Research on response of water quality to land use pattern in nine plateau lake watersheds of Yunnan Province, China

Environmental pollution includes point-source pollution and non-point source pollution. The former derives from within and can be controlled through technical measures, but the latter is wide-spread, scattered, random and highly uncertain, making it difficult to research on and control (Dennis et al. 1998; Wang et al. 2018; Zhang et al. 2019). Many studies in China have proved that land use or land cover is a major cause for non-point source pollution in lake watersheds, and the farmland is the largest contributor to the values of TN (total nitrogen content) and TP (total phosphorus content) among all types of land use (Geng et al. 2015; Zhao et al. 2017; Jin et al. 2018). Optimizing the land use pattern, using the land in a scientific way and strengthening farming management will be effective measures to control non-point source pollution and improve the water quality in watersheds (Santini & Valentini 2011; Meng & Xu 2015).

Among the nine plateau lakes in Yunnan, Dian Lake, Yilong Lake, Xingyun Lake and Qilu Lake are under severe pollution, Chenghai Lake and Yangzong Lake are under medium-level pollution, and only the water in Lugu Lake, Fuxian Lake and Erhai Lake meets the standard for good-quality water. It is an onerous and urgent task to protect the water environment in the watersheds of these lakes. In these three years, there have been many studies on the nine plateau lakes in Yunnan province of China, but most of them were about Dian Lake and Erhai Lake, and few were about Qilu Lake, Xingyun Lake and Yilong Lake. Moreover, most of these studies focused on topics including temporal-spatial distribution of water quality (Gao et al. 2013; Zhao et al. 2013), ecological evaluation (Yang et al. 2012a, 2012b; Zhang et al. 2012), the population structure and evolution (Dong & Wang 2013; Yang et al. 2018), value of ecological services (Li et al. 2018; Liu et al. 2018), water environment modelling and monitoring (Wei et al. 2013; Fan et al. 2018) and distribution of pollution (Dong et al. 2012; Xie et al. 2013). Studies on the correlation between land use patterns and water quality were mainly conducted in the Dian Lake watershed (Yang et al. 2012a, 2012b; Niu et al. 2014). All these studies concentrated on key and hotspot regions, but few studies took the nine plateau lakes as an integrated whole to explore the correlation between land use and water quality in the watersheds of these nine plateau lakes.

Therefore, this study probed into the land use pattern and functions in the watersheds of the nine plateau lakes in Yunnan province of China, explored the law of spatial distribution of land use types and analyzed the characteristics of land use patterns in these lake watersheds. On this basis, this study identified the correlation between the land use pattern in the watersheds of the nine plateau lakes and the water quality of the lakes. This study will be of practical significance for adjustment of the land use pattern in the watersheds and treatment of non-point source pollution, provide decision support for local government to carry out watershed management.

Nine plateau lake watersheds located in Yunnan Province, China. The Dian Lake watershed is in the upstream of Pudu River, a branch of Jinsha River. It spans from 24°29′N to 25°28′N, and from 102°29′E to 103°01′E, covering an area of 2,891.07 km², and the water-covered area in the watershed reaches 322.39 km². The altitude of the watershed decreases southwards and the watershed is subject to a climate with distinct dry and wet seasons. Located in Dali Bai Autonomous Prefecture, the Erhai Lake watershed stretches from 25°25′N to 26°10′N, and from 99°32′E to 100°27′E, covering an area of 2,609.07 km², and the water-covered area in the watershed reaches 263.90 km². Erhai Lake is a typical inland sag pond and has distinct dry and wet seasons. The Fuxian Lake watershed stretches from 24°13′N to 24°46′N, and from 102°39′ E to 103°00′E. The watershed covers 690.03 km², and the water-covered area reaches 221.30 km². Fuxian Lake is located between a basin on its north and an alluvial plain on its south, and it is subject to a climate with distinct dry and rain seasons. The Lugu Lake watershed stretches from 27°36′N to 27°47′N, and from 100°43′E to 100°54′E, covering an area of about 247.8 km². Lugu Lake is under administration of two provinces, with its eastern part administrated by Yanyuan Yi Autonomous County of Sichuan province and its western part by Ninglang Yi Autonomous County of Yunnan province. The Qilu Lake watershed covers an area of 340.8 km². The lake has a shoreline of 32 km, an average depth of 4.5 m and a volume of 168 million m³. Three rivers – Zhong River, Yaochong River and Daxin River, join Qilu Lake. Located in the north of Jiangchuan County in Yuxi City, the Xingyun Lake watershed covers an area of 378 km². The lake stretches 10.5 km from north to south, and its average width from west to east is 3.2 km, with a shoreline of 36.3 km. The watershed has a typical plateau monsoon climate, with distinct dry and rain seasons. Chenghai Lake stretches from 26°27′N to 26°38′N, and from 100°38′E to 100°41′E. The lake covers an area of 74.6 km², with a runoff area of 318 km², an average depth of 25.7 m and is able to store 1.95 billion m³ of water. The watershed is subject to the temperate montane monsoon climate, with warm temperature and abundant sunshine. Located 30 km east of Kunming city, Yangzong Lake stretches from 24°46′N to 24°59′N, and from 102°54′E to 103°04′ E, covering an area of 192 km². The Yangzong Lake watershed has a typical subtropical climate and is subject to the monsoon. Located in Shiping County of Honghe Hani and Yi Autonomous Prefecture in Yunnan province, Yilong Lake is a sag pond and covers an area of 44.4 km². With an average depth of 2.93 m, the lake can reach as deep as 6.75 m and store water as much as 130 million m³. Though subject to the subtropical plateau montane monsoon climate, Yilong Lake shows features of a wide range of climates from tropic climates to frigid climates.

