- Email: [email protected]
Modeling the wetland restorability based on natural and anthropogenic impacts in Sanjiang Plain, China
Ecological Indicators 91 (2018) 429–438
Contents lists available at ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
Original Articles
Modeling the wetland restorability based on natural and anthropogenic impacts in Sanjiang Plain, China ⁎
Yi Qu, Chunyu Luo, Hongqiang Zhang, Hongwei Ni , Nan Xu
T
⁎
National and Local Joint Laboratory of Wetland and Ecological Conservation, Institute of Natural Resources and Ecology, Heilongjiang Academy of Sciences, Harbin 150040, China
A R T I C LE I N FO
A B S T R A C T
Keywords: Wetland restoration planning Restorability Prioritization Agricultural landscape GIS
Wetlands are reducing dramatically due to rapid agricultural occupation and urbanization, resulting in severe ecosystem degradation and environmental deterioration. Comprehensive wetland restoration planning that fully considers ecosystem-level processes has the potential to improve restoration efficiency and regional sustainability. Most evaluating criteria of restorability are confined to the natural impacts on the suitability of successful wetland restoration. However, anthropogenic impacts received little attention; especially in areas where most wetlands had been encroached by intensively used farmlands. Utilization intensity is also an important factor that influences the restorability of wetlands. Here, we propose a GIS-based Restorability Index (RESI) model to evaluate the wetland restorability of the Sanjiang Plain, which is the largest marsh area of China as well as the largest grain production base. In the first step, eight criteria including natural conditions (stream order, overland flow length, saturation index, and soil characteristics) and anthropogenic impacts (land use, reclamation history, density of grain yield, and power density of agricultural machinery) were selected as influence factors of wetland restoration. For the second step, the RESI value was calculated for each grid cell via integration of spatially quantified criteria and these RESI values were classified into five levels to prioritize wetland restoration implementation. Finally, we designed a restoration plan according to the results of our restorability analysis and the requirements of regional sustainable development. This model offers a valuable tool for priority ranking of wetland restoration implementations in the agricultural landscape in a spatially explicit way.
1. Introduction Substantial areas of wetlands have been degraded or have disappeared during the agricultural and urban expansion of recent years. Many wetland restoration scholars are increasingly focusing on efforts of planning wetland restoration at a landscape scale since a larger scale would be conducive to the integrity of wetland restoration (MorenoMateos and Comin, 2010; Liu et al., 2016). Many have viewed large scale wetland restoration planning as an effective way to improve water quality and regional security (Crumpton, 2001; Zedler, 2003; Verhoeven et al., 2006). In agricultural landscapes, the peak flood period could be reduced by 50% if the watershed contains 5%-10% wetlands (DeLaney, 1995). A GIS-based model is effective in producing a large-scale wetland restoration plan. Dale and Siobhan (2005) used this method to evaluate the suitability of wetland restoration for the Cuyahoga River at watershed scale, providing spatially explicit guidance for wetland restoration efforts. Francisco et al. (2014) prioritized wetland restoration and construction with the explicit purpose of water quality improvement in agricultural watersheds, utilizing a similar ⁎
Corresponding authors. E-mail addresses: [email protected] (H. Ni), [email protected] (N. Xu).
https://doi.org/10.1016/j.ecolind.2018.04.008 Received 26 October 2017; Received in revised form 31 March 2018; Accepted 4 April 2018 1470-160X/ © 2018 Elsevier Ltd. All rights reserved.
method. Wetland restoration planning to systematically recover wetlands at the watershed and landscape scale provides the potential for restoring ecological processes that help to maintain the stability of whole regions (Dale and Siobhan, 2005). The likelihood of wetland restoration for sustainable long-term projects in an agricultural landscape not only involves natural factors, but also relates to factors in human disturbance. Natural factors such as hydrology, topography, soils, and geomorphology provide templates for wetland development (Bedford, 1996; Peng et al., 2010; Wang et al., 2011; Patenaude et al., 2015). These are the triggers of wetland formation and restoration, which are positively related to the likelihood of restoration success. However, anthropogenic impacts such as land use, reclamation history, power density of agricultural machinery, total fertilization amount, and farming times per year are negative impedances for the probability of successful restoration (Hatvany, 2009), since these activities affect the soil seed bank, consequentially reducing the probability of wetland restoration (Hong et al., 2012; Wang et al., 2015). All these criteria should be incorporated into the GIS model to integrate landscape variables that impact wetland biological and
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
reclaimed farmlands for many years is much more costly and complex than restoration of weeds formed by degradation of wetlands. In this study, reclamation history, grain yield density, and power density of agricultural machinery were selected as intensity criteria, affecting wetland restoration. These factors are directly related to the seed bank in the soil, which determines the degree of wetland restoration under natural conditions.
biogeochemical characteristics and sequentially prioritize implementations of wetland restoration. Our research develops a GIS-based wetland restorability model to evaluate the probability of successful wetland restoration in the agricultural landscape of the Sanjiang Plain (China). This study assumes that anthropogenic also have impacts on the potential of wetland restoration and aims to assess the restorability from both natural and anthropogenic perspectives and then prioritize the restoration implementations and actions according to modeled restorability and requirements of regional sustainability development. Our results will guide adjustments of human activities to accelerate the restoration process. If areas with high restorability were successfully identified and restored, the efficiency for wetland restoration would increase greatly.
3.2. Measurement and unification of criteria Topographic saturation, stream order, and overland flow distance were generated in ArcGIS from the digital elevation model (DEM) data which was obtained from the National Fundamental Geographic Information System (http://nfgis.nsdi.gov.cn/). The saturation index was calculated using slope and flow accumulation as follows:
2. Study area
SI = ln(α/tanβ)
The Sanjiang plain is located in the northeast of China. It is the largest freshwater swamp alluvial plain formed by three rivers. Furthermore, it is also the most important Chinese grain production area. Agricultural and construction development resulted in drastic decreases of wetlands, overall degrading wetland function during the last four decades. Remote sensing images of the Sanjiang Plain in different periods indicate that these natural wetlands had mostly been converted into farmlands and landscape fragmentation was obviously increased (Fig. 1). The widespread loss and degradation of wetlands resulted in a series of ecological and environmental problems such as massive floods, soil pollution, and shortage of water resources. Key areas of lost and degraded wetlands need to be identified and their restoration needs to be prioritized to provide references for a regional wetland restoration plan.
(1)
where SI represents the saturation index without unit, α indicates the area of up-slope drainage, and β is the local slope (Beven and Kirkby, 1979). The flow length was generated via the Flow length module of ArcGIS, which is the sum distance of each grid to the next down-stream grid. Stream orders and corresponding sub-watersheds were generated based on flow accumulation and the Strahler Stream Order. Soils were classified into three types: hydromorphic soil, high-water-content soil, and low/no-water-content soil. Each class was assigned a value representing the prospective contribution to wetland restoration. Land use/cover was scaled from 0 to 10 according to the restoration complexity of each type. The reclamation history was generated via comparison among 1995, 2000, 2005, 2010 and 2015 land use maps that were provided by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC). Grain yield density and power density of agricultural machinery were queried from statistical yearbooks that were obtained from statistical department of different counties. These factors were all numerical values that could be graded. For all factors, a unified and standardized measurement system was necessary to evaluate wetland restorability. The boundary values of each factor were determined by overall consideration of the relative importance of different factors from related publication (Dale and Siobhan, 2005; Ouyang et al., 2011) and classification methods (Natural Break Jenks). According to the effect-level of different intervals per factor, six class values were assigned to each level according to decreasing restorability (Table. 1). Level 1 indicates high restorability; level 2 indicates moderately high restorability; level 3 indicates intermediate restorability; level 4 indicates moderately low restorability; and level 5 indicates low restorability. Level 6 was excluded from the plan, as it is unsuitable for restoration. The maps were prepared in a rasterized format for subsequent analysis.
