sufficient biomass. The optimization part in the work of Voivontas et al. [2] adopted this approach. The site-selection analysis in this study used supply-area modeling, because in a large region like Guangdong it is reasonable to set limits on transportation costs. The modeling was implemented using the second approach discussed above. This is because it is still not clear what biomass-fuel conversion technology and power plant scale will be appropriate in Guangdong, but it is relatively easy to set a reasonable transportation cost threshold. In addition, technically it is more straightforward to use the available GIS tools 32 (2008) 35 – 43 39 to define supply areas for candidate sites based on a uniform threshold. Compared with the work of Voivontas et al. [2], however, our analysis is based on much more spatially detailed information about the usable biomass. The site selection analysis of this study ranked all the candidate sites based on the usable biomass in their supply areas. Site rank and associated biomass information could then be used by the planner to evaluate whether it is feasible at all to set up power plants with a certain type and scale in the region. If even at those most optimal sites the surrounding usable biomass is not sufficient, a conclusion could be drawn that the region cannot support biomass power plants. If the evaluation result is positive, then those optimal sites should be the best candidate sites for building the plants. The analytical process for ranking all the sites and identifying optimal sites is described as follows. Since the transportation cost should be calculated along the roads, the first operation is to aggregate the biomass values over the areas to certain points on the roads [4]. These points are called biomass points herein. Note that in this study these biomass points were merely used for conducting network analysis and do not necessarily correspond to biomass storage centers or transportation stations in the real world. In this study, following Zhan et al. [4] intersections of two or more road lines (i.e., nodes in the road network) were designated to be biomass points. However, if a road link (i.e., road section between two intersections) is too long, the accuracy of the transportation cost calculation will be low. To mitigate this problem, a long road link was broken into short sections and the joint between two sections was also designated as a biomass point. In the final road dataset used, the distance from any biomass point to its nearest biomass point is o25 km along roads and there were a total of 1108 biomass points across the province. The biomass values at pixels were then aggregated to their nearest biomass points. After aggregation, each biomass point obtained a biomass amount value, which is the sum of the values from the pixels in its neighborhood. In other words, each biomass point can be considered as the representative point of a small piece of area on the ground. Technically, this was achieved by building Thiesson polygons for the biomass points and calculating zonal statistics (ArcGISs) using the Thiesson polygons and the biomass layer displayed by Fig. 1. The biomass points were also designated as the candidate sites for building biomass energy power plants. This is different from the candidate designation of Zhan et al. [4], who treated every location in the region as a candidate site and finally created a surface of optimality. This is also different from the strategy of Voivontas et al. [2], who used geometric centroids of administration units as the initial candidate sites. Our candidate designation was based on the fact that power plants, if built, should be close to roads. The next step is to find all biomass points that are within the supply area (defined by the transportation cost threshold) of a candidate site, and then aggregate the biomass values from those biomass points and attach the value to that candidate site. Transportation cost was measured simply by distance and 100 km was chosen as the transportation cost threshold. This threshold was determined based on the experience in the US [25] and our field survey in Guangdong. Once all candidate sites Author's personal copy ARTICLE IN PRESS 40 BIOMASS AND BIOENERGY 32 (2008) 35 – 43 24 - 1500 1500 - 3000 3000 - 4500 4500 - 6000 6000 - 7200 Road Unit: Mt weighted by distance Fig. 2 – Scores of candidate sites. receive values through this process, the sites are ranked and optimal sites are identified. If more than one optimal site need to be identified, it should be realized that two optimal sites may have overlapping supply areas. Mapping the supply areas is a simple way to identifying non-competing optimal