The data concerning the current situation of land use in this study are from remote-sensing images captured by Landsat 8 OLI in 2016, the resolution ratio of which is 30 m. DEM (Digital Elevation Model) data are the original elevation data obtained by SRTM 90 m DEM from the Geospatial Data Cloud of Chinese Academy of Sciences. ENVI5.1, ArcGIS10.0 and other software were used to extract data for current land use conditions and elevation. ENVI5.1 was used to calibrate the downloaded original remote sensing images, and after amplifying the images, the wavebands of 5, 4 and 3 were selected for false-color combination. According to the ground-object characteristics of field surveys, this study selected the training samples, made classification and evaluated the accuracy of classification. Aside from the major research object; that is, the tea gardens, this study referred to the Standard for Current Land Use Classification (GB/T 21010-2007) and classified the land into six types – farmland, woodland, grassland, land for buildings, water-covered land and land for other purposes (including unused land, rock-exposed barren land and others). By processing the images, including image smoothing, image splicing and cutting, the classification result (Figure 1) was obtained, and data of land use types and their respective area were extracted by ArcGIS 10.0.

In order to improve the accuracy of remote sensing image interpretation, an improved decision tree classification method is adopted. Combined with the spectral characteristics of land use, the ratio index, NDVI value and texture index are introduced into decision-making classification, and the remote sensing data extraction model of decision tree is established to improve the accuracy of remote sensing image classification.

According to ‘2016 Bulletin of Environmental Conditions in Yunnan’, this study obtained water quality indices of the nine plateau lakes in Yunnan province over 12 months, including the permanganate index, trophic state index, total nitrogen content (TN), total phosphorus content (TP), biochemical oxygen demand (BOD) and chemical oxygen demand (COD). The statistics used to analyze the correlation between water quality and land use of the nine plateau lakes are the means of each water quality index over a span of 12 months (Table 1).

where r is the correlation coefficient between two variables and ︱r︱represents the correlation level between two variables. When r > 0, the two variables are positively correlated; when r < 0, the variables are negatively correlated; and when r = 0, it means there is no correlation. As︱r︱approaches 1, the level of correlation increases. To be specific, when ︱r︱is between 0.8 and 1, the correlation is extremely strong, when ︱r︱falls between 0.6 and 0.8, the correlation is strong; when ︱r︱ is between 0.4 and 0.6, the correlation is at a medium level; when ︱r︱ is between 0.2 and 0.4, the correlation is weak; when ︱r︱ falls between 0 and 0.2, the variables are barely or not correlated.

The correlation coefficient obtained needs to be verified by the significance p-value, and when p ≤ 0.05, the correlation is deemed significant. The software SPSS 22 is used for analysis, and in the result correlation coefficient r, * means significant correlation at the p ≤ 0.05 level, and ** means significant correlation of variables at the p ≤ 0.01 level. Due to the limited number of samples, there are errors in the correlation analysis result, and to make the study more scientific, only the significant correlation at the p ≤ 0.01 level is analyzed.