3. Methods Based on previous studies on wetland restoration (Dale and Siobhan, 2005; Ouyang et al., 2011) and the current situation of agricultural encroachment in the Sanjiang Plain, we identified eight criteria for successful wetland restoration as variables. These include: topographic saturation, soil characteristics, stream order, overland flow distance, land use, reclamation history, grain yield density, and power density of agricultural machinery. We integrated these criteria into GISbased model to produce a restorability index (RESI) to assess the likelihood of restoration success. Then, RESI was classified to prioritize wetland restoration implementations and actions. Finally, we designed a restoration plan according to the results of our RESI analysis and the requirements of regional sustainable development (Fig. 2). 3.1. Criteria influencing restorability Criteria related to the wetland restoration model were selected based on two aspects: natural impacts and anthropogenic impacts. All used criteria were highly related to the probability of successful wetland restoration. It should be noted that these criteria are salient for wetland restoration under an agricultural context, but they are not all inclusive. Soil characteristics and topographic saturation are the most representative physical parameters, forming wetlands. The river order and overland flow distance are neighborhood parameters forming the landscape context for a given wetland (Dale and Siobhan, 2005). These perimeters cover most of the natural impacts relative to the wetland restoration success. Land use/cover and land use intensity are the main impacts of human activities. They are highly related to the successful restoration of wetlands because land use types reflect the opportunity costs and land use intensity reflect the implementation costs that are affected by the probability and complexity of successful restoration (Hatvany, 2009; Wang et al., 2015; Yang et al., 2016). For example, restoration of
3.3. Assignment of factor weights To build the restorability model, it is necessary to identify the relative weights of factors. Principal component analysis (PCA) is a statistical procedure to convert a set of possibly correlated variables into a set of values of linearly uncorrelated variables and it is also useful in determining relative importance of factors (Benasseni, 2010; Randjelovic et al., 2013). The quantitative process was conducted in SPSS version 15.0 (Chicago, IL, USA) as follows (Zhang and Dong, 2004). Firstly, we extracted unified values of identified factors to randomly created 1399 points and used these values as inputs of PCA in SPSS. Six principle components (PCs) were extracted to reach the target of exceeding 80% of cumulative contributions (Joliffe and Morgan, 1992). PCs are uncorrelated variables that are obtained by multiplying the original correlated factors and by the eigenvectors (list of coefficients). 430
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 1. Changes of wetland resources in the study area.
431
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 2. Research methods and analytical framework.
been applied to evaluate the restoration suitability of wetlands to identify the spatial distribution of most suitable areas that should be recovered. The restorability modeling techniques are part of the multicriteria evaluation (MCE) framework (Carver, 1991a,b), which was originally used to analyze planning and policy (Bethel et al., 2014). Our model employed linear-weighted summation, which is the simplest technique in the general class. In this study, RESI was calculated according to corresponding weights derived via PCA (Eq. (4)).
Then, component loadings (CLs) were generated by projections of the original variables on the subspace of the PCs. Next, coefficients of PCs to express each factor were calculated by weighted mean of CLs and eigenvalues of PCs. The coefficients of PCs are defined as: n
Ci =
∑
CLj × λj (2)
j=1
where C is the coefficient of PC, CL is the component loading,λis the eigenvalue, i is the number of factors, j is the number of PCs, and n is the total number of factors. Finally, the weights of factors were obtained by weighted mean of coefficients and cumulative contribution rates. The weights of factors are defined as:
8
RESI =
∑
(4)
where RESI is the Restorability Index; wi and fi represent the weights and unified ranking values of factor i, respectively. Each criterion was converted into raster format and reclassified in ArcGIS software. All raster layers were summed based on Eq. (4) to produce a map of RESI, thus revealing the spatial characteristic of restorability. The continuous values of RESI were classified via the Natural Break (Jenks) method (Jenks, 1967), which is a classification method identifying break points by grouping similar values and maximizing differences of two classes (Economic and Social Research Institute, 2007). The RESI values were classified into five classes to facilitate the differentiation of ranks from both perspectives of natural condition and human disturbance, utilizing this method, which can be described as follows: RESI 5, most suitable natural conditions and lightly disturbed by human activities; RESI 4, highly suitable natural conditions and lightly disturbed by human activities; RESI 3, moderately suitable natural conditions and moderately disturbed by human
Cj × Rj (3)
j= 1
wi fi
i=1
n
Wi =
∑
where W is the weight of each factor, C is the coefficient of PC, R is the rate of cumulative contribution, and i, j and n have the same meanings as those in Eq. (2). 3.4. RESI calculation and classification RESI is a numerical linear index calculated by the restorability model and it represents the likelihood of successful restoration for a given degraded wetland. RESI translates existing knowledge of wetland restoration requirements into standard, quantitative measures of both natural environment and human impacts. The restorability model had Table 1 Normalized scores of restorability evaluation criteria. Criteria
Excluded 0
Level 5 2
Level 4 4
Level 3 6
Level 2 8
Level 1 10
Natural impacts
Topographic saturation Soil characteristic Stream order Overland flow length (km)
—— —— Others ——
4.4–10.9 Others 4 35.4–54.9
10.9–15.7 —— —— 24.0–35.4
15.7–23.5 Hydric inclusion 2,3 15.0–24.0
2.50–31.4 —— —— 6.4–15.0
31.4–51.8 Hydric soil 1 0–6.4
Anthropogenic impacts
Land use
Developed areas and forests
Bare lands
Farmlands
Bottomlands
Grass
Water and wetlands
Reclamation history (years) Grain yield density (ton/km2) Power density of agricultural machinery (kilowatt/km2)
—— —— ——
15–20 284.9–409.5 185.3–300.8
10–15 214.3–284.9 127.6–185.3
5–10 161.7–214.3 99.7–127.6
1–5 106.1–161.7 65.9–99.7
0–1 26.5–106.1 47.0–65.9
432
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
activities; RESI 2, least suitable natural conditions and highly disturbed by human activities; RESI 1, unsuitable natural conditions and highly disturbed by human activities. RESI 5 and 4 represent degraded or completely lost wetlands with high probability of successful restoration.
Table 2 Weights of assignment resulted from PCA method.
3.5. Sensitivity analysis Sensitivity analysis is an effective method to determine the most sensitive determinants of wetland restorability by varying values of each component. All factors were selected as variables for the sensitivity analysis. Each factor has current values as the baseline measurement. The sensitivity analysis was processed by sequentially increasing or decreasing the values by 20% of their present baseline values per step, while others remained unchanged. Seven scenarios per factor were generated in a process to identify the most sensitive factors. The number of planning units with very high restorability (RESI 5) was selected as the dependent variable to detect the most sensitive influential impact. This analysis aided the identification of determinants that lead to significant variations of the dependent variable via small value amendments.
Primary criteria
Sub-criteria
Weights of subcriteria
Natural impacts
(1) (2) (3) (4)
0.0509 0.2304 0.0872 0.3063
Anthropogenic impacts
(5) Land use (6) Reclamation history (years) (7) Power density of agricultural machinery (8) Grain yield density
Topographic saturation Soil characteristics Stream order Overland flow distance (m)
0.0876 0.0819 0.0017 0.1539
Table 3 Area and proportion of each class of the Restorability Index values.
3.6. Restoration prioritizing and planning Although the restorability model provided bases to evaluate restoration suitability, implementing all the wetland restorations at the same time is not feasible due to limitations of resources and time. Thus, wetland restoration needs to be prioritized via considering the current situation of remaining wetlands. Except for four counties that are characterized by forest landscapes from past to present, thus being only little affected by floods, other counties with restoration potential were prioritized by a combination of vulnerability and restorability, represented by existing wetland ratio and percentage of high restorability occupation. The grades were prioritized according to the scatter plot of Fig. 3. Unsafe areas with low restorability and safe areas with high restorability were grouped into GradeⅠ; safe areas with low restorability and very safe area with high restorability were grouped into GradeⅡ; and very safe areas with low restorability were grouped into GradeIII.
Levels
Descriptions
Area (km2)
Percentage of the total area (%)
RESI RESI RESI RESI RESI
Most suitable Highly suitable Moderately suitable Least suitable Unsuitable
14,973 25,961 27,638 26,832 11,375
14 24 26 25 11
5 4 3 2 1
weights showed that, in all anthropogenic impacts, grain yield density has more significance in the restoration probability of degraded wetlands, while in all natural impacts, overland flow distance and soil character play important role in wetland restoration potential. 4.2. RESI levels and assessment According to the current land use pattern, the most suitable areas for wetland restoration cover 14,973 km2 (accounting for 14% of the whole study area). Adding the highly suitable areas increases the total extent to 40,934 km2, which corresponds to 38% of the total study area (Table. 3). These high restorability areas are mainly located in the conjunction zones of different rivers and around the Lake Xingkai, including the administrative counties of Luobei, Tongjiang, Fuyuan, Suibin, Fujin, Baoqing, Minshan, and Hulin (Fig. 4). All these areas have not been encroached by farmlands for a long time (within the consuming limitation of soil seed banks) and have not been intensively utilized. In Fujin County, the wetlands that had been converted into farmlands for less than five years have a very high probability of recovery. The RESI values revealed a larger proportion (64%) of the degraded area to be suitable for wetland restoration, indicating great potential for the recovery of ecological patterns and species habitats. In the Sanjiang Plain, remaining wetlands occupy 8532 km2, covering less than 14% of the total study area. These areas can be naturally recovered if they degraded or were polluted, since they are still within the wetland category. However, degraded wetlands that were converted into grasslands and farmlands require more efforts and the complexity of restoration depends on the time of reclamation, which is determined by the soil seed bank. Due to the age limit of the soil seed bank, farmlands that are reclaimed within 20 years were considered capable of being recovered without artificial cultivation (Wang et al., 2015). The extent of these farmlands has been defined as the effective restoration area (approximately 5777 km2), which is less costly to restore. Of the effective restoration area, about 39% fell into the most suitable category of restoration, 22% fell into a high suitability level, and 21% fell into a moderate level of suitability. Therefore, almost 85% of the lost or degraded wetlands can be recovered to their natural condition.
