Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin

Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin

Physics and Chemistry of the Earth xxx (2015) xxx–xxx Contents lists available at ScienceDirect Physics and Chemistry of the Earth journal homepage:...

2MB Sizes 1 Downloads 23 Views

Physics and Chemistry of the Earth xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Physics and Chemistry of the Earth journal homepage: www.elsevier.com/locate/pce

Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin Xinxin Zhang a, Tatiana Ermolieva b, Juraj Balkovic b,c, Aline Mosnier b, Florian Kraxner b, Junguo Liu a,⇑ a

School of Nature Conservation, Beijing Forestry University, Qinghua East Road 35, Haidian District, 100083 Beijing, China Ecosystems Services and Management, International Institute for Applied System Analysis, Schlossplatz l, 2361 Laxenburg, Austria c Department of Soil Science, Faculty of Natural Sciences, Comenius University in Bratislava, 842 15 Bratislava, Slovak Republic b

a r t i c l e

i n f o

Article history: Received 28 September 2014 Received in revised form 31 January 2015 Accepted 14 May 2015 Available online xxxx Keywords: Land use Downscaling Cross-entropy Shared Socioeconomic Pathway scenarios Heihe River Basin

a b s t r a c t Downscaling methods assist decision makers in coping with the uncertainty regarding sustainable local area developments. In particular, they allow investigating local heterogeneities regarding water, food, energy, and environment consistently with global, national, and sub-national drivers and trends. In this paper, we develop a conceptual framework that integrates a partial equilibrium Global Biosphere Management Model (GLOBIOM) with a dynamic cross-entropy downscaling model to derive spatially explicit projections of land uses at 1-km spatial resolution from 2010 to 2050 relying on aggregate land demand projections. The fusion of the two models is applied in a case study in Heihe River Basin to analyze the extent of potential cropland, grassland, and unused land transformations, which may exacerbate already extensive water consumption caused by rapid expansion of irrigated agriculture in the case study region. The outcomes are illustrated for two Shared Socioeconomic Pathway scenarios. The kappa coefficients show that the downscaling results are in agreement with the land use and land cover map of the Heihe River Basin, which indicates that the proposed approach produces realistic local land use projections. The downscaling results show that under both SSP scenarios the cropland area is expected to increase from 2010 to 2050, while the grassland area is projected to increase sharply from 2010 to 2030 and then gradually come to a standstill after 2030. The results can be used as an input for planning sustainable land and water management in the study area, and the conceptual framework provides a general approach to creating high-resolution land-use datasets. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Increasing population, growing human demands, climate change, and tropospheric pollution significantly impact Earth’s ice-free land surface, leading to the transformations of almost 50% of the ice-free land (Turner et al., 2007). Land use and land cover changes have become a major driver of global and local environmental changes (Lambin and Meyfroidt, 2011). Integrated systems analysis of future land use changes induced by socio-economic, population, technological trends is important for global and regional policy support discussions assisting decision makers in coping with the uncertainty associated with planning future development (Havlík et al., 2011; Mancosu et al., 2015). Increasing demands for land resources may have serious impacts on local water, soil, air, and increase health risks to humans. For example, transformation of forest into cropland causes changes in water, soil, and air quality (Fitzherbert et al., 2008). ⇑ Corresponding author. E-mail address: [email protected] (J. Liu).

Over the last decades, many spatially explicit land use and land cover models have been developed (Chomitz and Gray, 1996; Huang et al., 2009; Lotze-Campen et al., 2010; Schmidt et al., 2011; Li et al., 2012) to meet land management needs and evaluate possible future land use scenarios at global and regional levels (Britz et al., 2011; Busch, 2006; Dietrich et al., 2014; Ewert et al., 2005; Heubes et al., 2013; Lambin, 1997; Pérez-Vega et al., 2012; Rounsevell et al., 2006). However, regional models often oversee global drivers, while aggregate models cannot account for local, possibly alarming, water, food, energy, and environment heterogeneities. Improving our understanding of the complex processes involved in planning land use systems requires new research methods to integrate land use planning models at different scales. The downscaling procedure is based on cross-entropy theory (Shannon, 1948), which provides a powerful tool for data estimation when information in locations is not observed directly or is available in the form of balance equations or expert opinions when traditional statistical methods are inapplicable (Ermoliev et al., 2014). The challenge of estimating future land use scenarios by the downscaling method is the qualitative interpretation and the

http://dx.doi.org/10.1016/j.pce.2015.05.007 1474-7065/Ó 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