With remote images, this study interpreted the classification result, extracted data about the type and area of land use, and hence obtained the area and proportion of each type of land in the watersheds of the nine plateau lakes (Figure 2).

On the basis of analysis of the spatial distribution features of each land-use type in a watershed, the following characteristics are obtained.

In the Dian Lake watershed, the woodland takes up the largest area, covering 132,400,000 hm² and accounting for 45.79% of the total area in the watershed. It is followed by farmland, which covers an area of 67,000 hm² and accounts for 23.19% of the watershed's total area. Most of the farmland is along the lake's eastern bank and is scattered in other sites in the watershed. The land used for buildings, which concentrates on the mild slopes in the watershed's northern part, also takes up a large proportion, covering 47,000 hm² and accounting for 16.25% of the total area of the watershed. Grassland, which is scattered within the watershed, takes up a small proportion, covering a mere area of 4,100 hm² and accounting for 1.43% of the total area of the watershed.

In the Xingyun Lake watershed, the farmland takes up the largest area, covering 20,400 hm² and accounting for 52.76% of the total area of the watershed. The woodland that concentrates on the northern and southeastern parts of the watershed comes in second place, covering 8,700 hm² and accounting for 22.47% of the total area of the watershed. The land for buildings and water bodies cover an area of 3,700 hm² and 4,200 hm², taking up a respective percentage of 9.62 and 10.81% of the total area of the watershed. Land for buildings is mainly along the southwestern bank of the lake. The grassland, which is mainly distributed along the skirts of the watershed, covers a small area of 1,200 hm² and takes up merely 3.16% of the total area of the watershed.

In the Yilong Lake watershed, the woodland takes up the largest area, covering 17,800 hm² and accounting for 49.26% of the total area of the watershed. It is followed by farmland, which covers 7,500 hm² and takes up 20.72% of the total area of the watershed. The farmland is mainly along the western bank of the lake and stretches from southwest to southeast. The land for buildings covers 4,300 hm² and takes up 12.07% of the total area in the watershed. Buildings are mainly found in the northwest of the watershed, and along the northwestern bank and southeastern bank of Yilong Lake. Land for buildings along the lake's southeastern bank is mainly between the valleys or along the rivers. The grassland takes up a small area, covering merely 1,500 hm² and accounting for 4.16% of the total area in the watershed, and it can only be found on mild slopes in the watershed.

Farmland takes up the largest area in the Qilu Lake watershed, covering 17,900 hm² and accounting for 50.1% of the total area in the watershed. Except the water-covered areas, the watershed is largely occupied by farmland. The woodland that is mainly found in the southern part of the watershed comes in second place, covering 7,200 hm² and accounting for 20.09% of the total area of the watershed. The land for buildings, mainly found along the bank of the lake, covers 3,200 hm² and accounts for 8.81% of the total area of the watershed. The area of grassland takes up a small area, covering merely 1,200 hm² and accounting for 2.47% of the total area of the watershed. The grassland is mainly found in the northwest of the watershed and a few areas in the southern part of the watershed.

In the Chenghai Lake watershed, woodland takes up the largest area, covering 17,500 hm² and accounting for 51.76% of the total area of the watershed. Farmland covers 4,000 hm² and makes up 11.81% of the watershed's total area. Most of the farmland is located to the south of the lake and some is scattered along the bank of the lake. The land for buildings, covering 2,400 hm² and accounting for 6.98% of the watershed's total area, is mainly found along the southern bank of the lake. The grassland that covers 1,600 hm² and takes up 4.88% of the watershed's area in total can be found mainly in the northeast of the watershed.

In the Yangzong Lake watershed, woodland takes up the largest area, covering 6,000 hm² and accounting for 31.47% of the total area, and spreads mainly in the watershed's southern part. Farmland covers an area of 4,800 hm² and takes up 25.31% of the watershed's area in total. The farmland is distributed in the southwest of the watershed, mainly along the bank of Yangzong Lake. The grassland covers an area of 3,100 hm² and takes up 15.97% of the total area of the watershed. Grassland is mainly found in the northwest and southeast of the watershed, and it is also scattered in other sites within the watershed. The land for buildings, mainly distributed in the north of the watershed, covers an area of 1,400 hm² and takes up 7.54% of the total area of the watershed.