sites. Two measurements were used to compare the supply areas for ranking. The first is the total amount of usable biomass in the supply area, as described above. The second is an efficiency score calculated using a distance-decaying function. This function is borrowed from the MAXATTEND model in the location-allocation tools of ArcInfos. It decreases linearly as the distance from a biomass point to the candidate site increases. The efficiency score of a candidate site is calculated as follows: X Pi ð1  di =DÞ, (4) s¼ i2N where N refers to the set of biomass points that are within the supply area of the candidate site under test; Pi is the amount of usable biomass at biomass point i; di is the road distance from i to the candidate site; and D is the cost threshold and in our case D ¼ 100 km. The efficient scores of all the candidate sites are shown in Fig. 2. 5. Results Fig. 3 shows three optimal sites identified based on the efficiency score, labeled as 1, 2, and 3, respectively, in order of their optimality. Fig. 3 also shows their supply areas and the biomass points within the supply areas. The procedure for identifying the three (or more, if necessary) optimal sites is as follows: (1) locate the candidate site with the highest efficiency score. Label it as the most optimal site (site 1) and map its supply area. (2) Locate and map the other candidate sites and their supply areas in a descending order based on the efficiency score. During this process, if the supply area of a candidate site has no or very little overlap with the supply area(s) of the already identified optimal site(s), the current candidate site is labeled as the next optimal site. Among the three sites shown by Fig. 3, only the supply areas of sites 1 and 3 have a small overlap. Similarly, Fig. 4 shows three optimal sites identified based on the total usable biomass in the supply area. These three sites are not competing at all. Site 1s in Figs. 3 and 4 are at the same location, meaning that it is the most optimal site in terms of both efficiency score and total biomass. Some statistics of the optimal sites shown by Figs. 3 and 4 are given in Table 3. Although only the top three sites in each scenario are presented here, using the calculation results more, progressively less optimal, sites can be identified if necessary. Based on this information, the planner and the engineers can decide whether it is feasible to build biomass power plant(s) in Guangdong; if yes, how many power plants can be supported in this province; what the scale of a specific power plant should be; and what energy generation technology is appropriate for a power plant in that certain scale. Author's personal copy ARTICLE IN PRESS BIOMASS AND BIOENERGY 41 32 (2008) 35 – 43 2 1 3 Optimal sites Supply biomass points Supply area Overlapping supply area Roads Fig. 3 – Optimal sites selected for Guangdong: based on the efficiency score. 4 1 5 Optimal sites Supply biomass points Supply area Roads Fig. 4 – Optimal sites selected for Guangdong: based on the total biomass within the supply area. Author's personal copy ARTICLE IN PRESS 42 BIOMASS AND BIOENERGY Table 3 – Statistics of the optimal sites Site no. 1 2 3 4 5 6. Efficiency score (Mt weighted by distance) Total usable biomass (PJ) Size of supporting area (km2) 7210 5830 5510 5710 5090 21.7 15.1 14.1 18.6 15.7 15,800 11,600 10,800 14,300 15,700 Discussion This paper presents a case study that utilizes spatial information technologies in planning biomass power plants, demonstrating a procedure from accurately and efficiently mapping the biomass using remotely sensed data to selecting optimal sites for the power plants using GIS tools. This paper also proposes a generic model to incorporate information from ecological modeling, economic modeling, governmental statistical data, and field survey to accurately estimate the amount and distribution of usable biomass. The case study illustrates the process of identifying one or a few optimal sites. Eventually, the presented approach evaluates all candidate sites based on the transportation efficiency and/or the amount of available biomass, thus setting a base for further considerations of other factors, such as land-use type, power transmission cost, environmental impact, economical interest, and political and cultural issues. For example, some areas in Guangdong have a great demand for electricity because of rapid industrial development and thus may be given higher priorities in locating the power plants. The worsening air pollution problems in Hong Kong have required the reduction of using coal-fuel in the Pearl River Delta, which is both an environmental and a political factor that needs to be taken