2

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx

demand for detail and precision spatial information (Mancosu et al., 2015). How to recognize the specific data at local level when the information is limited, partially correct and conflicting remains challenging. The goal of this paper is to derive land use projections through coupling the Global Biosphere Management Model (GLOBIOM, Havlík et al., 2011) with the recursive cross-entropy downscaling model. In this approach, future land use changes computed by GLOBIOM at the resolution of homogenous units varying from 10- to 50-km resolution are subsequently downscaled to a 1-km resolution grid covering the study area of the Heihe River Basin, China. The fusion of the GLOBIOM and the downscaling model is able to analyze a variety of global–local interdependencies, including those determined by plausible socio-economic, technological, and population trends summarized in the Shared Socioeconomic Pathways (SSPs) scenarios (Kriegler et al., 2012; Moss et al., 2010; Van Vuuren et al., 2012). The approach is illustrated for two alternative future development scenarios. A short description of the GLOBIOM and the scenarios are in Sections 2.2 and 2.4 respectively. We apply the approach in the Heihe River Basin, China, to investigate the extent of possible land use transformations, which may exacerbate water use in the region. The Heihe River Basin, the second largest inland river basin in the arid region, is located in Northwest China with a total basin area of 0.234 million km2 (Fig. 1). It stretches from southern permanent ice and now in Qilian Mountains to northern oases in Ejin Banner westward through western Shiyang River Basin into eastern Shule River Basin. The average annual precipitation varies considerably from south to north, ranging from 480 mm in the upstream of the basin to even less than 20 mm downstream. Because the Heihe River Basin has once been considered to have abundant water resources, it had been used as an important grain producing base in northwestern China (Qi and Luo, 2005). Nowadays, heavy dependence on irrigated agriculture has led to overexploitation of water and land resources, environment degradation in the upper and middle parts of the basin (Qi and Luo, 2005, 2007). This results in a sharp decrease of water resources also in the lower reach of the Heihe River, and severe deterioration of the ecosystems in the Ejin Banner Oasis (Qi and Luo, 2005, 2007). Rapid land use changes in

Heihe River Basin are to a large extent due to socioeconomic developments, increasing population and demands at regional and national levels in China and worldwide. 2. Methods 2.1. Conceptual framework Fig. 2 shows the conceptual framework of downscaling future land uses by integrating an economic partial equilibrium model with a downscaling approach. The output is the spatial distribution of future land uses at fine resolution (e.g., 1 km resolution in this paper). This conceptual framework includes two steps. (1) Use economic partial equilibrium model (GLOBIOM model) to obtain future land use projections in 30 major regions. (2) Apply a dynamic recursive cross entropy-based downscaling model to create future land use maps at 1 km grid resolution. 2.2. GLOBIOM Model GLOBIOM (Havlík et al., 2011) is a global recursive dynamic partial equilibrium model running for major world countries and regions. It integrates the agricultural, bioenergy and forestry sectors enabling the analysis of land use competition and land use transformations driven by increasing demands for food, feed, and biofuels. Land use projections are calculated within the available land resources depending on productivity of these resources, demand and prices for respective land use products accounting for potential export–import flows. Aggregate land use projections for 30 major regions are derived by GLOBIOM model for the period from 2000 to 2050. 2.3. Recursive cross-entropy downscaling model The fusion of GLOBIOM with a dynamic recursive cross entropy-based downscaling model has been developed (Ermolieva et al., 2013; Ermoliev et al., 2014) for the analysis of spatially-explicit land use changes consistent with global land

Fig. 1. Location of the Heihe River Basin in China.

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx

3

Fig. 2. Conceptual framework of the downscaling future land uses.

use demand projections derived by GLOBIOM. In the studies, the land use projections have been derived at the resolution of homogeneous simulation units varying from 10 to 50 km. For the case study in the Heihe River Basin, on the one hand, the resolution of the estimated land use maps is too coarse; on the other hand, the local river basin-specific information is not explicitly considered. Therefore, future land use projections in China from GLOBIOM (Ermolieva et al., 2013; Havlík et al., 2011) are further disaggregated to estimate the spatially resolved land use projections for Heihe River Basin at the grid level of 1 km resolution. In the downscaling model, Atir denotes the area of land use type i at time t in r coarse resolution unit (CRU), which is calculated according to a recursive equation.

Atir ¼ Airt1 þ

X t X t DAjir  DAijr ; j

ð1Þ

j

where DAtijr is area change from land use type i to j at time t. Let ztijlr be the fraction of aggregate CRU level area change DAtijr from land use type i into j in CRU r allocated to 1 km grid cell l. To derive future projections of land use type i, time t, in grid cell l, i.e., Atilr , we need to find out unknown shares ztijlr by solving the following system of equations: t1 Ailr þ

X

ztjilr DAtjir 

j

X ztijlr ¼ 1:

X ztijlr DAtijr ¼ Atilr ;

ð2Þ

j

ð3Þ

l

At 1 km grid level, the base year 2000, t = 1, estimates of area in different land uses area are available from historical Heihe River Basin land use and land cover maps at 1 km spatial resolution. Then total available land imposes the following constraint on land use transformations in a grid cell:

Atilr 6 1:

ð4Þ

The system (2) and (3) may be underdetermined, i.e., there may be fewer equtions than unknows and, therefore, there may be an infinite number of feasible solutions ztijlr satisfying (2) and (3). To

ensure the unique solution requires application of cross-entropy principle. We find the unknown shares ztijlr by minimizing a type of cross-entropy function (5) under constraints (2)

min CEðztijlr ; qtijlr Þ ¼ t fz

ijlr

g

 XXX ztijlr lnðztijlr =qtijlr Þ ;

ð5Þ

i¼1 j¼1 l¼1

X qtijlr ¼ 1;

ð6Þ

l

0 6 qtijlr 6 1;

ð7Þ

where qtijlr is some prior distribution of ztijlr satisfying constraints (6) and (7). For different land use types, the prior may include different factors and trends. In this study, according to local information in Heihe River Basin, we use the following equation to determine land use specific priors:

aV ðNPPÞlr þ bV ðGDPÞlr þ cV ðPOPÞlr qijlr ¼ X ; aV ðNPPÞlr þ bV ðGDPÞlr þ cV ðPOPÞlr

ð8Þ

l

where V ðNPP Þlr ; V ðGDPÞlr , and V ðPOP Þlr refer to the value of net primary production (NPP), GDP, and population density at gird cell l. a, b, and c are defined to the value ‘‘1’’ or ‘‘0’’. We use the General Algebraic Modeling System (GAMS) with NLP solvers (Brooke et al., 1998) to solve the optimization problem. Each optimization run attempts to simultaneously allocate all land use types into grid cells across Heihe River Basin subject to the defined constraints, using the prior distributions as starting point. 2.4. Information sources The goal of the downscaling is to spatially disaggregate future land use projections from CRUs (about 50 km grid) to a 1 km grid in Heihe River Basin. The information used to guide the spatial allocation includes the following: (a) Future land use projections (in CRUs, for 2010–2050) calculated by GLOBIOM within an internal collaborative project at IIASA (Ermolieva et al., 2013; Ermoliev et al., 2014; Havlík et al., 2011). Various scenarios relevant to global (including climate) change discussions have been released. For example, the radiative forcing

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

4

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx

scenarios of the Representative Concentration Pathways (RCPs; van Vuuren et al., 2011), the climate scenarios of the Coupled Model Intercomparison Project Phase 5 (CMIP5; Taylor et al., 2012), the socio-economic scenarios of the Shared Socioeconomic Pathways (SSPs; Kriegler et al., 2012; Moss et al., 2010; O’Neill et al., 2012; van Vuuren et al., 2012). RCPs are a set of four greenhouse gas concentration trajectories developed for the climate modelling community. However, in this paper we make use of two SSPs scenarios (SSP2 and SSP3) coming from SSP database at IIASA. The scenarios are chosen because they incorporate main land use change drivers, such as technological progress, population and economic growth, without including parameters of the RCPs. With the scenarios we illustrate the applicability of the integrated GLOBIOM-downscaling methodology for the analysis of local land use trends in the river basin. Without aiming at the detailed comparison, the differences between the scenario results in the short-term are expected to be not substantial, while in the long run even at this resolution the estimates may diverge. Here we present only a summary of the SSP2 and SSP3 storylines and for more details refer to the IIASA SSP database (IIASA, 2012). In SSP2, which is considered as a baseline (Dynamics as Usual) scenario, current trends in economic and population growth, technological developments, emissions and environmental pollution continue, with some progress towards achieving development goals, reductions in resource and energy intensity at historic rates, and slowly decreasing fossil fuel dependency. Per-capita income levels grow at a medium pace on the global average, with slowly converging income levels between developing and industrialized countries. In SSP 3 (Fragmented World), the world is separated into regions. Countries address energy and food security goals within their regions. Population growth in this scenario is high. Unmitigated emissions are relatively high, driven by high population growth, use of local energy resources and slow technological change in the energy sector. A regionalized world restricts trade flows. (b) Cross-entropy based downscaling of cropland, grassland, and unused land relies on existing land-cover data for the Heihe River Basin. Land-cover maps were combined with socio-economic data or other data in the case study area, including gross domestic product (GDP), population density (POP), and net primary productivity (NPP), to determine prior distributions involved in cross-entropy downscaling principle. The socio-economic drivers are likely to have a greater effect on future land use transformations than climate change at least for the next 30–40 years (Ewert et al., 2007; Rounsevell et al., 2006). The historical land cover data at 1 km resolution (for the years of 2000 and 2011) together with POP (2000), annual NPP (1998–2002), and GDP (2000) were obtained from the Heihe Data Research Group (http://westdc.westgis.ac.cn/heihe). 2.5. Model validation The downscaling model provides gridded maps of land area, which means that the downscaling results reflect how much area of certain land use types occurs in each grid cell. Our final spatial resolution is 1 km resolution. Hence, the value of land use is between 0 and 1 km2. For validation purposes, if the area of a certain land use type in a grid cell is equal to or more than 50% (P0.5 km2), this grid cell is classified within that particular land use type (Ramankutty and Foley, 1999). Finally, the cropland, grassland, unused land and other land use types were included into an aggregated land use map. The kappa coefficient is commonly used for accuracy assessment of remote sensing image classifications (Foody, 2002) and results of spatial models (Monserud and Leemans, 1992). The coefficient is an index of classification accuracy. Kappa statistics have been used to compare the model-derived downscaling results