In this watershed, woodland takes up the largest area of 135,500 hm² and accounts for 51.94% of the watershed's total area. The farmland, covering an area of 44,600 hm² and making up 15.08% of the total area of the watershed, is mainly distributed in the north of the watershed and along the northern, western and southern bank of the lake. The grassland, covering an area of 40,100 hm² and taking up 13.37% of the total area of the watershed, is scattered sparsely within the watershed. The land for buildings, covering an area of 13,600 hm² and taking up 8.2% of the total area of the watershed, is mainly distributed along the southwestern bank of the lake.

The woodland in this watershed takes up the largest proportion compared with the proportion it takes up in other watersheds. Woodland in this watershed covers an area of 12,900 hm² and makes up 51.97% of the total area. Except for the water-covered areas, woodland is the major type of land use in this watershed. Farmland is mainly distributed along the southeastern bank of the lake, covering an area of 2,000 hm² and taking up 8.1% of the total area of the watershed. Grassland is mainly distributed along the rivers, covering an area of 2,700 hm² and taking up 10.74% of the total area of the watershed. The land for buildings covers an area of 1,500 hm² and takes up 5.89% of the total area.

Apart from water-covered areas, farmland takes up the largest area in this watershed, covering 24,300 hm² and taking up 31.15% of the total area of the watershed. The woodland is mainly distributed in the northwest of the watershed, covering 14,00 hm² and taking up 20.24% of the total area in the watershed. Grassland is mainly found in the northwest of the watershed, covering 7,300 hm² and taking up 10.59% of the total area of the watershed. Land for buildings is mainly distributed along the northern bank of the lake, covering 1,300 hm² and taking up 5.91% of the total area of the watershed.

Based on analysis of the spatial distribution of each type of land in different watersheds, the following characteristics are concluded.

Farmland takes up the largest proportion of 52.76% in the Xingyun Lake watershed, followed by 50.10% in the Qilu Lake watershed. The proportions it takes up in Chenghai Lake watershed, Erhai Lake watershed and Lugu Lake watershed are small and below 15.08%. In the Lugu Lake watershed, it takes up the smallest proportion of 8.10%. The woodland takes up large proportions in the Dian Lake watershed, Yilong Lake watershed, Chenghai Lake watershed, Erhai Lake watershed and Lugu Lake watershed, all above 45%. In particular, the proportion it takes up in the Lugu Lake watershed, 51.97%, is the highest. Woodland takes up small proportions in other watersheds and has the minimum proportion of 20.09% in the Qilu Lake watershed. The grassland takes up small proportions in the nine watersheds, all below 16%, and the smallest proportion it takes, 3.16%, occurs in the Xingyun Lake watershed. The largest proportion that the land for buildings takes up is in the Dian Lake watershed, reaching 16.25%, followed by 12.07% in the Yilong Lake watershed. The land for buildings takes up small proportions in other watersheds, and the smallest proportion it takes up, 5.89%, occurs in the Lugu Lake watershed. Water-covered land takes up the largest proportion of 22.53% in the Lugu Lake watershed, followed by 22.50% in the Chenghai Lake watershed. The smallest proportion the water-covered land takes up, 10.11%, is in the Erhai Lake watershed.