4. Results 4.1. Weights of criteria Weights of different criteria in the restorability model were generated via the PCA method and the results are presented in Table. 2. Due to extensive agricultural encroachment, substantially degraded wetland resulted from the high density of agricultural activities. The
4.3. Sensitivity analysis Fig. 3. Prioritization based on vulnerability and restorability.
The planning unit numbers of the most suitable sites for restoration 433
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 4. Spatial distribution characteristics of Restorability Index in the Sanjiang Plain.
434
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 5. Number of planning units of the most suitable restoration sites in 7 scenarios: 1, the specified factor reduced by 60% on the baseline value; 2, the specified factor reduced by 40% on the baseline value; 3, the specified factor reduced by 20% on the baseline value; 4, the baseline scenario; 5, the specified factor increased by 20% on the baseline value; 6, the specified factor increased by 40% on the baseline value; 7, the specified factor increased by 60% on the baseline value.
drastically increased, when overland flow length, soil characteristic, and grain yield density varied through scenarios 1–7 (Fig. 5), indicating that overland flow length, soil characteristic, and grain yield density are the most three sensitive factors for restoration of degraded or lost wetlands. Although overland flow length and soil characteristic are very hard and costly to be changed into suitable condition in restoration on the ground, they imply that areas with suitable condition both of these factors will give us quick responses when we implementing wetland restoration. Grain yield density is also sensitive to restoration, which indicates that wetlands with relatively low grain yield density would recover more rapidly. Power density of agricultural machinery has the lowest weight but it showed a sudden sensitivity increase in the last, which indicates that when the strength of reclamation is reduced to a certain degree the restoration will speed up. Land use is a little more sensitive than stream order, reclamation history and topographic saturation. The result revealed that, if the wetland had not converted into other land use types, natural factors can help them in their recovery. Furthermore, if the wetland had been converted into farmland, land use and power density of agricultural machinery became the main impact, influencing the probability of successful wetland restoration. This
indicates that if a restoration plan were in place for these farmlands, the power agricultural machinery on the lands have to be controlled. 4.4. Restorability prioritization and restoration plan Data from existing wetland and degraded wetland revealed that 17 counties now have more than 5% of wetlands occupation and are in a safe situation (Table. 4). Wetlands in other counties have the potential for restoration and may improve the ecological security in these counties. After restoration (except for four excluded counties), all counties would be in the safe category. In our prioritization results, five counties were classified into the first grade of restoration action (three in safe wetland ratio and in high restorability section, two in unsafe wetland ratio section); seven counties (three in safe wetland ratio and in low restorability section, four in very safe wetland ratio and high restorability section) were classified into the second grade of restoration action; and seven counties (in very safe wetland ratio and low restorability section) were classified as the last grade of restoration action (Fig. 6). The first grade counties in unsafe section underwent severe wetland degradation and are currently threatened by natural disasters. The first grade counties in safe section have very high restorability, which means wetlands in these counties are easy to recover. The second grade in safe section is relatively hard to recover but easy to become unsafe. The second grade in very safe section is relatively simple to recover. The wetlands in the last grade are hard to recover and they are currently in very safe situation; therefore, there is no need for immediate restoration action. If degraded or lost wetlands were recovered, 11 counties would have more than 10% of wetlands and others would have a wetland ratio between 5% and 10%, indicating the reduction of passive impacts according to the conclusion of DeLaney.
Table 4 Proportion of existing wetland and degraded wetland in the Sanjiang Plain. County name
Area of existing wetlands (km2)
Area of degraded wetlands (km2)
Ratio of existing wetlands (%)
Ratio of degraded wetlands (%)
Wetland ratio after restoration (%)
Fuyuan Tongjiang Mishan Suibin Hulin Jiamusi Raohe Fujin Baoqing Luobei Tangyuan Huachuan Hegang Jidong Yilan Boli Huanan Qitaihe Youyi
2522.65 1993.81 2188.87 623.53 1482.45 245.31 841.27 1023.97 1128.40 729.69 354.20 191.56 287.29 200.73 275.75 255.71 251.68 74.53 25.97
507.94 715.65 388.57 220.22 719.26 49.95 193.65 1906.75 188.85 256.81 70.06 81.80 27.42 3.30 178.69 101.23 92.27 16.83 119.45
40.27 32.31 28.34 18.61 15.89 13.03 12.75 12.46 11.37 10.82 10.24 8.50 6.30 6.21 5.99 5.73 5.69 4.27 1.54
8.11 11.60 5.03 6.57 7.71 2.65 2.94 23.20 1.90 3.81 2.02 3.63 0.60 0.10 3.88 2.27 2.09 0.96 7.08
48.38 43.90 33.37 25.18 23.60 15.69 15.69 35.65 13.27 14.63 12.26 12.13 6.91 6.31 9.88 8.00 7.78 5.24 8.62
5. Discussion Data of spatial changes showed severe degradation of wetlands in the study area and wetland restoration is urgent. This study identified natural and anthropogenic factors and their weights in wetland restoration. It also successfully assessed restorability and restoration prioritization of the study area. The weights identified showed that part but not all anthropogenic factors play important roles in wetland restoration and some factors (both natural and anthropogenic) that were expected to be important have not been supported by data. In this section, we will discuss the necessity of wetland restoration, some 435
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 6. Prioritization of implementing actions for all counties in the Sanjiang Plain.
436
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
restoration plan. With regard to the implementation cost, the management cost is not the only aspect we need to take into account: opportunity cost also deserves consideration (Plantinga et al., 2001; Naidoo et al., 2006). Furthermore, for the technologies of wetland restoration, many studies tend to be site specific (Millard et al., 2013; Kershner, 1997) and mainly focus on one single influencing factor. Studies that unify the sites into the whole wetland system require further research.
issues in modelling and implementations, and our further researches. 5.1. Necessity for large extent restoration During the past four decades, the area of intact wetlands in the Sanjiang Plain drastically decreased due to irrational use of wetlands (such as burning and draining) for agricultural expansion. Their extent reduced by more than 75% of the original area. The habitats and corridors for species have been fragmented which significantly influence species’ habitats and gene flow (Naugle et al., 2001; Ceresa et al., 2015). Existing research indicated that if watersheds comprised 5–10% of wetlands, the peak flood period would reduce significantly. Although extant wetlands occupied 7.85% of the study area, these marshes and water were distributed unevenly within the watersheds of the whole plain, particularly in the densely constructed and agriculture areas. The wetlands in statuary nature reserves cover only 3.88% of the total study area and about half of these reserves had been replaced by agriculture. If the wetlands outside the conservation reserves disappeared or were to degrade, the entire region would be out of ecological security control and become unstable. If the recommended restoration plan was implemented by restoring lost and degraded wetlands, adding existing nature protected areas, the total extent of wetlands would increase from 7.85% to 13.17% of the study area. This would increase the area and connectivity of habitats for local species and thus improving the stability and biodiversity of the entire study area.
5.4. Further research To improve the accuracy of the model, we need to incorporate more factors influencing wetland restoration in our continuing research. Beyond the factors used for this study, the diversity of seed bank, microbial environment, total amount of fertilizer, population density, and other related determinants need to be considered. Based on the field sampling data, a regression correlation analysis could be completed to build a more reliable restorability model. Due to a lack of test plots for the control of wetland restoration in the Sanjiang Plain, we have to resort to natural reserves that have projects of returning farmlands to wetlands to seek for the variation of wetland restoration by comparing restored areas with adjacent natural wetlands. We expect such a statistical method to generate more objective results and we will compare the results with outcomes of this study to analyse the rationality of both models.
5.2. Assessment of the model
6. Conclusions
Despite the advantages of RESI model technologies, random uncertainties in manual operation of data processing are inevitable; for example, vector format of the land use, interpreted by remote sensing images had an accuracy of 85%, which is short of the 100% target. A further problem is that DEM data and its ramifications were not up-todate and the resolution does not allow finer scale analyses. All these processes caused uncertainty in the results. To increase accuracy of the model, more up-to-date and reliable data relying on technology development are required. Our results of weights assigning showed that both natural and anthropogenic factors have impacts on wetland restoration. Among all identified factors, overland flow length, soil characteristics and grain yield density are more important and sensitive to restoration. However, topographic saturation was not assigned a high weight as expected and it is not sensitive to changes. This might have been due to the correlation between flow length and topographic saturation, or topographic saturation might have less influence on marshes than river and lake wetlands. For the anthropogenic factors, the intensity of land use is more important than the way of land use. These results give us enlightenment in further implementation of wetland restoration. Other factors such as diversity of seed banks, microbial environment, and total amount of fertilizer also impact strongly on wetland restorability (Wang et al., 2015). They should be included to conduct more detailed and comprehensive further research for wetland restoration.