and land cover maps provided by Heihe Data Research Group. Most reliable land use and land cover maps of the Heihe River Basin are available for 2011. We assumed that the land use and land cover map between 2010 and 2011 in Heihe River basin has not changed. Hence, a confusion matrix was constructed to express the accuracy of land use projections and the overall agreement was determined using spatial analysis. Precise comparison of the two maps showed substantial agreement of 70% (Landis and Koch, 1977; Liu et al., 2014). Table 1 (Landis and Koch, 1977) presents the scales of kappa coefficient. 3. Results and discussion 3.1. Validation of recursive cross-entropy downscaling model Table 2 shows that the overall agreement of the Heihe River Basin map and the aggregated land use map is 74%, which is high. Based on the common criterion of kappa coefficient, Table 2 illustrates that there is substantial agreement for cropland with kappa coefficient of 0.62 under both SSP2 and SSP3 scenarios. The accuracy assessment of unused land is in moderate agreement with kappa coefficient of 0.54 under SSP2 and 0.53 under SSP3 scenario. The grassland is in fair agreement with kappa coefficient of 0.36 under both scenarios. However, the grassland has the lowest accuracy. We must highlight that GLOBIOM land use projections are based on GLC2000 land-cover map (Havlík et al., 2011), while the downscaled maps at 1 km grid are based on the historical regional land use and land cover information available for Heihe River Basin. Different land use classification and methodology among these products contribute to the uncertainty of the land use downscaling results in Heihe River Basin. The accuracy assessment of grassland between GLC2000 land-cover map and the land use and land cover data for Heihe River Basin in 2000 is in fair agreement with the lowest coefficient of 0.26, while the cropland is 0.31 and unused land is 0.56. The grass land data in GLC2000 have caused the low accuracy of the downscaled results. As a whole, the kappa statistics show that the downscaling results have good agreement with the land use and land cover map of Heihe River Basin, which indicates that this recursive cross-entropy downscaling model can produce promising land use products. 3.2. Land use and land cover change in Heihe River Basin from 2010 to 2050 The downscaling model described in section 2 was performed for cropland, grassland, and unused land in Heihe River Basin. The output represents the spatial distribution of land uses at 1-km gird. Figs. 3 and 4 show the downscaling results for cropland, grassland, and unused land area from 2010 to 2050 in HBR under SSP2 and SSP3 scenarios. The cropland area is expected to increase in the midstream and downstream basins ranging from 0.005 km2 to 0.5 km2 from 2010 to 2050 under both scenarios. But a wide range of cropland area is less than 0.05 km2 in downstream. The grassland area will increase significantly in the downstream basin

Table 1 Interpretation of Kappa. Kappa coefficient

Agreement

<0 0.01–0.20 0.21–0.40 0.41–0.60 0.61–0.80 0.81–0.99

Less than chance agreement Slight agreement Fair agreement Moderate agreement Substantial agreement Almost perfect agreement

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

5

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx Table 2 Comparison of cropland, grassland, and unused land maps between land use projection and land use and land cover map. Land use projection for 2010 SSP2

SSP3

Land use and land cover map for 2011

Kappa coefficient

Overall agreement

Cropland

Grassland

Unused land

Other

Total

Cropland Grassland Unused land Other Total

2856 804 794 303 4757

401 8790 3749 3402 16,342

631 9507 60,472 1772 72,382

337 1209 2132 713 4391

4225 20310 67,147 6190 97,872

0.62 0.36 0.54

0.74

Cropland Grassland Unused land Other Total

2855 761 837 304 4757

399 8766 3779 3398 16,342

631 9474 60,509 1768 72,382

337 1191 2153 710 4391

4222 20,192 67,278 6180 97,872

0.62 0.36 0.53

0.74

Fig. 3. Dynamics of cropland, grassland, and unused land area from 2010 to 2050 under SSP2 scenario.

from 2010 to 2050, where the area in most grid cells increase by 10–30%. There are slight changes in unused land area in most regions of the downstream basins but keeping decreasing tendency from 2010 to 2050. The unused land area will increase in some grid cells of the upstream and midstream basins after 2030, and the change of area under SSP2 scenarios is more obvious than SSP3 scenario.

3.3. Comparison of different land area by county in Heihe River Basin between SSP2 and SSP3 scenarios The Heihe River Basin covers 18 counties in Qinghai, Gansu, and Inner Mongolia. Fig. 5 shows the comparison of cropland, grassland, and unused land area between SSP2 and SSP3 scenarios for 18 counties. The cropland will keep increasing in Alashanyouqi,

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

6

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx

Fig. 4. Dynamics of cropland, grassland, and unused land area from 2010 to 2050 under SSP3 scenario.