Figure 3 shows the analysis result of water quality indices in the nine plateau lakes. The permanganate index reaches its highest at 18.45 mg/L in Yilong Lake, and the index remains above 7.87 mg/L in Dian Lake, Xingyun Lake and Qilu Lake. In Chenghai Lake, Yangzong Lake, Erhai Lake, Lugu Lake and Fuxian Lake, the permanganate index is relatively low and stays below 3.087.87 mg/L, with the index in Lugu Lake marking the lowest at 0.65 mg/L. The trophic state index stays above 40.51 in seven of the nine lakes, except Lugu Lake and Fuxian Lake where it is below 20.71. The highest trophic state index, 66.17, occurs in Xingyun Lake. The TN value is high in Dian Lake, Xingyun Lake, Yilong Lake and Qilu Lake, staying above 1.61 mg/L. Qilu Lake marks the highest TN value of 2.93 mg/L. But in other lakes, including Chenghai Lake, Yangzong Lake, Erhai Lake, Lugu Lake and Fuxian Lake, the TN value is below 0.78 mg/L, with Lugu Lake marking the minimum of 0.12 mg/L. The TP value reaches its highest at 0.21 mg/L in Xingyun Lake. Its value is also high in Dian Lake and Qilu Lake, where it stays above 0.10 mg/L. In the other six lakes, the TP value remains below 0.05 mg/L. The BOD value reaches its highest at 6.50 mg/L in Xingyun Lake, followed by 6.00 mg/L in Qilu Lake. The index is also high in Dian Lake and Xingyun Lake, but it stays below 2.27 mg/L in the other five lakes, with Lugu Lake marking the lowest value of 0.41 mg/L. The COD value is high in Dian Lake, Xingyun Lake, Yilong Lake and Qilu Lake, where it stays above 38.58 mg/L, with Yilong Lake marking the highest value of 82.94 mg/L. In the other five lakes, the COD value is below 24.50 mg/L.

To sum up, in 2016, the water quality in Lugu Lake and Fuxian Lake is the best and meets the Level I and Level II water standards; the water quality in Erhai Lake and Yangzong Lake is good and meets the Level III water standard; the water in Chenghai Lake is under mild pollution and meets the Level IV water standard; the water in Dian Lake and Qilu Lake is under medium pollution and meets the Level V water standard; water in Yilong Lake and Xingyun Lake is under severe pollution and is below the Level V water standard.

By using SPSS, Pearson correlation analysis was conducted to analyze the land use type and water quality in the nine plateau lake watersheds. With the proportion of each type of land in these watersheds as the independent variable and the water quality indices as dependent variables, this study analyzed their correlation. The analysis result is as follows.

As shown in the correlation analysis result (Table 2), the proportion of farmland is positively correlated to all the water quality indices. The correlation coefficients that the proportion of farmland has with the trophic state index, TN, TP and BOD are high and exceed 0.560; the proportion of farmland has significant correlation with the TP value at the 0.05 level, with a high coefficient of 0.770; it has significant correlation to the BOD value at the 0.01 level, with the correlation coefficient reaching 0.838. In summary, the water quality deteriorates as the proportion of farmland increases. The proportion of woodland is negatively correlated to the permanganate index, the trophic state index, TN value, TP value and BOD value, with low correlation coefficients that stay below 0.486. That is to say, the water quality improves as the proportion of woodland increases. The proportion of grassland is negatively correlated to water quality indices, with high correlation coefficients that stay above 0.636. Among the indices, the proportion of grassland shows significant correlation to the trophic state index, TN, TP and COD at the 0.05 level, which means that the water quality improves as the area of grassland rises. The proportion of land for buildings is positively correlated to all the water quality indices. The correlation coefficients it has with the permanganate index, trophic state index, TN and COD are high, reaching above 0.574. Among these indices, there is significant positive correlation with the trophic state index and COD at the 0.05 level, which means that the water quality deteriorates as the area of land for buildings increases. The proportion of water-covered land is positively correlated to all the water quality indices, with high correlation coefficients that stay above 0.500, and it shows significant positive correlation with the trophic state index, TN and BOD at the 0.05 level, which means the water quality improves as the proportion of water bodies increases.

According to analysis result of the correlation between the land use type and water quality indices in the nine plateau lake watersheds, this study identified the response relationship between the proportion of the type of land and the water quality index. With the multiple linear regression model built based on the least squares method, this study established the response relationship between the land use structure and water quality indices in the nine plateau lake watersheds. Variables without significant correlation as shown in the correlation analysis result are removed and only variables with significant correlation at the p<0.05 level are selected. That is, the selected variables include the correlation coefficients of the farmland with TP and BOD, the correlation coefficients of the grassland with the trophic state index, TN, TP and COD, the coefficients of the land for buildings with the trophic state index and COD, the coefficients of water-covered land with the trophic state index, TN, BOD, and the correlation coefficient between the land for other purposes and TN. With the proportion of the type of land use as the independent variable and the water quality index as the dependent variable, this study built a regression model between water quality of the lakes and the land use pattern.