In this study, we proposed a GIS-based model for evaluating the restorability of wetland in Sanjiang Plain, China. This model successfully identifies priority areas that have high potential for wetland restoration based on both natural and anthropogenic impacts. It provides a useful tool for prioritization of wetland restoration implementations in the agricultural landscape in a spatially explicit way. In further research, we will incorporate more factors that influence wetland restoration to improve the accuracy of the model and apply a regression correlation analysis with more field sampling data to improve the reliability. In general, the model proposed here allows for wetland managers and planners to effectively identify wetlands that should be given priority for restoration in agricultural landscape and at the same time reflects the relative sensitivity of different influencing factors. This model is adaptable to other regions and its resulted wetland restoration plan will guide adjustments of human activities to accelerate the restoration process and thus improve the restoration efficiency greatly. Acknowledgements We thank the China Wetland Scientific Database for the provided wetland distribution data set. Funding for data surveys and collection came from the National Natural Science Foundation of China (Project Number 41501583). The English writing of the manuscript was carefully edited by Mr. Thomas S., Professional Editor, Friedrich-Alexander Universität Erlangen.
5.3. Implementing the plan
References
To implement wetland restoration on a regional scale, we have to consider three problems: the first problem is management and control, i.e. the type of policy to incorporate the degraded or lost wetlands into a system; the second problem is the cost of implementing restoration and the balance between economic development and natural conservation; finally, the problem of restoration technologies. For management and control, incorporating restoration into existing wetland conservation networks is a feasible method because strict restriction of reserves greatly contributes to the restoration process. Moreover, many areas within nature reserves have undergone wetland degradation, up to wetland loss and they are good starting points for a step-by-step
Bedford, B.A., 1996. The need to define hydrologic equivalence at the landscape scale for freshwater wetland mitigation. Ecol. Appl. 6, 57–68. http://dx.doi.org/10.2307/ 2269552. Benasseni, J., 2010. On the algebraic relation between principal components corresponding to two different sets of weights. Linear Multilinear Algebra 58 (4), 425–429. Bethel, M.B., Brien, L.F., Esposito, M.M., Miller, C.T., Buras, H.S., Laska, S.B., Philippe, R., Peterson, K.J., Richards, C.P., 2014. Sci-TEK: A GIS-based multidisciplinary method for incorporating traditional ecological knowledge into Louisiana's coastal restoration decision-making processes. J. Coastal Res. 297 (5), 1081–1099. http://dx.doi.org/10. 2112/JCOASTRES-D-13-00214.1. Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area model of
437
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
s11273-013-9303-9. Moreno-Mateos, D., Comin, F.A., 2010. Integrating objectives and scales for planning and implementing wetland restoration and creation in agricultural landscapes. J. Environ. Manage. 91, 2087–2095. http://dx.doi.org/10.1016/j.jenvman.2010.06.002. Naidoo, R., Balmford, A., Polasky, S., Rouget, M., 2006. Integrating economic costs into conservation planning. Trends Ecol. Evol. 21 681e687. Naugle, D.E., Johnson, R.R., Estey, M.E., Higgins, K.F., 2001. A landscape approach to conserving wetland bird habitat in the prairie pothole region of eastern South Dakota. Wetlands 21 (1), 1–17. http://dx.doi.org/10.1672/0277-5212(2000) 020[0588:ALATCW]2.0.CO. Ouyang, N.L., Lu, S.L., Wu, B.F., Zhu, J.J., Wang, H., 2011. Wetland restoration suitability evaluation at the watershed scale-A case study in upstream of the Yongdinghe river. Procedia Environ. Sci. 10, 1926–1932. Patenaude, T., Smith, A.C., Fahrig, L., 2015. Disentangling the effects of wetland cover and urban development on quality of remaining wetlands. Urban Ecosyst. 18 (3), 663–684. http://dx.doi.org/10.1007/s11252-015-0440-1. Peng, Y.L., Wu, N., Gao, X.F., Fang, Z.Q., Xiao, W.Y., 2010. Soil seed banks in lakeshore wetlands: relation to the extant vegetation. Pol. J. Ecol. 58 (3), 449–457. Plantinga, A.J., Alig, R., Cheng, H.T., 2001. The supply of land for conservation uses: evidence from the conservation reserve program. Resour. Conserv. Recycl. 31 (3), 199–215. http://dx.doi.org/10.1016/S0921-3449(00)00085-9. Randjelovic, D., Stankovic, J., Jankovic-Milic, V., Stankovic, J., 2013. Weight coefficents determination based on parameters in factor analysis. Metal. Int. 18 (3), 128–131. Verhoeven, J.T.A., Arheimer, B., Yin, C., Hefting, M.M., 2006. Regional and global concerns over wetlands and water quality. Trends Ecol. Evol. 21, 96–103. http://dx.doi. org/10.1016/j.tree.2005.11.015. Wang, M.N., Wang, H., Qin, D.Y., Lu, C.Y., Li, Y.P., 2011. Modelling the artificial recharge of a wetland and its influence on regional hydrological process in China: a case study. Ecohydrology 4 (4), 589–596. http://dx.doi.org/10.1002/eco.175. Wang, G.D., Wang, M., Lu, X.G., Jiang, M., 2015. Effects of farming on the soil seed banks and wetland restoration potential in Sanjiang Plain, Northeastern China. Ecol. Eng. 77, 265–274. Yang, W.H., Liu, Y.B., Cutlac, M., Boxall, P., Weber, M., Bonnycastle, A., Gabor, S., 2016. Integrated economic-hydrologic modeling for examining cost-effectiveness of wetland restoration scenarios in a canadian prairie watershed. Wetlands 36 (3), 577–589. http://dx.doi.org/10.1007/s13157-016-0768-1. Zedler, J.B., 2003. Wetlands at your service: reducing impacts of agriculture at the watershed scale. Front. Ecol. Environ. 1, 65–72. http://dx.doi.org/10.2307/3868032. Zhang, W.T., Dong, W., 2004. SPSS Statistical Analysis Advantage Textbook. Higher Education Press, Beijing.
basin hydrology. Hydrol. Sci. Bull. 24, 43–69. Carver, S., 1991a. Site search and multi-criteria evaluation. Plann. Outlook 34 (1), 27–36. http://dx.doi.org/10.1080/00320719108711886. Carver, S., 1991b. Integrating multi-criteria evaluation with geographical information systems. Int. J. Geog. Inf.Sci. 5 (3), 321–339. http://dx.doi.org/10.1080/ 02693799108927858. Ceresa, F., Belda, E.J., Kvist, L., Rguibi-Idrissi, H., Monrós, J.S., 2015. Does fragmentation of wetlands affect gene flow in sympatric Acrocephalus warblers with different migration strategies? J. Avian Biol. 46 (6), 577–588. http://dx.doi.org/10.1111/jav. 00589. Crumpton, W.G., 2001. Using wetlands for water quality improvement in agricultural watersheds; the importance of a watershed scale approach. Water Sci. Technol. 44, 559–564 11804150. Dale, W., Siobhan, F., 2005. Modeling the suitability of wetland restoration potential at the watershed scale. Original Res. Art. Ecol. Eng. 24 (4), 359–377. http://dx.doi.org/ 10.1016/j.ecoleng.2005.01.012. DeLaney, T.A., 1995. Benefits to downstream flood attenuation and water quality as a result of constructed wetlands in agricultural landscapes. J. Soil Water Conserv. 50, 620–626. Economic and Social Research Institute (ESRI), 2007. ArcGIS Desktop Help 9.2, USA. Francisco, A.C., Ricardo, S., Nadia, D.C., Mercedes, G., Adriá, M., 2014. A protocol to prioritize wetland restoration and creation for water quality improvement in agricultural watersheds. Ecol. Eng. 66 (3), 10–18. http://dx.doi.org/10.1016/j.ecoleng. 2013.04.059. Hatvany, M.G., 2009. Wetlands and reclamation. Int. Encycl. Hum. Geogr. 241–246. Hong, J.M., Liu, S., Shi, G.P., Zhang, Y.Q., 2012. Soil seed bank techniques for restoring wetland vegetation diversity in Yeyahu Wetland, Beijing. Ecol. Eng. 42, 192–202. Jenks, G.F., 1967. The data model concept in statistical mapping. International Yearbook of Cartography, 7: 186e190. Joliffe, I.T., Morgan, B.J., 1992. Principal component analysis and exploratory factor analysis. Stat. Methods Med. Res. 1 (1), 69–95. Kershner, J.L., 1997. Setting riparian/aquatic restoration objectives within a watershed context. Restor. Ecol. 5 (4S), 15–24. http://dx.doi.org/10.1111/j.1526-100X.1997. 00015. Liu, G.D., Tian, K., Sun, J., Xiao, D., Yuan, X., 2016. Evaluating the effects of wetland restoration at the watershed scale in Northwest yunnan plateau, China. Wetlands 36 (1), 169–183. http://dx.doi.org/10.1007/s13157-015-0727-2. Millard, K., Redden, A.M., Webster, T., Stewart, H., 2013. Use of GIS and high resolution LiDAR in salt marsh restoration site suitability assessments in the upper Bay of Fundy, Canada. Wetlands Ecol. Manage. 21 (4), 243–262. http://dx.doi.org/10.1007/
438
Contents lists available at ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
Original Articles
Modeling the wetland restorability based on natural and anthropogenic impacts in Sanjiang Plain, China ⁎
Yi Qu, Chunyu Luo, Hongqiang Zhang, Hongwei Ni , Nan Xu
T
⁎
National and Local Joint Laboratory of Wetland and Ecological Conservation, Institute of Natural Resources and Ecology, Heilongjiang Academy of Sciences, Harbin 150040, China
A R T I C LE I N FO
A B S T R A C T
Keywords: Wetland restoration planning Restorability Prioritization Agricultural landscape GIS
Wetlands are reducing dramatically due to rapid agricultural occupation and urbanization, resulting in severe ecosystem degradation and environmental deterioration. Comprehensive wetland restoration planning that fully considers ecosystem-level processes has the potential to improve restoration efficiency and regional sustainability. Most evaluating criteria of restorability are confined to the natural impacts on the suitability of successful wetland restoration. However, anthropogenic impacts received little attention; especially in areas where most wetlands had been encroached by intensively used farmlands. Utilization intensity is also an important factor that influences the restorability of wetlands. Here, we propose a GIS-based Restorability Index (RESI) model to evaluate the wetland restorability of the Sanjiang Plain, which is the largest marsh area of China as well as the largest grain production base. In the first step, eight criteria including natural conditions (stream order, overland flow length, saturation index, and soil characteristics) and anthropogenic impacts (land use, reclamation history, density of grain yield, and power density of agricultural machinery) were selected as influence factors of wetland restoration. For the second step, the RESI value was calculated for each grid cell via integration of spatially quantified criteria and these RESI values were classified into five levels to prioritize wetland restoration implementation. Finally, we designed a restoration plan according to the results of our restorability analysis and the requirements of regional sustainable development. This model offers a valuable tool for priority ranking of wetland restoration implementations in the agricultural landscape in a spatially explicit way.