into account. In real-world planning, these factors need to be put together with the optimality information to give a more comprehensive evaluation to a given candidate site. Some limitations in the presented case study should be explicitly pointed out, as they may provide guidance for the future work. First, we were not able to acquire sufficient information to accurately estimate values for parameters c (ecological retaining), e (competing economical uses), and l (harvest cost) in Eq. (1). In the real world, however, these factors may have critical impacts on the amount and distribution of usable biomass. For example, in our study almost all of the optimal sites that we identified to a great extent rely on forest biomass and did not take into account the current government policy that prohibits forest harvest for any use. Another issue is the amount of crop residuals that should be returned to soil. There are some debates on whether crop residuals should be used as an 32 (2008) 35 – 43 important energy source [10], and our field work in Guangdong indeed found that a very high percentage of rice residuals were returned to field in order to maintain soil quality. However, it is not clear whether that high percentage is also applied to other crops and is consistent across the entire province, and whether that high percentage is absolutely necessary. Accurately estimating those parameter values requires intensive field work and sophisticated modeling, which should be included in the agendas of future projects. Second, the NPP data used for this study are for the year 2004, while the most updated land-use data we acquired are based on 1998 information. It is well known that Guangdong has observed a significant land-use change during the past decade [26], so it is expected that there are errors when applying the land-use data to the NPP data. However, most land-use changes (mainly urban sprawl) occurred in the mid-southern part of the province on the Pearl River Delta [27], where the vegetation density has been relatively low anyway. The northern and eastern parts of the province, the traditional agricultural and forest areas, are the main sources of biomass and fortunately have experienced much less land-use changes since 1998; thus, we consider that the use of the currently available landuse data is justifiable. Nevertheless, more updated data should provide more accurate results and our method is flexible enough to easily incorporate any available land-use datasets. Third, the road dataset we acquired only contains the lines of major roads. With this dataset, in some areas, the road densities are low and the Thiessen polygons created for the biomass points are fairly large. Since it has to be assumed that the transportation within a Thiessen polygon follows straight lines and its cost has been neglected in this study, large Thiessen polygons reduce the accuracy of the analysis. To mitigate this problem, we have broken long road lines into shorter sections to increase the number of biomass points. However, more detailed road data, especially if the data contain information about speed limit, paving condition, slope gradient, etc., will lead to more accurate estimate of transportation cost. Fourth, still due to the limitation in data availability, the evaluations for the locations nearing the province border may be inaccurate, since it has to be assumed that these locations do not receive biomass from neighboring provinces. This edge effect problem is not uncommon in preliminary studies on planning biomass power plants [4]. Fifth, in this study we did not consider the biomass that is not directly from vegetation, such as biomass from municipal wastes and manure. The amount and distribution of municipal wastes and manure cannot be estimated from remotely sensed data, but once these data are available, they can be coupled with the vegetation biomass estimates for the GIS analysis. This study sets an example of using detailed spatial data and developed spatial information technologies to achieve a more scientific planning for building biomass power plants. The knowledge acquired from this pilot project is valuable when the project is expanded to cover a larger area of China. Author's personal copy ARTICLE IN PRESS BIOMASS AND BIOENERGY Acknowledgment The authors highly appreciate the support from The Rockefeller Center for Public Policy and the Social Sciences, Dartmouth College. R E F E R E N C E S [1] Noon CE, Daly MJ. GIS-based biomass resource assessment with BRAVO. Biomass & Bioenergy 1996;10:101–9. [2] Voivontas D, Assimacopoulos D, Koukios EG. Assessment of biomass potential for power production: a GIS-based method. Biomass & Bioenergy 