The Chaishitan reservoir is one of the centerpieces of the Nanpanjiang River and is located upstream of the Zhujiang River (i.e., the Pearl River). The actual maximum water level of the Chaishitan reservoir is 1,643.74 m, and its corresponding capacity is of 0.381 billion m³. However, Yiliang city and its surroundings often suffer severe drought. For example, in 2010, the drought areas of Yiliang county totaled 196.2 km², and 104 km² of croplands produced nothing due to drought at the end of that year (Anonymous, 2011). In addition, more than 100,000 people had difficulties in accessing safe drinking water because the vast majority of the surface rivers had dried up and the groundwater level had severely decreased in 2010. Therefore, it is urgent to establish an irrigation project (i.e., Chaishitan WCP) from the Chaishitan reservoir to meet the needs of surrounding Yiliang and Shilin counties.

Based on the Current Land Use Classification (GB/T 21010-2007) (CMDLRDC (Cadastral Management Division of Land and Resources Department of China) and LSPI (Lands Surveying and Planning Institute), 2007) issued by China's Land and Resource Ministry, we adopted Current Land Use Classification (Table 1). As the national standard, China's Land Use Status Classification is mandatory, principled and guiding; therefore, we adopted this unified standard in order to improve the level of land management and promote its application in Yiliang county of Yunnan province. Secondly, China's Land Use Status Classification conforms to China's national conditions; namely, it meets the relevant laws and regulations of China's domestic standards, which makes it easier for the construction project to obtain approval. Thirdly, we adopted the nationwide uniform standards, which facilitate the smooth development of the project's construction work; it will be easier to compare our results with similar projects in China and to provide reference for future related projects (Chen et al., 2013).

The methods of remote sensing image interpretation are widely used and have been applied in previous studies (e.g., Qiu and Wang, 2010; Wang et al., 2013, 2016; Booth et al., 2016). The first-level classification was based on land use and land cover (LULC) characteristics, and the secondary classification is primarily based on the first survey characteristics of land management. One scene of Operational Land Imager (OLI) data from Landsat 8 was obtained in March 9, 2015 as the main source for the second survey of LULC classification. The spatial resolution of panchromatic and multispectral OLI was 15 and 30 m, respectively. Three main steps were taken for the classification. First, we performed a multi-resolution image segmentation to generate image objects to which the classification algorithm could be applied. For this segmentation, we used a “scale parameter” of 20, which was determined by visual interpretation of the segmentation results. The segmentation process identified objects that were homogeneous and included the features (i.e., spectral values, shape, texture, etc.) that can be used for classification (Walker and Blaschke, 2008; Qiu and Wang, 2010). Second, once the segmentation was achieved, we utilized a combination of fuzzy rules and a standard nearest neighbor (SNN) algorithm to classify each image object. A total of 200 training samples were selected and referenced with the auxiliary data (including digital topographic maps). We printed 20 images wherein land use or vegetation types could not be determined from the data alone, brought these images into the field, and either found the specific positions or asked the local people to identify the location. In total, 20 field surveys for the unidentified land use or vegetation types were conducted. Google Earth and local specific maps were also referenced in order to identify their representative classes. Based on the spectral and spatial information of these samples, we created the SNN feature space and fuzzy rule algorithms that were collectively used for classification. Each image object was assigned a probability of belonging to each LULC class, and the final class of image objects was decided on the basis of which assigned class has the highest probability. Finally, we refined the classification with manual adjustment to improve the overall quality of classification. The classification was performed using Definiens 7.0 software. Classification accuracy was assessed by comparing the reference collection with classified imagery (Congalton, 1991). Based on the derived LULC map, we further calculated the percentage for each LULC type in ArcGIS 9.3 (ESRI).

We overlaid the proposed CAD digital maps onto the above remote sensing image from Landsat 8, and then we interpreted the image again and predicted LULC changes immediately after the WCP has been built in 2016. Other ancillary data included 1:50,000 topographic maps, 1:250,000 land-use maps from 2014, district administrative maps and related land resources survey data, inventory reports and related maps.

The field survey on plant species structure and diversity adopted a previously used sampling protocol (Wang et al., 2011) to investigate the community composition and structure of typical vegetation in the region. The proposed layout of the Chaishitan WCP has one main-channel and two branch channels in Yiliang county, and one major main-channel and five branch channels in Shilin county. According to local people who are familiar with the proposed layout of Channels, we found the specific positions of the future channels passing away, and we set 21 plant diversity sampling plots along the channels or near the channels (Figure 1). Once the water conservancy established, most plant species will disappear because of habit loss, therefore, we get the plant diversity before the water conservancy established and predict how could plant diversity change based on the CAD maps and modeling (i.e., we assumed the plant species in or nearby proposed channels will disappear once the water conservancy established in 2016).