1. Introduction Substantial areas of wetlands have been degraded or have disappeared during the agricultural and urban expansion of recent years. Many wetland restoration scholars are increasingly focusing on efforts of planning wetland restoration at a landscape scale since a larger scale would be conducive to the integrity of wetland restoration (MorenoMateos and Comin, 2010; Liu et al., 2016). Many have viewed large scale wetland restoration planning as an effective way to improve water quality and regional security (Crumpton, 2001; Zedler, 2003; Verhoeven et al., 2006). In agricultural landscapes, the peak flood period could be reduced by 50% if the watershed contains 5%-10% wetlands (DeLaney, 1995). A GIS-based model is effective in producing a large-scale wetland restoration plan. Dale and Siobhan (2005) used this method to evaluate the suitability of wetland restoration for the Cuyahoga River at watershed scale, providing spatially explicit guidance for wetland restoration efforts. Francisco et al. (2014) prioritized wetland restoration and construction with the explicit purpose of water quality improvement in agricultural watersheds, utilizing a similar ⁎
Corresponding authors. E-mail addresses: [email protected] (H. Ni), [email protected] (N. Xu).
https://doi.org/10.1016/j.ecolind.2018.04.008 Received 26 October 2017; Received in revised form 31 March 2018; Accepted 4 April 2018 1470-160X/ © 2018 Elsevier Ltd. All rights reserved.
method. Wetland restoration planning to systematically recover wetlands at the watershed and landscape scale provides the potential for restoring ecological processes that help to maintain the stability of whole regions (Dale and Siobhan, 2005). The likelihood of wetland restoration for sustainable long-term projects in an agricultural landscape not only involves natural factors, but also relates to factors in human disturbance. Natural factors such as hydrology, topography, soils, and geomorphology provide templates for wetland development (Bedford, 1996; Peng et al., 2010; Wang et al., 2011; Patenaude et al., 2015). These are the triggers of wetland formation and restoration, which are positively related to the likelihood of restoration success. However, anthropogenic impacts such as land use, reclamation history, power density of agricultural machinery, total fertilization amount, and farming times per year are negative impedances for the probability of successful restoration (Hatvany, 2009), since these activities affect the soil seed bank, consequentially reducing the probability of wetland restoration (Hong et al., 2012; Wang et al., 2015). All these criteria should be incorporated into the GIS model to integrate landscape variables that impact wetland biological and
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
reclaimed farmlands for many years is much more costly and complex than restoration of weeds formed by degradation of wetlands. In this study, reclamation history, grain yield density, and power density of agricultural machinery were selected as intensity criteria, affecting wetland restoration. These factors are directly related to the seed bank in the soil, which determines the degree of wetland restoration under natural conditions.
biogeochemical characteristics and sequentially prioritize implementations of wetland restoration. Our research develops a GIS-based wetland restorability model to evaluate the probability of successful wetland restoration in the agricultural landscape of the Sanjiang Plain (China). This study assumes that anthropogenic also have impacts on the potential of wetland restoration and aims to assess the restorability from both natural and anthropogenic perspectives and then prioritize the restoration implementations and actions according to modeled restorability and requirements of regional sustainability development. Our results will guide adjustments of human activities to accelerate the restoration process. If areas with high restorability were successfully identified and restored, the efficiency for wetland restoration would increase greatly.
3.2. Measurement and unification of criteria Topographic saturation, stream order, and overland flow distance were generated in ArcGIS from the digital elevation model (DEM) data which was obtained from the National Fundamental Geographic Information System (http://nfgis.nsdi.gov.cn/). The saturation index was calculated using slope and flow accumulation as follows:
2. Study area
SI = ln(α/tanβ)
The Sanjiang plain is located in the northeast of China. It is the largest freshwater swamp alluvial plain formed by three rivers. Furthermore, it is also the most important Chinese grain production area. Agricultural and construction development resulted in drastic decreases of wetlands, overall degrading wetland function during the last four decades. Remote sensing images of the Sanjiang Plain in different periods indicate that these natural wetlands had mostly been converted into farmlands and landscape fragmentation was obviously increased (Fig. 1). The widespread loss and degradation of wetlands resulted in a series of ecological and environmental problems such as massive floods, soil pollution, and shortage of water resources. Key areas of lost and degraded wetlands need to be identified and their restoration needs to be prioritized to provide references for a regional wetland restoration plan.
(1)
where SI represents the saturation index without unit, α indicates the area of up-slope drainage, and β is the local slope (Beven and Kirkby, 1979). The flow length was generated via the Flow length module of ArcGIS, which is the sum distance of each grid to the next down-stream grid. Stream orders and corresponding sub-watersheds were generated based on flow accumulation and the Strahler Stream Order. Soils were classified into three types: hydromorphic soil, high-water-content soil, and low/no-water-content soil. Each class was assigned a value representing the prospective contribution to wetland restoration. Land use/cover was scaled from 0 to 10 according to the restoration complexity of each type. The reclamation history was generated via comparison among 1995, 2000, 2005, 2010 and 2015 land use maps that were provided by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC). Grain yield density and power density of agricultural machinery were queried from statistical yearbooks that were obtained from statistical department of different counties. These factors were all numerical values that could be graded. For all factors, a unified and standardized measurement system was necessary to evaluate wetland restorability. The boundary values of each factor were determined by overall consideration of the relative importance of different factors from related publication (Dale and Siobhan, 2005; Ouyang et al., 2011) and classification methods (Natural Break Jenks). According to the effect-level of different intervals per factor, six class values were assigned to each level according to decreasing restorability (Table. 1). Level 1 indicates high restorability; level 2 indicates moderately high restorability; level 3 indicates intermediate restorability; level 4 indicates moderately low restorability; and level 5 indicates low restorability. Level 6 was excluded from the plan, as it is unsuitable for restoration. The maps were prepared in a rasterized format for subsequent analysis.
3. Methods Based on previous studies on wetland restoration (Dale and Siobhan, 2005; Ouyang et al., 2011) and the current situation of agricultural encroachment in the Sanjiang Plain, we identified eight criteria for successful wetland restoration as variables. These include: topographic saturation, soil characteristics, stream order, overland flow distance, land use, reclamation history, grain yield density, and power density of agricultural machinery. We integrated these criteria into GISbased model to produce a restorability index (RESI) to assess the likelihood of restoration success. Then, RESI was classified to prioritize wetland restoration implementations and actions. Finally, we designed a restoration plan according to the results of our RESI analysis and the requirements of regional sustainable development (Fig. 2). 3.1. Criteria influencing restorability Criteria related to the wetland restoration model were selected based on two aspects: natural impacts and anthropogenic impacts. All used criteria were highly related to the probability of successful wetland restoration. It should be noted that these criteria are salient for wetland restoration under an agricultural context, but they are not all inclusive. Soil characteristics and topographic saturation are the most representative physical parameters, forming wetlands. The river order and overland flow distance are neighborhood parameters forming the landscape context for a given wetland (Dale and Siobhan, 2005). These perimeters cover most of the natural impacts relative to the wetland restoration success. Land use/cover and land use intensity are the main impacts of human activities. They are highly related to the successful restoration of wetlands because land use types reflect the opportunity costs and land use intensity reflect the implementation costs that are affected by the probability and complexity of successful restoration (Hatvany, 2009; Wang et al., 2015; Yang et al., 2016). For example, restoration of
3.3. Assignment of factor weights To build the restorability model, it is necessary to identify the relative weights of factors. Principal component analysis (PCA) is a statistical procedure to convert a set of possibly correlated variables into a set of values of linearly uncorrelated variables and it is also useful in determining relative importance of factors (Benasseni, 2010; Randjelovic et al., 2013). The quantitative process was conducted in SPSS version 15.0 (Chicago, IL, USA) as follows (Zhang and Dong, 2004). Firstly, we extracted unified values of identified factors to randomly created 1399 points and used these values as inputs of PCA in SPSS. Six principle components (PCs) were extracted to reach the target of exceeding 80% of cumulative contributions (Joliffe and Morgan, 1992). PCs are uncorrelated variables that are obtained by multiplying the original correlated factors and by the eigenvectors (list of coefficients). 430
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 1. Changes of wetland resources in the study area.