Ejin Banner, Jinchuan, Qilian, Subei, Sunan, and Yongchang counties from 2010 to 2050 under SSP2 scenario. In SSP3 scenario the trend is almost the same as in SSP2 excluding Jinta, Shandan and Yumen counties, where the cropland area decreases after 2040 under SSP2 scenario. The grassland area in Ejin Banner will substantially increase over 2000 km2 under SSP2 and SSP3. In Qilian and Sunan counties, there is a decrease in the grassland area after 2030 under SSP2 scenario. The unused land area is expected to decrease in Ejin Banner, Jinta, and Subei counties and to keep almost unchanged after 2030 under SSP2 and SSP3 scenarios. But in Qilian and Shandan counties, the unused land area will decrease and then increase after 2030 under SSP2 scenario. From the comparison of total area of cropland, grassland, and unused land for Heihe River Basin (Fig. 6), we find that the total cropland area will increase rapidly and then slow down after 2040 under SSP2 scenario, while the cropland area will keep on increasing sharply from 2010 to 2050 under SSP3 scenario. The total grassland area will increase significantly and then decline after 2040 under SSP2 scenario. There is a rapid increase in total grassland area before 2030 under SSP3 scenario, which will gradually come to a standstill after 2030. For the unused land, the total area will keep descending tendency from 2010 to 2050 SSP3

scenario. While the total unused land area will increase after 2040 under SSP2 scenario. The increased cropland area in Heihe River Basin from 2010 to 2050 is likely to be attributed to the increasing population, which will lead to the higher demand for agriculture development (Qi et al., 2006; Zhang et al., 2003). Conversion of unused land into grassland in the downstream basin takes place in locations with the program of restoring grazing area to grassland and water diversion in Heihe River Basin. Although Heihe River Basin has experienced rapid socioeconomic development, extensive exploitation of land resources in upstream and midstream led to a sharp decrease in water resources and severe eco-environment problem (i.e., desertification) in the downstream of the Heihe River (Chen et al., 2005; Qi et al., 2003; Zeng et al., 2012). After enforcement of water diversion project in Heihe River Basin, the eco-environment in downstream has been improved, especially for the partial recovery of oasis in Ejin Banner (Sun et al., 2011). 4. Conclusions and future work We introduced an integrated framework to project future land uses at 1-km resolution by combining projections from a global

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

7

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx

2040

0.9

2050

2010

2020

2030

2040

2050

0.8

2

Cropland area (×10 km )

0.4 0.3 0.2

alashanyouqi alashanzuoqi ejin banner zhongmushandan ganzhou gaotai jiayuguan jinchuan jinta linze minle qilianxian shandan Subei sunan suzhou yongchang yumen

0.1

2010

2020

2030

2040

2

Grassland area (×10 km )

5 4 3 2 1

0.3 0.2 0.1 0.0

2010

2020

2030

2040

2050

7 6

2020

2030

2040

2050

5 4 3 2 1 0

60

2010

2020

2030

2040

2050

2

2010

Unusedland area (×10 km )

2

60

0.4

alashanyouqi alashanzuoqi ejin banner zhongmushandan ganzhou gaotai jiayuguan jinchuan jinta linze minle qilianxian shandan Subei sunan suzhou yongchang yumen

6

0

0.5

3

7

0.6

8

2050

8

0.7

alashanyouqi alashanzuoqi ejin banner zhongmushandan ganzhou gaotai jiayuguan jinchuan jinta linze minle qilianxian shandan Subei sunan suzhou yongchang yumen

3

0.5

alashanyouqi alashanzuoqi ejin banner zhongmushandan ganzhou gaotai jiayuguan jinchuan jinta linze minle qilianxian shandan Subei sunan suzhou yongchang yumen

50

3

50 40 30 20 10 0

SSP2

40 30 20 10 0

alashanyouqi alashanzuoqi ejin banner zhongmushandan ganzhou gaotai jiayuguan jinchuan jinta linze minle qilianxian shandan Subei sunan suzhou yongchang yumen

3

2

Grassland area (×10 km )

2030

0.6

9

3

2020

0.7

0.0

Unusedland area (×10 km )

2010

0.8

alashanyouqi alashanzuoqi ejin banner zhongmushandan ganzhou gaotai jiayuguan jinchuan jinta linze minle qilianxian shandan Subei sunan suzhou yongchang yumen

3

2

Cropland area (×10 km )

0.9

SSP3

Fig. 5. Total area of cropland, grassland, and unused land from 2010 to 2050 by county in the Heihe river basin under SSP2 and SSP3.