We performed plant diversity field investigations twice; the first time was conducted from May to June in 2015 for 27 days with seven people involved in the field work, and the second time was conducted from August to September in 2015 for 42 days with 11 people involved in the field work. We conducted field work according to different land use types: i.e., Secondary needle and broadleaf mixed forest (SF); Wasteland (Wetland) (WL); Eucalyptus plantation (EP); Orchard (OR), and Farmland (FL) (Table 2). The number of sampling sites with each kind of land use type varied from three to eight; eight sampling sites fell into Farmland. In total, 21 sampling sites were investigated (Table 2, Supplementary Material A). The plant diversity was investigated at three different layers; i.e., tree, shrub, and herb layers. At each sampling site, we investigated one 20 × 20 m tree plot: five 2 × 2 m shrub plots and five 1 × 1 m herb plots were surveyed within the tree plot at its four corners and center. In total, we had 21 tree plots, 105 shrub plots, and 105 herb plots. We recorded each species name, diameter at Breast Height (DBH) for trees, height, crown width, coverage for herb, and origin.

Five kinds of main vegetation types were found in the field:

There was a high vascular plant taxonomic diversity (over 1,000 plant species) found in the investigated fields. A total of 1076 vascular plant species (including subspecies, varieties and form, Supplementary Material B) were identified in the Chaishitan irrigation region, belonging to 165 families and 647 genera (Supplementary Material B). Of these, 23 species (13 families and 16 genera) belong to pteridophytes, accounting for 0.88% of all pteridophyte species in China, whereas 60 species (8 families and 31 genera) are gymnosperms, accounting for 0.8% of gymnosperms species in China. The remaining 993 species (144 families and 600 genera) are angiosperms, accounting for 3.58% of all angiosperms in China. Angiosperm species accounts for 92.29% of all vascular species.

All 1,076 vascular species could be divided into three categories: (1) 295 species (87 families and 226 genera) are wild native species; (2) 748 species (138 families and 462 genera) alien are cultivated species, including vegetables, fruit, aromatic plant, cash crops, and a very large number of garden plant species; (3) 33 species (16 families and 30 genera) are alien invasive species, such as Eupatorium adenophorum Hort. Berol. ex Kunth, Conyza canadensis L. and Ageratum conyzoides Sieber ex Steud.

Poaceae, Fabaceae, and Rosaceae accounted for the most families in Chaishitan irrigation region. Poaceae was represented by the most species (70 species, 48 genera), in which 46 species are wild native species. Fabaceae and Rosaceae were represented by the second (62 species and 36 genera) and the third highest number of species (61 species and 21 genera), respectively. In addition, 49 families were represented by only one species.

Plant taxonomic diversity indices (i.e., d, He′ and Je) differed in different vegetation types. In tree layer, index d, He′ and Je are the highest in farmland compared to the other four vegetation types (Figure 4A). In shrub layer, d and He′ are the highest in Wasteland (WL); however, Je is the highest in Farmland (FL) (Figure 4B). In the herb layer, d, He′ and Je are the highest in the secondary needle- and broad-leaved mixed forest (SF) (Figure 4C).

Some rare or protected plant species were found in the Greenhouses or Nurseries; i.e., cultivars were mainly for ornamental or medicinal use, and were not wild species, such as Alsophila spinulosa (Wall. ex Hook.) R. M. Tryon, Ginkgo biloba Linn., Juglans mandshurica Maxim., Pseudolarix amabilis (Nelson) Rehd., Glyptostrobus pensilis (Staunt.) Koch, Metasequoia glyptostroboides Hu et Cheng, Taxus wallichiana var. chinensis (Pilg.) Florin, Liriodendron chinense (Hemsl.) Sargent., Pachylarnax sinica (Law) N. H Xia et C. Y. Wu, Phoebe zhennan S. Lee et F. N. Wei, Fagopyrum dibotrys (D. Don) Hara, Davidia involucrate Baill., Sinojackia xylocarpa Hu, and Kolkwitzia amabilis Graebn. were found in our field investigation. These species are listed as national key protected plants species. However, as these were not wild species, their genetic diversity is lower. Furthermore, as there were a large number of individuals of these species, they have a lower conservation value. There were numerous clones of the protected species in the greenhouses and nurseries.

The accuracy of remote sensing depends on the data extraction process (Zhu et al., 2016). In our study, the remote sensing image was from OLI data based on Landsat 8: its resolution is 30 m. Thus, the accuracy defines the meaningfulness of the data. For example, the percentage decline of Farmland and Forestland was 0.079 and 0.066%, respectively. However, if the sensing accuracy is only 70%, these changes are not meaningful at all, because the simple classification error is much greater than the magnitude of changes. Unfortunately, high resolution images are either simply not available or prohibitively expensive, as are images from SPOT (Satellite Pour l'Observation de la Terre) or Quickbird. In this study, based on OLI data from Landsat 8, the CEI before and after Chaishitan WCP was established is 0.92 and 0.89, respectively. Although the CEI decreased by 0.03, both CEI values indicate that the ecological state is healthy, according to the criteria in Zheng et al. (1994). These values indicate that the Chaishitan WCP will make the CEI decrease slightly, but will not decrease the rank of the overall ecological quality of the Chaishitan region.