431
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 2. Research methods and analytical framework.
been applied to evaluate the restoration suitability of wetlands to identify the spatial distribution of most suitable areas that should be recovered. The restorability modeling techniques are part of the multicriteria evaluation (MCE) framework (Carver, 1991a,b), which was originally used to analyze planning and policy (Bethel et al., 2014). Our model employed linear-weighted summation, which is the simplest technique in the general class. In this study, RESI was calculated according to corresponding weights derived via PCA (Eq. (4)).
Then, component loadings (CLs) were generated by projections of the original variables on the subspace of the PCs. Next, coefficients of PCs to express each factor were calculated by weighted mean of CLs and eigenvalues of PCs. The coefficients of PCs are defined as: n
Ci =
∑
CLj × λj (2)
j=1
where C is the coefficient of PC, CL is the component loading,λis the eigenvalue, i is the number of factors, j is the number of PCs, and n is the total number of factors. Finally, the weights of factors were obtained by weighted mean of coefficients and cumulative contribution rates. The weights of factors are defined as:
8
RESI =
∑
(4)
where RESI is the Restorability Index; wi and fi represent the weights and unified ranking values of factor i, respectively. Each criterion was converted into raster format and reclassified in ArcGIS software. All raster layers were summed based on Eq. (4) to produce a map of RESI, thus revealing the spatial characteristic of restorability. The continuous values of RESI were classified via the Natural Break (Jenks) method (Jenks, 1967), which is a classification method identifying break points by grouping similar values and maximizing differences of two classes (Economic and Social Research Institute, 2007). The RESI values were classified into five classes to facilitate the differentiation of ranks from both perspectives of natural condition and human disturbance, utilizing this method, which can be described as follows: RESI 5, most suitable natural conditions and lightly disturbed by human activities; RESI 4, highly suitable natural conditions and lightly disturbed by human activities; RESI 3, moderately suitable natural conditions and moderately disturbed by human
Cj × Rj (3)
j= 1
wi fi
i=1
n
Wi =
∑
where W is the weight of each factor, C is the coefficient of PC, R is the rate of cumulative contribution, and i, j and n have the same meanings as those in Eq. (2). 3.4. RESI calculation and classification RESI is a numerical linear index calculated by the restorability model and it represents the likelihood of successful restoration for a given degraded wetland. RESI translates existing knowledge of wetland restoration requirements into standard, quantitative measures of both natural environment and human impacts. The restorability model had Table 1 Normalized scores of restorability evaluation criteria. Criteria
Excluded 0
Level 5 2
Level 4 4
Level 3 6
Level 2 8
Level 1 10
Natural impacts
Topographic saturation Soil characteristic Stream order Overland flow length (km)
—— —— Others ——
4.4–10.9 Others 4 35.4–54.9
10.9–15.7 —— —— 24.0–35.4
15.7–23.5 Hydric inclusion 2,3 15.0–24.0
2.50–31.4 —— —— 6.4–15.0
31.4–51.8 Hydric soil 1 0–6.4
Anthropogenic impacts
Land use
Developed areas and forests
Bare lands
Farmlands
Bottomlands
Grass
Water and wetlands
Reclamation history (years) Grain yield density (ton/km2) Power density of agricultural machinery (kilowatt/km2)
—— —— ——
15–20 284.9–409.5 185.3–300.8
10–15 214.3–284.9 127.6–185.3
5–10 161.7–214.3 99.7–127.6
1–5 106.1–161.7 65.9–99.7
0–1 26.5–106.1 47.0–65.9
432
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
activities; RESI 2, least suitable natural conditions and highly disturbed by human activities; RESI 1, unsuitable natural conditions and highly disturbed by human activities. RESI 5 and 4 represent degraded or completely lost wetlands with high probability of successful restoration.
Table 2 Weights of assignment resulted from PCA method.
3.5. Sensitivity analysis Sensitivity analysis is an effective method to determine the most sensitive determinants of wetland restorability by varying values of each component. All factors were selected as variables for the sensitivity analysis. Each factor has current values as the baseline measurement. The sensitivity analysis was processed by sequentially increasing or decreasing the values by 20% of their present baseline values per step, while others remained unchanged. Seven scenarios per factor were generated in a process to identify the most sensitive factors. The number of planning units with very high restorability (RESI 5) was selected as the dependent variable to detect the most sensitive influential impact. This analysis aided the identification of determinants that lead to significant variations of the dependent variable via small value amendments.
Primary criteria
Sub-criteria
Weights of subcriteria
Natural impacts
(1) (2) (3) (4)
0.0509 0.2304 0.0872 0.3063
Anthropogenic impacts
(5) Land use (6) Reclamation history (years) (7) Power density of agricultural machinery (8) Grain yield density
Topographic saturation Soil characteristics Stream order Overland flow distance (m)
0.0876 0.0819 0.0017 0.1539
Table 3 Area and proportion of each class of the Restorability Index values.
3.6. Restoration prioritizing and planning Although the restorability model provided bases to evaluate restoration suitability, implementing all the wetland restorations at the same time is not feasible due to limitations of resources and time. Thus, wetland restoration needs to be prioritized via considering the current situation of remaining wetlands. Except for four counties that are characterized by forest landscapes from past to present, thus being only little affected by floods, other counties with restoration potential were prioritized by a combination of vulnerability and restorability, represented by existing wetland ratio and percentage of high restorability occupation. The grades were prioritized according to the scatter plot of Fig. 3. Unsafe areas with low restorability and safe areas with high restorability were grouped into GradeⅠ; safe areas with low restorability and very safe area with high restorability were grouped into GradeⅡ; and very safe areas with low restorability were grouped into GradeIII.
Levels
Descriptions
Area (km2)
Percentage of the total area (%)
RESI RESI RESI RESI RESI
Most suitable Highly suitable Moderately suitable Least suitable Unsuitable
14,973 25,961 27,638 26,832 11,375
14 24 26 25 11
5 4 3 2 1
weights showed that, in all anthropogenic impacts, grain yield density has more significance in the restoration probability of degraded wetlands, while in all natural impacts, overland flow distance and soil character play important role in wetland restoration potential. 4.2. RESI levels and assessment According to the current land use pattern, the most suitable areas for wetland restoration cover 14,973 km2 (accounting for 14% of the whole study area). Adding the highly suitable areas increases the total extent to 40,934 km2, which corresponds to 38% of the total study area (Table. 3). These high restorability areas are mainly located in the conjunction zones of different rivers and around the Lake Xingkai, including the administrative counties of Luobei, Tongjiang, Fuyuan, Suibin, Fujin, Baoqing, Minshan, and Hulin (Fig. 4). All these areas have not been encroached by farmlands for a long time (within the consuming limitation of soil seed banks) and have not been intensively utilized. In Fujin County, the wetlands that had been converted into farmlands for less than five years have a very high probability of recovery. The RESI values revealed a larger proportion (64%) of the degraded area to be suitable for wetland restoration, indicating great potential for the recovery of ecological patterns and species habitats. In the Sanjiang Plain, remaining wetlands occupy 8532 km2, covering less than 14% of the total study area. These areas can be naturally recovered if they degraded or were polluted, since they are still within the wetland category. However, degraded wetlands that were converted into grasslands and farmlands require more efforts and the complexity of restoration depends on the time of reclamation, which is determined by the soil seed bank. Due to the age limit of the soil seed bank, farmlands that are reclaimed within 20 years were considered capable of being recovered without artificial cultivation (Wang et al., 2015). The extent of these farmlands has been defined as the effective restoration area (approximately 5777 km2), which is less costly to restore. Of the effective restoration area, about 39% fell into the most suitable category of restoration, 22% fell into a high suitability level, and 21% fell into a moderate level of suitability. Therefore, almost 85% of the lost or degraded wetlands can be recovered to their natural condition.