land use model with information on local biophysical processes, maps of natural resources (land and water), and expert opinions. For this, we applied a fusion of GLOBIOM partial equilibrium model with a recursive cross-entropy based downscaling procedure allowing the downscaling of land use from coarse spatial resolution to fine grid cells at 1-km resolution. In the application of the downscaling model to Heihe River Basin, the downscaling results evaluated by kappa statistics show good agreement with the land use and land cover maps of Heihe River Basin. Despite the promising results, the question remains whether more detailed information and spatially explicit data of Heihe River Basin would

produce better downscaling estimates of spatial distribution of land uses and land use changes. The obvious way forward is to improve the quality of the input data, in particular by harmonizing the global (GLC2000) and the local basin-level data sources. More information can be also included to calculate the priors for the downscaling model. The proposed fusion of a large-scale GLOBIOM with the recursive cross-entropy based downscaling procedure allows the disaggregation of global or regional land uses to the scales required for local studies with hydrology and environmental models such as GEPIC (Liu et al., 2007)and SWAT (Zang et al., 2012), to investigate

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

5.0 4.9 4.8 4.7 4.6 2010

2020

2030

2040

2050

2

29

SSP2 SSP3

3

30

3

SSP2 SSP3

Total unused land area (×10 km )

2

5.1

3

2

Total grassland area (×10 km )

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx

Total cropland area (×10 km )

8

28 27 26 2010

2020

2030

2040

2050

125 124

SSP2 SSP3

123 122 121 120 2010

2020

2030

2040

2050

Fig. 6. Total area of cropland, grassland, and unused land from 2010 to 2050 in Heihe River Basin under SSP2 and SSP3 scenarios.

the effects of future land use changes on local water scarcity, or to study the interactions between climate change and land use at local scale by obtaining the aggregated land use projections for the same period as in the climate change. Meanwhile, these outputs can give the feedback to adjust the downscaling model. Our model can help the decision makers to recognize that either global or local data are limited, not perfect and even conflicting. Besides, with the increasingly highlighting the requirement to incorporate the ecosystem service into the real-world decision-making (Ian et al., 2013), previous studies have shown that land use decisions have significant influence on the values of ecosystem services at local level (Goldstein et al., 2012; Polasky et al., 2011). Therefore, our approach provides strategic planners with a tool to envisage the outcome of global and national trends and assess the implications of alternative decisions and planning strategies at fine spatial scales for the analysis of sustainable local land use and management strategies. Acknowledgements This study was supported by the National Natural Science Foundation of China (41161140353, 91325302, 91425303), International S&T Cooperation Program from the Ministry of Science and Technology of China (2012DFA91530), the National Program for Support of Top-notch Young Professionals, the Fundamental Research Funds for the Central Universities (TD-JC-2013-2). The present work was partially developed within the framework of the Panta Rhei Research Initiative of the International Association of Hydrological Sciences (IAHS) by the working group ‘‘Water Scarcity Assessment: Methodology and Application’’. We also thank Potsdam Institutes of Climate Change Impacts (PIK) in Germany for supporting J. Liu’s visits, and the University of Leeds for providing J. Liu a Cheney Senior Fellowship. The development of downscaling procedures contributes to EU SIGMA and AGRICISTRADE projects, and the project on integrated analysis for robust food, energy, water provision, conducted jointly by International Institute for Applied Systems Analysis, Laxenburg, Austria, and National Academy of Sciences of Ukraine (IIASA-NASU) project. References Britz, W., Verburg, P.H., Leip, A., 2011. Modelling of land cover and agricultural change in Europe: combining the CLUE and CAPRI-Spat approaches. Agr. Ecosyst. Environ. 142, 40–50. Brooke, A., Kendrick, D., Meeraus, A., Raman, R., 1998. General Algebraic Modeling System (GAMS): A user’s guide. Washington, USA. Busch, G., 2006. Future European agricultural landscapes—what can we learn from existing quantitative land use scenario studies? Agr. Ecosyst. Environ. 114, 121– 140. Chen, Y., Zhang, D., Sun, Y., Liu, X., Wang, N., Savenije, H.H.G., 2005. Water demand management: a case study of the Heihe River Basin in China. Phys. Chem. Earth Pt. A/B/C 30, 408–419. Chomitz, K.M., Gray, D.A., 1996. Roads, land use, and deforestation: a spatial model applied to Belize. World Bank Econ. Rev. 10, 487–512.

