• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Evaluating the Importance of Environmental Variables on Spatial Distribution of Caspian cobra Naja oxiana (Eichwald,1831) in Iran

    2019-06-26 03:03:42ElmiraKAZEMIMohammadKABOLIRasoulKHOSRAVIandNematollahKHORASANI
    Asian Herpetological Research 2019年2期

    Elmira KAZEMI, Mohammad KABOLI, Rasoul KHOSRAVI and Nematollah KHORASANI

    1 Department of Environment, Faculty of Natural Recources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran

    2 Department of Environmental Science, Faculty of Natural Resources, University of Tehran, Karaj, Iran

    3 Department of Natural Resources and Environmental Engineering, School of Agriculture, Shiraz University, Shiraz,Iran

    Abstract Over recent years, the population of Caspian cobra Naja oxiana has declined in its distribution range in Iran due to habitat destruction and overhunting. Consequently, their small and isolated populations in fragmented landscapes are facing genetic and demographic threats. Evaluating the spatial distribution pattern of Naja oxiana, identifying core habitat patches and improving landscape connectivity among the patches have a significant role in the long-term survival of the species. This study predicts the spatial distribution map of the Caspian cobra considering the factors affecting the predictive power of the distribution models, including sampling bias in presence points, correct selection of background locations, and input model parameters. The sampling bias in presence points was removed using spatial filtering. Several models were run using 19 environmental variables that eventually led to the selection of the effective habitat variables and best MaxEnt distribution model. We also used an ensemble model (EM) of habitat suitability methods to predict the potential habitats of the species. Topographical roughness, shrublands, average annual precipitation, and sparse rangeland with a density of ≤ 20% had the most effect on the spatial distribution of Caspian cobra. The evaluation of models confirmed that the EM has more predictive performance than MaxEnt in predicting the distribution of Naja oxiana.

    Keywords Naja oxiana, MaxEnt, habitat suitability, landscape connectivity, distribution modelling

    1. Introduction

    Habitat fragmentation is one of the most important threats to biodiversity that leads to reduction in habitat area, increase in habitat isolation and reduction in ecological function of the remaining habitat patches(Ripple et al., 2014). In a fragmented landscape, the survival of small and isolated populations is affected by individual's movement between patches (Wiens, 2001;Niebuhr et al., 2015). Therefore, the conservation and monitoring of threatened and vulnerable species in a fragmented landscape requires maintaining the landscape connectivity between habitat patches. To improve landscape connectivity, managers need to have sufficient information about spatial patterns of species distribution and the effects of anthropogenic and landscape factors on the species distribution and dispersal. Landscape connectivity modelling can be improved by learning about how species interact between biotic and abiotic elements, as well as the spatial distribution of species in landscape (Bellamy et al., 2013). In addition, managing and preserving of landscape connectivity is affected by precise and reliable data of species distribution.

    Ecological niche models (ENMs) or spatial distribution models (SDMs; Warren and Seifert, 2011) are efficient tools in determining the factors affecting species presence, disease spread, and predicting the probable changes in the distribution range of invasive and nonnative species (Khanum et al., 2013). SDMs provide an accurate prediction of habitat suitability. Since accurate absence data for many species are scarce, in recent years,much progress has been made to improve and develop SDMs based on only presence data. Maximum Entropy(MaxEnt) have been introduced a powerful tool for mapping species distribution with a limited number of present data (Elith et al., 2006; Phillips et al., 2009).

    Although ENMs have high performance in ecological studies and conservation biology, these algorithms may suffer from over fitting and biased prediction. Such challenges may occur in modelling due to 1) sampling bias in presence data, 2) inaccurate selection of absence data,

    3) lack of sufficient data to assess the model accuracy, and

    4) false choice of spatial scale of environmental variables(Veloz, 2009; Barve et al., 2011). The application of ENMs using the biased presence points, especially data resulting from opportunistic observations can lead to a model over-fitting and reduce the accuracy and reliability of the model (Phillips et al., 2009; Boria et al., 2014). In addition to sampling bias in presence points, uncertainty in the predictions from these algorithms should be considered. Each algorithm has inherent limitations and errors that result in different levels of efficiency and accuracy in modelling (Araújo and New, 2007; Araújo et al., 2011). Developing ensemble models (EMs) is a valuable method to overcome the uncertainties of each model and increase the overall accuracy of predictions.In EMs, a landscape is suitable only if all algorithms in the EM classify it that way (Zarco-Gonzalez et al., 2013).Over the past few years, EMs have been used in many studies focusing on predicting species distribution and assessing landscape connectivity (Araújo et al., 2011;Kujala et al., 2011; Rodriguez-Soto et al., 2011; Abade et al., 2014).