Environment Impact Assessments are often conducted before a project is established, can lead to a lack of land use/cover assessments after establishment. In this study, we predicted the changes of land use/cover based on the data from planning diagrams, and found that the structure of landscapes could change after the establishment of the Chaishitan WCP. Specifically, PV, SHDI, PAFRAC, and LDI decreased, while PD, CONTAG, and SPLIT increased, indicating increased landscape fragmentation along with decreases in landscape diversity. The Chaishitan WCP establishment could result in land use change and therefore most landscape indices would change; e.g., PV decreases because the channels or canals would occupy the areas normally containing some natural vegetation, resulting in natural vegetation decreases. On the other hand, the Chaishitan WCP establishment would make the landscape more fragmented, and therefore, the SPLIT will increase due to the generation of more patches.

In this study, the Chaishitan WCP is predicted to change the area values of different land types (Table 1); however, the land use change should not decrease plant diversity because of the three following reasons: first, the change in farmland area (0.161 km²) and forest land area (0.104 km²) is limited (Table 1). Second, as shown in our field investigation, the Chaishitan project will not occupy natural forest, for all affected forests are secondary forests, plantations or orchards (Tables 4A,B). The plant diversity in the farmland and artificial forestland is not high, as most species are cultivated species or alien species, and the species in the forest are widely distributed in that region. Finally, there is no special species sensitive to local habit (Supplementary Material B). On the contrary, Chaishitan WCP might increase the number of local plant species in the long term because once the project is in operation, the human interventions will be reduced, which may create more beneficial habitat or niche for more species due to sufficient water supply.

From a whole ecosystem perspective, the ability of the ecosystem to resist alien species invasion, plant diseases and pest invasion will increase because of sufficient water supply. In the long term, ecosystem stability will depend on multiple factors (e.g., improved soil texture), and the temporary biodiversity decrease should not be reflected in the CEI decrease in the future, because the ecosystem resilience will improve the plant diversity.

An Impact Assessment is an important tool for conservation and sustainable use of biodiversity (IAIA, 2005). There is a growing interest in promoting biodiversity consideration in impact assessments (e.g., IAIA, 2005; CBD (Convention on Biological Diversity), 2006). In order to support this trend, many guidelines and tools have been developed (e.g., World Bank, 2000; OECD (Organisation for Economic Cooperation Development), 2002). Wegner et al. (2005) points out the diversity in definitions and approaches to its assessment among EIA practitioners. Geneletti et al. (2003) further argued that accounting for uncertainty in biodiversity impact assessment—in data, methodologies, and value judgments provided by the experts—is important. Gontier et al. (2006) also address the gap between research in prediction tools and current practice in biodiversity assessment within environmental assessment. Post-project-analysis (PPA) refers to a method and system of tracing, monitoring and confirmatory assessing the environmental impact of constructed projects and the efficiency of preventive measures, as well as proposing remedial plans or measures, aiming to achieve the coordination between project construction and environment (Division of Bill, Resources and Environmental Commission of National People's Congress (DBRECNP), 2003). At present, the PPA in environmental impact in China is basically at the stage of discussing the concept, indicators, methods, content and procedure of the assessment as well as developing case studies (Li et al., 1997; Shen et al., 2005; Cai et al., 2007), while post-project analysis in construction projects is not yet widely conducted. In this study, our assessment was conducted before the project was established, i.e., Before-Project-Analysis (BPA), which could maximally decrease the environmental risk the project brings. In this study, we used modeling predict the land use and plant diversity change immediately after the project established, while as discussed previously, plant could restore to its mature state in a long term (e.g., 5 years, 10 years), therefore, it seems that another comprehensive field investigation is needed to understand what the irrigation system will change. The rare or endangered plant species were found in the greenhouse or nursery, and no wild individuals were found in the investigated field. Furthermore, the irrigation project will provide sufficient water once it is established, which will enrich local plant diversity.

The loss of biodiversity in non-natural reserves is much higher than the loss rate of nature reserves due to over-exploitation, habitat loss, etc. in China (Xia et al., 2005). In this study, we used the Yiliang WCP as an example and we set up hundreds of plant diversity plots to systematically investigate the plant diversity, abundance, and conservation status. Our real intent is to generate more attention the landscape and plant diversity of non-natural reserve areas in China.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.