2001;20:101–12. [3] Ranta T. Logging residues from regeneration fellings for biofuel production—a GIS-based availability analysis in Finland. Biomass & Bioenergy 2005;28:171–82. [4] Zhan FB, Chen X, Noon CE, Wu G. A GIS-enabled comparison of fixed and discriminatory pricing strategies for potential switchgrass-to-ethanol conversion facilities in Alabama. Biomass & Bioenergy 2005;28:295–306. [5] Baath H, Gallerspang A, Hallsby G, Lundstrom A, Lofgren P, Nilsson M, et al. Remote sensing, field survey, and long-term forecasting: an efficient combination for assessments of forest fuels. Biomass & Bioenergy 2002;22:145–57. [6] Lobell DB, Hicke LA, Asner GP, Field CB, Tucker CJ, Los SO. Satellite estimates of productivity and light use efficiency in the United States agriculture, 1982–1998. Global Change Biology 2002;8:722–35. [7] Ma J, Scott NR. Siting analysis of farm-based centralized anaerobic digester systems for distributed generation using GIS. Biomass & Bioenergy 2005;28:591–600. [8] Homepage of People’s Government of Guangdong Province. /http://www.gd.gov.cnS (Last visited on August 30, 2006). [9] Liao C, Yan Y, Wu Ch, et al. Study on the distribution and quantity of biomass residues resource in China. Biomass & Bioenergy 2004;27:111–7. [10] Lal R. World crop residues production and implications of its use as a biofuel. Environment International 2005;31:575–84. [11] Zhao M, Heinsch FA, Nemani RR, Running SW. Improvements of the MODIS terrestrial gross and net primary production global data set. Remote Sensing of Environment 2005;95:164–76. [12] Zhao M, Running SW, Nemani RR. Sensitivity of Moderate Resolution Imaging Spectroradiometer (MODIS) terrestrial primary production to the accuracy of meteorological reanalyses. Journal of Geophysical Research 2006;111:G01002. [13] Running SW, Nemani RR, Heinsch FA, Zhao MS, Reeves M, Hashimoto H. A continuous satellite-derived measure of global terrestrial primary production. BioScience 2004;54(6):547–60. 32 (2008) 35 – 43 43 [14] Turner DP, Ritts WD, Cohen WB, Maeirsperger TK, Gower ST, Kirschbaum A, et al. Site-level evaluation of satellite-based global terrestrial gross primary production and net primary production monitoring. Global Change Biology 2005;11:666–84. [15] Turner DP, Ritts W, Zhao M, Kurc S, Dunn A, Wofsy S, et al. Assessing interannual variation in MODIS-based estimates of gross primary production. IEEE Transactions on Geoscience and Remote Sensing 2006;44(7):1899–907. [16] Turner DP, Ritts WD, Cohen WB, Gower ST, Running SW, Zhao M, et al. Evaluation of MODIS NPP and GPP products across multiple biomes. Remote Sensing of Environment 2006;102:282–92. [17] Heinsch FA, Zhao M, Running SW, et al. Evaluation of remote sensing based terrestrial productivity from MODIS using regional tower eddy flux network observations. IEEE Transactions on Geoscience and Remote Sensing 2006;44(7):1908–25. [18] Steingrobe B, Schmid H, Gutser R, Classen N. Root production and root mortality of winter wheat grown on sandy and loamy soils in different farming systems. Biology and Fertility of Soils (online publication) 2001. [19] Schlesinger WH. Biogeochemistry: an analysis of global change. San Diego: Academic Press; 1997. [20] Enquist BJ, Niklas KJ. Global allocation rules for patterns of biomass partitioning in seed plants. Science 2002;295:1517–20. [21] Lasco RD. Forest carbon budgets in Southeast Asia following harvesting and land cover change. Science in China (Series C) 2002;45(Supp.):55–64. [22] MOA/DOE Project Expert Team. Assessment of biomass resource availability in China. Beijing: China Environmental Science Press; 1998. [23] Bureau of Statistics of Guangdong Province. Guangdong agriculture statistical yearbook 1999. Beijing: China Statistics Press; 1999. [24] Li J, Hu R, Song Y, Shi J, Bhattacharya SC, Abdul Salam P. Assessment of sustainable energy potential of non-plantation biomass resources in China. Biomass & Bioenergy 2005;29:167–77. [25] Aden A, Ruth M, Ibsen K, Jechura J, Neeves K, Sheehan J, et al. Lignocellulosic biomass to ethanol process design and economics utilizing co-current dilute acid prehydrolysis and enzymatic hydrolysis for corn stover. Technical report of National Renewable Energy Laboratory 2002, NREL/TP-51032438, available from /http://www.nrel.gov/docs/fy02osti/ 32438.pdfS (last visited on June 27, 2006). [26] Li X, Yeh AGO. Analyzing spatial restructuring of land use patterns in a fast growing region using remote sensing and GIS. Landscape and Urban Planning 2004;69(4):335–54. [27] Seto KC, Kaufmann RK. Modeling the drivers of urban land use change in the Pearl River Delta, China: integrating remote sensing with socioeconomic data. Land Economics 2003;79(1):106–21.