    The Caspian cobra Naja oxiana, also called the Central Asian cobra, is a member of the family Elapidae. It has the easternmost distribution for the Naja subgenus in the Palaearctic region. Naja oxiana lives in dry, rocky habitats and areas with sparse vegetation. Its distribution ranges from northern and north-eastern Iran to southern Turkmenistan, southern Uzbekistan, western Tajikistan,northern Pakistan, north and east Afghanistan, and northern India (Wüster and Thorpe, 1989; Wüster, 1990;Wallach et al., 2009). In recent years, the population of the species has dramatically declined in northern habitats due to illegal overhunting (Darvish and Rastegar-Pouyani, 2012). Moreover, declining population size as a result of habitat degradation and excessive exploitation for medical purposes such as experiments on anti-cancer compounds (Wüster et al., 2007; Pook et al., 2009)can put this species at risk of extinction. Also, drying up of some rivers and springs due to reduced rainfall has influenced the distribution of the species. Habitat fragmentation increases patch isolation and reduces the ecological suitability of the remnant patches and landscape connectivity between them. These changes may negatively affect gene flow among subpopulations in a metapopulation pattern (Young, 2000). Hence, assessing species distribution, identifying core habitat patches, as well as managing and preserving landscape connectivity among patches plays an important role in directing future research, conservation management programs and ensuring the long-term survival of the species.

    The present study was designed with the aim of using the EMs together with considering the factors affecting the performance of the models. We had three objectives:1) to determine the effects of environmental and human factors on the distribution pattern of Caspian Cobra; 2) to determine habitat patches and the landscape connectivity among patches; 3) to evaluate how the input parameters of predictive distribution models (i.e. regularization parameter, the number of input variables, and feature types) affect model performance.

    2. Material and Methods

    2.1. The study area and occurrence pointsThe present study was conducted in the distribution range of Caspian cobra in the north and northeast of Iran within the geographical range of 38°-30° N to 62°-51° E (480 000 km2; Figure 1). A total of 47 presence points were obtained from 2015-2016, based on direct observations and genetic sampling. The coordinates of all the presence points were recorded using a hand-held multi-channel Global Positioning System (GPS). The average annual precipitation of the area varies between 50 and 960 mm.The most important factors causing habitat fragmentation and landscape changes in the landscape include road network development, expansion of agricultural activities,and land use changes.

    2.2. Eco-geographical variablesNineteen ecological and ecogeographical variables were selected based on the literature review and were categorized in five main classes including topography, bioclimatic variables(Hijmans et al., 2005), anthropogenic, vegetation, and land cover (Table 1). Because of the high correlation (r >0.70) between bioclimatic variables, only five bioclimatic variables were used in this study including average annual temperature (Bio1), temperature seasonality(Bio4), maximum temperature of the warmest month(Bio5), average annual precipitation (Bio12), and precipitation seasonality (Bio15). Focal mean elevation,topographic roughness, and slope were extracted using a digital elevation map (Shuttle Radar Topography Mission (SRTM elevation model; http://srtm.csi.cgiar.org) and the Surface Gradient and Geomorphometric

    Modelling toolbox (Evans et al., 2014) in the ArcGIS.In addition, Normalized Difference Vegetation Index(NDVI) was obtained in 2012 at 250-m resolution using MODIS Vegetation Indices (MOD13) C5 User's Guide(Solano et al., 2010). NDVI layers were summarized into several important components using principal component analysis (PCA). Only the first two components were used in distribution model. We obtained a digital map of vegetation cover from the Iranian Forests, Rangeland and Watershed Management Organization (FRWMO, 2009)and reclassified it into five land cover (LULC) types,including sparse rangeland with a density of ≤ 20%,mixture of grassland-scrubland with a density of ≥ 25%,shrublands, forests, and barelands. To provide continuity for the extracted categories, the proportional extent of each land cover type was calculated within a 2 km × 2 km grid network by running the ArcMap Neighbourhood statistic tool. Furthermore, the density of agricultural land,rural areas, geographical distance to residential areas, and the distance to the roads were also used as anthropogenic variables. All nineteen layers were resampled to a 1-km2cell size.