Dietrich, J.P., Schmitz, C., Lotze-Campen, H., Popp, A., Müller, C., 2014. Forecasting technological change in agriculture—an endogenous implementation in a global land use model. Technol. Forecast. Soc. 81, 236–249. Ermoliev Y., Ermolieva, T., Havlík, P., Mosnier, A., Leclere, D., Obersteiner, M., Kostyuchenko, Y., 2014. Estimating local-global dependencies of land use systems by downscaling from GLOBIOM model. Published by Committee for Systems Analysis and Presidium of National Academy of Sciences (Ukraine), National Member Organization of the International Institute for Applied Systems Analysis (IIASA). pp. 228–240, ISBN 978-966-02-7376-4, Kyiv. Ermolieva, T., Havlík, P., Mosnier, A., Obersteiner, M., Yermoliev, Y., 2013. Dynamic recursive procedure for downscaling land cover changes from GLOBIOM model. Interim Reprot IR-13-005. International Institute for Applied Systems Analysis, IIASA, Laxenburg, Austria. Ewert, F., Rounsevell, M.D.A., Reginster, I., Metzger, M.J., Leemans, R., 2005. Future scenarios of European agricultural land use: I. Estimating changes in crop productivity. Agr. Ecosyst. Environ. 107, 101–116. Ewert, F., Porter, J.R., Rounsevell, M.D.A., 2007. Crop models, CO2, and climate change. Science 315, 459–460. Fitzherbert, E.B., Struebig, M.J., Morel, A., Danielsen, F., Brühl, C.A., Donald, P.F., Phalan, B., 2008. How will oil palm expansion affect biodiversity? Trends Ecol. Evol. 23, 538–545. Foody, G.M., 2002. Status of land cover accuracy assessment. Remote Sens. Environ. 80, 185–201. Goldstein, J.H., Caldarone, G., Duarte, T.K., Ennaanay, D., Hannahs, N., Mendoza, G., Polasky, S., Wolny, S., Daily, G.C., 2012. Integrating ecosystem-service tradeoffs into land-use decisions. P. Natl. Acad. Sci. USA 109, 7565–7570. Havlík, P., Schneider, U., Schmid, E., Bottcher, H., Fritz, S., Skalsky, R., Aoki, K., Decara, S., Kindermann, G., Kraxner, F., Leduc, S., McCallum, I., Mosnier, A., Sauer, T., Obersteiner, M., 2011. Global land use implications of firest and second generation biofuel targets. Energ. Policy 39, 5690–5702. Heubes, J., Schmidt, M., Stuch, B., Márquez, J.R.G., Wittig, R., Zizka, G., Thiombiano, A., Sinsin, B., Schaldach, R., Hahn, K., 2013. The projected impact of climate and land use change on plant diversity: an example from West Africa. J. Arid Environ. 96, 48–54. Huang, B., Xie, C., Tay, R., Wu, B., 2009. Land-use-change modeling using unbalanced support-vector machines. Environ. Plann. B 36, 398–416. Ian, J., Harwood, A.R., Mace, G.M., Watson, R.T., Abson, D.J., Andrews, B., Binner, A., Crowe, A., Day, B.H., Dugdale, S., Fezzi, C., Foden, J., Hadley, D., Haines-Young, R., Hulme, M., Kontoleon, A., Lovett, A.A., Munday, P., Pascual, U., Paterson, J., Perino, G., Sen, A., Siriwardena, G., van Soest, D., Termansen, M., 2013. Bring ecosystem services into economic decision-making: land use in the United Kingdom. Science 341, 45–50. IIASA, 2012. Supplementary note for the SSP data base. International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria. . Kriegler, E., O’Neill, B.C., Hallegatte, S., Kram, T., Lempert, R.J., Moss, R.H., Wilbanks, T., 2012. The need for and use of socioeconomic scenarios for climate change analysis: a new approach based on shared socio-economic pathways. Global Environ. Chang. 22, 807–822. Lambin, E.F., 1997. Modelling and monitoring land-cover change processes in tropical regions. Prog. Phys. Geogr. 21, 375–393. Lambin, E.F., Meyfroidt, P., 2011. Global land use change, economic globalization, and the looming land scarcity. P. Natl. Acad. Sci. USA 108, 3465–3472. Landis, J.R., Koch, G.G., 1977. The measurement of observer agreement for categorical data. Biometrics 33, 159–174. Li, R., Guan, Q., Merchant, J., 2012. A geospatial modeling framework for assessing biofuels-related land-use and land-cover change. Agr. Ecosyst. Environ. 161, 17–26. Liu, J., Williams, J.R., Zehnder, A.J.B., Hong, Y., 2007. GEPIC—modeling wheat yield and crop water productivity with high resolution on a global scale. Agric. Syst. 94, 478–493. Liu, R., Zhang, K., Zhang, Z., Borthwick, A.G.L., 2014. Land-use suitability analysis for urban development in Beijing. J. Environ. Manage. 145, 170–179. Lotze-Campen, H., Popp, A., Beringer, T., Müller, C., Bondeau, A., Rost, S., Lucht, W., 2010. Scenarios of global bioenergy production: the trade-offs between agricultural expansion, intensification and trade. Ecol. Model. 221, 2188–2196.