4. Results 4.1. Weights of criteria Weights of different criteria in the restorability model were generated via the PCA method and the results are presented in Table. 2. Due to extensive agricultural encroachment, substantially degraded wetland resulted from the high density of agricultural activities. The
4.3. Sensitivity analysis Fig. 3. Prioritization based on vulnerability and restorability.
The planning unit numbers of the most suitable sites for restoration 433
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 4. Spatial distribution characteristics of Restorability Index in the Sanjiang Plain.
434
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 5. Number of planning units of the most suitable restoration sites in 7 scenarios: 1, the specified factor reduced by 60% on the baseline value; 2, the specified factor reduced by 40% on the baseline value; 3, the specified factor reduced by 20% on the baseline value; 4, the baseline scenario; 5, the specified factor increased by 20% on the baseline value; 6, the specified factor increased by 40% on the baseline value; 7, the specified factor increased by 60% on the baseline value.
drastically increased, when overland flow length, soil characteristic, and grain yield density varied through scenarios 1–7 (Fig. 5), indicating that overland flow length, soil characteristic, and grain yield density are the most three sensitive factors for restoration of degraded or lost wetlands. Although overland flow length and soil characteristic are very hard and costly to be changed into suitable condition in restoration on the ground, they imply that areas with suitable condition both of these factors will give us quick responses when we implementing wetland restoration. Grain yield density is also sensitive to restoration, which indicates that wetlands with relatively low grain yield density would recover more rapidly. Power density of agricultural machinery has the lowest weight but it showed a sudden sensitivity increase in the last, which indicates that when the strength of reclamation is reduced to a certain degree the restoration will speed up. Land use is a little more sensitive than stream order, reclamation history and topographic saturation. The result revealed that, if the wetland had not converted into other land use types, natural factors can help them in their recovery. Furthermore, if the wetland had been converted into farmland, land use and power density of agricultural machinery became the main impact, influencing the probability of successful wetland restoration. This
indicates that if a restoration plan were in place for these farmlands, the power agricultural machinery on the lands have to be controlled. 4.4. Restorability prioritization and restoration plan Data from existing wetland and degraded wetland revealed that 17 counties now have more than 5% of wetlands occupation and are in a safe situation (Table. 4). Wetlands in other counties have the potential for restoration and may improve the ecological security in these counties. After restoration (except for four excluded counties), all counties would be in the safe category. In our prioritization results, five counties were classified into the first grade of restoration action (three in safe wetland ratio and in high restorability section, two in unsafe wetland ratio section); seven counties (three in safe wetland ratio and in low restorability section, four in very safe wetland ratio and high restorability section) were classified into the second grade of restoration action; and seven counties (in very safe wetland ratio and low restorability section) were classified as the last grade of restoration action (Fig. 6). The first grade counties in unsafe section underwent severe wetland degradation and are currently threatened by natural disasters. The first grade counties in safe section have very high restorability, which means wetlands in these counties are easy to recover. The second grade in safe section is relatively hard to recover but easy to become unsafe. The second grade in very safe section is relatively simple to recover. The wetlands in the last grade are hard to recover and they are currently in very safe situation; therefore, there is no need for immediate restoration action. If degraded or lost wetlands were recovered, 11 counties would have more than 10% of wetlands and others would have a wetland ratio between 5% and 10%, indicating the reduction of passive impacts according to the conclusion of DeLaney.
Table 4 Proportion of existing wetland and degraded wetland in the Sanjiang Plain. County name
Area of existing wetlands (km2)
Area of degraded wetlands (km2)
Ratio of existing wetlands (%)
Ratio of degraded wetlands (%)
Wetland ratio after restoration (%)
Fuyuan Tongjiang Mishan Suibin Hulin Jiamusi Raohe Fujin Baoqing Luobei Tangyuan Huachuan Hegang Jidong Yilan Boli Huanan Qitaihe Youyi
2522.65 1993.81 2188.87 623.53 1482.45 245.31 841.27 1023.97 1128.40 729.69 354.20 191.56 287.29 200.73 275.75 255.71 251.68 74.53 25.97
507.94 715.65 388.57 220.22 719.26 49.95 193.65 1906.75 188.85 256.81 70.06 81.80 27.42 3.30 178.69 101.23 92.27 16.83 119.45
40.27 32.31 28.34 18.61 15.89 13.03 12.75 12.46 11.37 10.82 10.24 8.50 6.30 6.21 5.99 5.73 5.69 4.27 1.54
8.11 11.60 5.03 6.57 7.71 2.65 2.94 23.20 1.90 3.81 2.02 3.63 0.60 0.10 3.88 2.27 2.09 0.96 7.08
48.38 43.90 33.37 25.18 23.60 15.69 15.69 35.65 13.27 14.63 12.26 12.13 6.91 6.31 9.88 8.00 7.78 5.24 8.62
5. Discussion Data of spatial changes showed severe degradation of wetlands in the study area and wetland restoration is urgent. This study identified natural and anthropogenic factors and their weights in wetland restoration. It also successfully assessed restorability and restoration prioritization of the study area. The weights identified showed that part but not all anthropogenic factors play important roles in wetland restoration and some factors (both natural and anthropogenic) that were expected to be important have not been supported by data. In this section, we will discuss the necessity of wetland restoration, some 435
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
Fig. 6. Prioritization of implementing actions for all counties in the Sanjiang Plain.
436
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
restoration plan. With regard to the implementation cost, the management cost is not the only aspect we need to take into account: opportunity cost also deserves consideration (Plantinga et al., 2001; Naidoo et al., 2006). Furthermore, for the technologies of wetland restoration, many studies tend to be site specific (Millard et al., 2013; Kershner, 1997) and mainly focus on one single influencing factor. Studies that unify the sites into the whole wetland system require further research.
issues in modelling and implementations, and our further researches. 5.1. Necessity for large extent restoration During the past four decades, the area of intact wetlands in the Sanjiang Plain drastically decreased due to irrational use of wetlands (such as burning and draining) for agricultural expansion. Their extent reduced by more than 75% of the original area. The habitats and corridors for species have been fragmented which significantly influence species’ habitats and gene flow (Naugle et al., 2001; Ceresa et al., 2015). Existing research indicated that if watersheds comprised 5–10% of wetlands, the peak flood period would reduce significantly. Although extant wetlands occupied 7.85% of the study area, these marshes and water were distributed unevenly within the watersheds of the whole plain, particularly in the densely constructed and agriculture areas. The wetlands in statuary nature reserves cover only 3.88% of the total study area and about half of these reserves had been replaced by agriculture. If the wetlands outside the conservation reserves disappeared or were to degrade, the entire region would be out of ecological security control and become unstable. If the recommended restoration plan was implemented by restoring lost and degraded wetlands, adding existing nature protected areas, the total extent of wetlands would increase from 7.85% to 13.17% of the study area. This would increase the area and connectivity of habitats for local species and thus improving the stability and biodiversity of the entire study area.
5.4. Further research To improve the accuracy of the model, we need to incorporate more factors influencing wetland restoration in our continuing research. Beyond the factors used for this study, the diversity of seed bank, microbial environment, total amount of fertilizer, population density, and other related determinants need to be considered. Based on the field sampling data, a regression correlation analysis could be completed to build a more reliable restorability model. Due to a lack of test plots for the control of wetland restoration in the Sanjiang Plain, we have to resort to natural reserves that have projects of returning farmlands to wetlands to seek for the variation of wetland restoration by comparing restored areas with adjacent natural wetlands. We expect such a statistical method to generate more objective results and we will compare the results with outcomes of this study to analyse the rationality of both models.
5.2. Assessment of the model
6. Conclusions
Despite the advantages of RESI model technologies, random uncertainties in manual operation of data processing are inevitable; for example, vector format of the land use, interpreted by remote sensing images had an accuracy of 85%, which is short of the 100% target. A further problem is that DEM data and its ramifications were not up-todate and the resolution does not allow finer scale analyses. All these processes caused uncertainty in the results. To increase accuracy of the model, more up-to-date and reliable data relying on technology development are required. Our results of weights assigning showed that both natural and anthropogenic factors have impacts on wetland restoration. Among all identified factors, overland flow length, soil characteristics and grain yield density are more important and sensitive to restoration. However, topographic saturation was not assigned a high weight as expected and it is not sensitive to changes. This might have been due to the correlation between flow length and topographic saturation, or topographic saturation might have less influence on marshes than river and lake wetlands. For the anthropogenic factors, the intensity of land use is more important than the way of land use. These results give us enlightenment in further implementation of wetland restoration. Other factors such as diversity of seed banks, microbial environment, and total amount of fertilizer also impact strongly on wetland restorability (Wang et al., 2015). They should be included to conduct more detailed and comprehensive further research for wetland restoration.