    Table 1 Categories of 19 environmental variables used in the distribution modelling of Caspian cobra.

    Figure 1 The study area in the northeast and east part of Iran. The dashed polygons and black dots show protected areas (PAs) and presence points, respectively.

    2.3. Modelling

    2.3.1. Evaluation of sampling bias in presence points and selection of background points Given that many of the presence points were recorded during fieldwork or based on available reports of collected specimens, we eliminated the probable bias in the presence points using spatial filtering method in ArcGIS (SDMtoolbox; Brown,2014). The threshold of 500 meters was used as a spatial filter to remove spatial autocorrelation in occurrence points. We filtered presence points so that each point was at least 500-m away from other points, which resulted in 41 unique occurrence points with decreased spatial autocorrelation. Selecting appropriate background localities is another challenge that should be considered in ENMs. In this study, to avoid selecting background points from environmental conditions outside the range of the presence points, a buffer with a radius of 30 kilometres was considered around the presence points and only the background points that were located within this distance around presence points were used in the modelling.

    2.3.2. Optimization of distribution model's parameters and variables used for modelling The R package“MaxEntVariableSelection” was used to eliminate correlation between the environmental predictors and to prevent multicollinearity and achieve an accurate goodness of fit value (Jueterbock et al., 2016). The software provides the possibility to optimize the model's parameters, including the regularization parameter, the number of input variables in the model, and feature types.In this study, the regularization index was evaluated between two to four with an interval of 0.5. This software also applies a repeat process to remove variables with a significance level less than the defined threshold value,which will depend on the species of interest, as well as a defined threshold value for correlation between variables.The threshold of significance and threshold of correlation were selected as 2.0 and 0.7, respectively. Moreover, we used AUC-testand AICc to determine the most important variables affecting the species distribution, the best regularization parameter, and feature types.

    2.3.3. MaxEnt distribution model and ensemble modelling After selecting the most important variables affecting the distribution of Caspian cobra and optimizing the input parameters, the final distribution model was implemented using the MaxEnt algorithm in the R package “dismo” (Hijmans et al., 2016). The MaxEnt model was performed using the optimized regularization parameter, 4000 background localities, five runs, and 5000 replications. The evaluation of the MaxEnt model was done using AUC-test. Afterwards, distribution maps were converted to presence and absence map using the threshold “Maximum Training Sensitivity and Specificity(MTSS)” (Liu et al., 2013). In order to evaluate the predictive power of the model and increase the performance of the distribution model, in addition to the MaxEnt model, a second predictive species distribution model was conducted using ensemble modelling approach(Araújo and New, 2007). Hence, three algorithms were used to produce the EM: Maximum Entropy (MaxEnt),generalized linear model (GLM), and generalized boosted model (GBM). All three models were implemented using the important variables identified in the previous steps.The EM was calculated using the weighted average of the values of AUC obtained from each model and the method proposed by Marmion et al. (2009) and Rodriguez-Soto et al. (2011). All the three algorithms MaxEnt, GLM and GBM, as well as the EM were run in the R package“biomod2”. All models were calibrated by using the 75%of occurrence points as training data, and tested on the remaining 25% to assess model performance (see Ahmadi et al., 2017; Khosravi et al., 2018 for details). Evaluating the performance of all models and the EM was carrying out using the AUC index (Deleo, 1993).