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007

X. Zhang et al. / Physics and Chemistry of the Earth xxx (2015) xxx–xxx Mancosu, E., Gago-Silva, A., Barbosa, A., De Bono, A., Ivanov, E., Lehmann, A., Fons, J., 2015. Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy 46, 26–36. Monserud, R.A., Leemans, R., 1992. Comparing global vegetation maps with the Kappa statistic. Ecol. Model. 62, 275–293. Moss, R., Edmonds, J., Hibbard, K., Manning, M., Rose, S., van Vuuren, D., Carter, T., Emori, S., Kainuma, M., Kram, T., Meehl, G., Mitchell, J., Nakicenovic, N., Riahi, K., Smith, S., Stouffer, R., Thomson, A., Weyant, J., Wilbanks, T., 2010. The next generation of scenarios for climate change research and assessment. Nature 463, 747–756. O’Neill, B.C., Carter, T.R., Ebi, K.L., Edmonds, J., Hallegatte, S., Kemp-Benedict, E., Kriegler, E., Mearns, L., Moss, R., Riahi, K., van Ruijven, B., van Vuuren, D., 2012. Meeting Report of the Workshop on the Nature and Use of New Socioeconomic Pathways for Climate Change Research, National Center for Atmospheric Research, Boulder, CO, November 2–4. Pérez-Vega, A., Mas, J.F., Ligmann-Zielinska, A., 2012. Comparing two approaches to land use/cover change modeling and their implications for the assessment of biodiversity loss in a deciduous tropical forest. Environ. Modell. Softw. 19, 11– 23. Polasky, S., Nelson, E., Pennington, D., Johnson, K.A., 2011. The impact of land-use change on Ecosystem service, biodiversity and returns to landowners: A case study in the state of Minnesota. Environ. Resour. Econ. 48, 219–242. Qi, S., Luo, F., 2005. Water environmental degradation of the Heihe River Basin in arid northwestern China. Environ. Monit. Assess. 108, 205–215. Qi, S., Luo, F., 2007. Environmental degradation problems in the Heihe River Basin, northwest China. Water Environ. J. 21, 142–148. Qi, S., Wang, T., Feng, J., 2003. Classification of Land Degradation in the Heihe River Basin, Northwestern China. Ecol. Environ. 12, 427–430 (in Chinese). Qi, S., Luo, F., Xiao, H., 2006. Land use changes and its environmental impact in arid inland river basins: a case from the Heihe River Basin in northwestern China. Ecol. Environ. 15, 757–760 (in Chinese). Ramankutty, N., Foley, J.A., 1999. Estimating historical changes in global land cover: croplands from 1700–1992. Global Biogeochem. Cy. 13, 997–1027.

9

Rounsevell, M.D.A., Reginster, I., Araújo, M.B., Carter, T.R., Dendoncker, N., Ewert, F., House, J.I., Kankaanpää, S., Leemans, R., Metzger, M.J., Schmit, C., Smith, P., Tuck, G., 2006. A coherent set of future land use change scenarios for Europe. Agr. Ecosyst. Environ. 114, 57–68. Schmidt, M., Thiombiano, A., Zizka, A., König, K., Brunken, U., Zizka, G., 2011. Patterns of plant functional traits in the biogeography of West African grasses (Poaceae). Afr. J. Ecol. 49 (4), 490–500. Shannon, C.E., 1948. A mathematical theory of communication. Bell Syst. Technol. J. 27, 379–423. Sun, Z.Q., Sun, Z.G., Deng, X.D., 2011. Ecological degradation process and ecological conservation strategies in Ejina Oasis. J. Arid Land Resour. Environ. 25, 39–45 (in Chinese). Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2012. An overview of cmip5 and the experiment design. B. Am. Meteorol. Soc. 93, 485–498. Turner, B.L., Lambin, E.F., Reenberg, A., 2007. The emergence of land change science for global environmental change and sustainability. P. Natl. Acad. Sci. USA 104, 20666–20671. van Vuuren, D., Edmonds, J., Kainuma, M., Riahi, K., Thomson, A., Hibbard, K., Hurtt, G., Kram, T., Krey, V., Lamarque, J.-F., Masui, T., Meinshausen, M., Nakicenovic, N., Smith, S., Rose, S., 2011. The representative concentration pathways: an overview. Clim. Change 109, 5–31. van Vuuren, D.P., Riahi, K., Moss, R., Edmonds, J., Thomson, A., Nakicenovic, N., Kram, T., Berkhout, F., Swart, R., Janetos, A., Rose, S.K., Arnell, N., 2012. A proposal for a new scenario framework to support research and assessment in different climate research communities. Global Environ. Chang. 22, 21–35. Zang, C.F., Liu, J., Van der Velde, M., Kraxner, F., 2012. Assessment of spatial and temporal patterns of green and blue water flows under natural conditions in inland river basins in Northwest China. Hydrol. Earth Syst. Sci. 16, 2859–2870. Zeng, Z., Liu, J., Koeneman, P.H., Zarate, E., Hoekstra, A.Y., 2012. Assessing water footprint at river basin level: a case study for the Heihe River Basin in northwest China. Hydro. Earth Syst. Sci. 16, 2771–2787. Zhang, H., Zhang, B., Shi, H., 2003. Studies on the land use and land cover change in arid regions. J. Arid Land Resour. Environ. 17, 49–54 (in Chinese).

Please cite this article in press as: Zhang, X., et al. Recursive cross-entropy downscaling model for spatially explicit future land uses: A case study of the Heihe River Basin. J. Phys. Chem. Earth (2015), http://dx.doi.org/10.1016/j.pce.2015.05.007