In this study, we proposed a GIS-based model for evaluating the restorability of wetland in Sanjiang Plain, China. This model successfully identifies priority areas that have high potential for wetland restoration based on both natural and anthropogenic impacts. It provides a useful tool for prioritization of wetland restoration implementations in the agricultural landscape in a spatially explicit way. In further research, we will incorporate more factors that influence wetland restoration to improve the accuracy of the model and apply a regression correlation analysis with more field sampling data to improve the reliability. In general, the model proposed here allows for wetland managers and planners to effectively identify wetlands that should be given priority for restoration in agricultural landscape and at the same time reflects the relative sensitivity of different influencing factors. This model is adaptable to other regions and its resulted wetland restoration plan will guide adjustments of human activities to accelerate the restoration process and thus improve the restoration efficiency greatly. Acknowledgements We thank the China Wetland Scientific Database for the provided wetland distribution data set. Funding for data surveys and collection came from the National Natural Science Foundation of China (Project Number 41501583). The English writing of the manuscript was carefully edited by Mr. Thomas S., Professional Editor, Friedrich-Alexander Universität Erlangen.
5.3. Implementing the plan
References
To implement wetland restoration on a regional scale, we have to consider three problems: the first problem is management and control, i.e. the type of policy to incorporate the degraded or lost wetlands into a system; the second problem is the cost of implementing restoration and the balance between economic development and natural conservation; finally, the problem of restoration technologies. For management and control, incorporating restoration into existing wetland conservation networks is a feasible method because strict restriction of reserves greatly contributes to the restoration process. Moreover, many areas within nature reserves have undergone wetland degradation, up to wetland loss and they are good starting points for a step-by-step
Bedford, B.A., 1996. The need to define hydrologic equivalence at the landscape scale for freshwater wetland mitigation. Ecol. Appl. 6, 57–68. http://dx.doi.org/10.2307/ 2269552. Benasseni, J., 2010. On the algebraic relation between principal components corresponding to two different sets of weights. Linear Multilinear Algebra 58 (4), 425–429. Bethel, M.B., Brien, L.F., Esposito, M.M., Miller, C.T., Buras, H.S., Laska, S.B., Philippe, R., Peterson, K.J., Richards, C.P., 2014. Sci-TEK: A GIS-based multidisciplinary method for incorporating traditional ecological knowledge into Louisiana's coastal restoration decision-making processes. J. Coastal Res. 297 (5), 1081–1099. http://dx.doi.org/10. 2112/JCOASTRES-D-13-00214.1. Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area model of
437
Ecological Indicators 91 (2018) 429–438
Y. Qu et al.
s11273-013-9303-9. Moreno-Mateos, D., Comin, F.A., 2010. Integrating objectives and scales for planning and implementing wetland restoration and creation in agricultural landscapes. J. Environ. Manage. 91, 2087–2095. http://dx.doi.org/10.1016/j.jenvman.2010.06.002. Naidoo, R., Balmford, A., Polasky, S., Rouget, M., 2006. Integrating economic costs into conservation planning. Trends Ecol. Evol. 21 681e687. Naugle, D.E., Johnson, R.R., Estey, M.E., Higgins, K.F., 2001. A landscape approach to conserving wetland bird habitat in the prairie pothole region of eastern South Dakota. Wetlands 21 (1), 1–17. http://dx.doi.org/10.1672/0277-5212(2000) 020[0588:ALATCW]2.0.CO. Ouyang, N.L., Lu, S.L., Wu, B.F., Zhu, J.J., Wang, H., 2011. Wetland restoration suitability evaluation at the watershed scale-A case study in upstream of the Yongdinghe river. Procedia Environ. Sci. 10, 1926–1932. Patenaude, T., Smith, A.C., Fahrig, L., 2015. Disentangling the effects of wetland cover and urban development on quality of remaining wetlands. Urban Ecosyst. 18 (3), 663–684. http://dx.doi.org/10.1007/s11252-015-0440-1. Peng, Y.L., Wu, N., Gao, X.F., Fang, Z.Q., Xiao, W.Y., 2010. Soil seed banks in lakeshore wetlands: relation to the extant vegetation. Pol. J. Ecol. 58 (3), 449–457. Plantinga, A.J., Alig, R., Cheng, H.T., 2001. The supply of land for conservation uses: evidence from the conservation reserve program. Resour. Conserv. Recycl. 31 (3), 199–215. http://dx.doi.org/10.1016/S0921-3449(00)00085-9. Randjelovic, D., Stankovic, J., Jankovic-Milic, V., Stankovic, J., 2013. Weight coefficents determination based on parameters in factor analysis. Metal. Int. 18 (3), 128–131. Verhoeven, J.T.A., Arheimer, B., Yin, C., Hefting, M.M., 2006. Regional and global concerns over wetlands and water quality. Trends Ecol. Evol. 21, 96–103. http://dx.doi. org/10.1016/j.tree.2005.11.015. Wang, M.N., Wang, H., Qin, D.Y., Lu, C.Y., Li, Y.P., 2011. Modelling the artificial recharge of a wetland and its influence on regional hydrological process in China: a case study. Ecohydrology 4 (4), 589–596. http://dx.doi.org/10.1002/eco.175. Wang, G.D., Wang, M., Lu, X.G., Jiang, M., 2015. Effects of farming on the soil seed banks and wetland restoration potential in Sanjiang Plain, Northeastern China. Ecol. Eng. 77, 265–274. Yang, W.H., Liu, Y.B., Cutlac, M., Boxall, P., Weber, M., Bonnycastle, A., Gabor, S., 2016. Integrated economic-hydrologic modeling for examining cost-effectiveness of wetland restoration scenarios in a canadian prairie watershed. Wetlands 36 (3), 577–589. http://dx.doi.org/10.1007/s13157-016-0768-1. Zedler, J.B., 2003. Wetlands at your service: reducing impacts of agriculture at the watershed scale. Front. Ecol. Environ. 1, 65–72. http://dx.doi.org/10.2307/3868032. Zhang, W.T., Dong, W., 2004. SPSS Statistical Analysis Advantage Textbook. Higher Education Press, Beijing.
basin hydrology. Hydrol. Sci. Bull. 24, 43–69. Carver, S., 1991a. Site search and multi-criteria evaluation. Plann. Outlook 34 (1), 27–36. http://dx.doi.org/10.1080/00320719108711886. Carver, S., 1991b. Integrating multi-criteria evaluation with geographical information systems. Int. J. Geog. Inf.Sci. 5 (3), 321–339. http://dx.doi.org/10.1080/ 02693799108927858. Ceresa, F., Belda, E.J., Kvist, L., Rguibi-Idrissi, H., Monrós, J.S., 2015. Does fragmentation of wetlands affect gene flow in sympatric Acrocephalus warblers with different migration strategies? J. Avian Biol. 46 (6), 577–588. http://dx.doi.org/10.1111/jav. 00589. Crumpton, W.G., 2001. Using wetlands for water quality improvement in agricultural watersheds; the importance of a watershed scale approach. Water Sci. Technol. 44, 559–564 11804150. Dale, W., Siobhan, F., 2005. Modeling the suitability of wetland restoration potential at the watershed scale. Original Res. Art. Ecol. Eng. 24 (4), 359–377. http://dx.doi.org/ 10.1016/j.ecoleng.2005.01.012. DeLaney, T.A., 1995. Benefits to downstream flood attenuation and water quality as a result of constructed wetlands in agricultural landscapes. J. Soil Water Conserv. 50, 620–626. Economic and Social Research Institute (ESRI), 2007. ArcGIS Desktop Help 9.2, USA. Francisco, A.C., Ricardo, S., Nadia, D.C., Mercedes, G., Adriá, M., 2014. A protocol to prioritize wetland restoration and creation for water quality improvement in agricultural watersheds. Ecol. Eng. 66 (3), 10–18. http://dx.doi.org/10.1016/j.ecoleng. 2013.04.059. Hatvany, M.G., 2009. Wetlands and reclamation. Int. Encycl. Hum. Geogr. 241–246. Hong, J.M., Liu, S., Shi, G.P., Zhang, Y.Q., 2012. Soil seed bank techniques for restoring wetland vegetation diversity in Yeyahu Wetland, Beijing. Ecol. Eng. 42, 192–202. Jenks, G.F., 1967. The data model concept in statistical mapping. International Yearbook of Cartography, 7: 186e190. Joliffe, I.T., Morgan, B.J., 1992. Principal component analysis and exploratory factor analysis. Stat. Methods Med. Res. 1 (1), 69–95. Kershner, J.L., 1997. Setting riparian/aquatic restoration objectives within a watershed context. Restor. Ecol. 5 (4S), 15–24. http://dx.doi.org/10.1111/j.1526-100X.1997. 00015. Liu, G.D., Tian, K., Sun, J., Xiao, D., Yuan, X., 2016. Evaluating the effects of wetland restoration at the watershed scale in Northwest yunnan plateau, China. Wetlands 36 (1), 169–183. http://dx.doi.org/10.1007/s13157-015-0727-2. Millard, K., Redden, A.M., Webster, T., Stewart, H., 2013. Use of GIS and high resolution LiDAR in salt marsh restoration site suitability assessments in the upper Bay of Fundy, Canada. Wetlands Ecol. Manage. 21 (4), 243–262. http://dx.doi.org/10.1007/
438