As shown in Table 3, the coefficients of determination for fitting linear regression are all above 0.451, and adjusted coefficients of determination are above 0.373. The significance coefficients of regression P are all below 0.05, and the F testing values are all larger than the significance coefficients. Therefore, the regression model built in this study is of statistical importance, but as the number of statistical variables is limited, the fitting level of part of the model is not high.

In these nine plateau lake watersheds, woodland and farmland rank top two among all types of land in terms of the area they occupy. Woodland takes up the largest proportion of area in Dian Lake watershed, Yilong Lake watershed, Chenghai Lake watershed, Yangzong Lake watershed, Erhai Lake watershed and Lugu Lake watershed. In particular, the Lugu Lake watershed marks the highest proportion of 51.97%, followed by the proportion of farmland. Farmland ranks in second place in terms of the area it occupies in these nine lake watersheds. It claims the largest proportion in Yunxing Lake watershed, Qilu Lake watershed and Fuxian Lake watershed. The proportion farmland takes up in Xingyun Lake watershed is the largest among all the nine watersheds, reaching 52.76%, followed by the proportion taken up by woodland. The type of land that takes up the smallest proportion is the land for other purposes in these lake watersheds, except the Qilu Lake watershed where grassland marks the smallest proportion. The largest proportion of land for buildings occurs in Dian Lake watershed, followed by the proportion it takes up in Yilong Lake watershed. The smallest proportion of land for buildings occurs in Lugu Lake watershed, and Fuxian Lake watershed marks the second smallest proportion of land for buildings. The proportion of water-covered land is largest in Lugu Lake watershed, followed by Chenghai Lake watershed, and the Erhai Lake watershed marks the smallest proportion of water-covered land.

As for the spatial distribution of land use types, woodland takes up the largest proportion of land in Dian Lake watershed, Yilong Lake watershed, Chenghai Lake watershed, Yangzong Lake watershed, Erhai Lake watershed and Lugu Lake watershed. The reason for this distribution pattern is that these areas are mountainous, with large relief and steep slopes. In the watersheds of Xingyun Lake, Qilu Lake and Fuxian Lake, the farmland accounts for the major type of land use, because these watersheds have few mountains but lots of mild slopes, and most of the population rely on agriculture. In the watersheds of Dian Lake, Xingyun Lake and Yilong Lake, the land for buildings takes up large proportions because, on one hand, these watersheds are more urbanized, have more economic strength and accommodate more urbanites, and on the other hand, these watersheds have flat terrain and large areas of mild slopes.

Certainly, human factors including population density, industrial activities and tourist development are also accountable for changes in lake water quality, but land utilization is still the major contributing factor because all human activities determine the changes in land utilization patterns. Therefore, studying the response of lake water quality to land use patterns in the spatial-temporal dimension can help explore a reasonable industrial and economic structure of the study area and thus promote sustainable development.

The following conclusions are reached in this study. The proportion of farmland is positively correlated to the water quality indices, showing significant positive correlation with TP at the 0.05 level and significant correlation with BOD at the 0.01 level, which means the water quality deteriorates as the proportion of farmland increases. The proportion of woodland is in negative correlation with the permanganate index, the trophic state index, TN, TP and BOD, which means the water quality improves as the proportion of woodland increases. The proportion of grassland is negatively correlated to all the water quality indices, with large correlation coefficients. It shows significant negative correlation with the trophic state index, TN, TP and COD at the 0.05 level, which means the water quality improves as the proportion of grassland increases. The proportion of land for buildings is positively correlated to water quality indices and shows significant positive correlation with the trophic state index and COD at the 0.05 level, which means the water quality deteriorates as the proportion of land for buildings increases. The proportion of water-covered land is positively correlated to water quality indices, with large correlation coefficients. It shows significant positive correlation to the trophic state index, TN and BOD at the 0.05 level, which means the water quality improves as the proportion of water-covered land increases.

Based on the spatial distribution of different types of land use in the nine plateau lake watersheds, this study analyzed the correlation between land use types and water quality, and built a response model. It is expected to provide a theoretical basis for optimization of the land use pattern and pollution treatment in the lake watersheds.

The next study is to classify the water pollution status of nine plateau lakes, analyze the response differences between different pollution types of lakes and land use, and then formulate land use strategies.

The research was supported by The National Natural Science Foundation Regional Science Foundation Project of China (No. 41961040).

All relevant data are included in the paper or its Supplementary Information.