    3. Results

    After correcting the sampling bias in the presence points,41 points were used in the MaxEnt model. The AUC-testand AICc, showed that the optimal distribution model would be achieved with fewer variables. The number of environmental variables included in the final distribution model based on the AUC-testand AICc index was reduced to eight and five, respectively (Table 2). Moreover, the regularization parameter for the optimal model was three based on the AICc and two based on the AUC-test(Figure 2). The best species habitat suitability prediction was produced using average annual precipitation (BIO12),topographic roughness (ROUGH), sparse rangeland with a density of ≤ 20%, (SRD), density of agricultural fields(AGD), density of shrublands (SHR), distance to roads(ROD), distance to residential regions (URBD), and the first component of vegetation index (NDVI_1).

    Table 2 The best distribution model of Caspian cobra based on AICc and AUC-test in the MaxEnt model.

    Figure 2 The best regularization parameter in Caspian cobra predictive distribution model based on the AICc and AUC-test. The red circle shows the optimal regularization parameter.

    Evaluation of the final MaxEnt model using selected environmental variables based on the AUC-testshowed that MaxEnt had high predictive power in predicting the distribution and suitable habitats of Caspian cobra (AUC= 0.840). In Figure 3, the distribution map of suitable habitats is shown based on the suitability threshold.

    The highly suitable habitats for the species were more widely distributed in the north-eastern parts of Iran, which illustrates the importance of this region in protecting Naja oxiana across the study area. The most effective variables in the final MaxEnt model were identified in order of importance as 1) the density of shrublands, 2) sparse rangeland with a density of ≤ 20% and 3) topographical roughness. The suitable habitats map obtained from the MaxEnt model identified approximately 12% (58 610 km2) of the study area as suitable habitat for Naja oxiana.

    The response curves (Figure 4) of the most influential variables in the species distribution showed that density of shrublands provides highly suitable habitat for this species until a certain threshold (i.e. 0.2) after which it causes a negative effect on habitat suitability. Likewise, increasing the topographical roughness initially increases the habitat suitability, but after reaching a threshold, the habitat suitability decreases. The average annual precipitation has a strong, almost linear positive correlation with habitat suitability. The effect of sparse rangeland with a density of ≤ 20% indicated that it is equal probability of presence from 0 to approx. 0.75 and then steep decrease.

    3.1. Evaluation of Naja oxiana distribution using EM

    Figure 3 Potential distribution of Naja oxiana in northeastern Iran using MaxEnt. Predictions based on Average annual precipitation: BIO12(mm); Topographic roughness: ROUGH; Shrublands: SHR (%); Sparse rangeland with a density of ≤ 20%: SRD. The predicted probability of presence, with values ranging from 0 to 100, is depicted by colours.

    Figure 4 Response curves of the most influential predictors for distribution of Naja oxiana modelling with MaxEnt. (Average annual precipitation: BIO12 (mm); Topographic roughness: ROUGH; Shrublands: SHR (%); Sparse rangeland with a density of ≤ 20%: SRD).These curves show the mean response of the cross-validated models with 10 replicate runs (red) and the mean ± one standard deviation (blue).

    Figure 5 shows the distribution map of Caspian cobra obtained from the EM (MaxEnt, GLM, and GBM). The EM identified nearly 8% (42 574 km2) of the study area as suitable habitat for Naja oxiana, which had a relatively high overlap with the suitable habitats map derived from the MaxEnt. As was expected, the suitable habitat area in the EM was smaller than the MaxEnt model since the EM was an ensemble of three distribution algorithms and it produced a smaller suitable habitat area as compared to MaxEnt, which was based on only one algorithm. The evaluation of the three algorithms separately showed that all these three models have an efficient performance in predicting the distribution of Naja oxiana (MaxEntAUC=0.832; GLMAUC= 0.785; GBMAUC= 0.988). The GLM model was similar to MaxEnt, while the GBM showed better performance in predicting the distribution of the species. Additionally, the AUC index was calculated as 0.932 for the EM, which illustrates it had more power compared to the MaxEnt and GLM models. Assessing the most effective variables in the EM showed that four variables including sparse rangeland with a density of ≤20%, the first component of vegetation index, average annual precipitation and topographic roughness have the most impact on species habitat suitability.

    Figure 5 Potential distribution of Naja oxiana in northeastern Iran using ensemble modelling. Predictions based on Average annual precipitation: BIO12 (mm); Topographic roughness: ROUGH; Shrublands: SHR (%); Sparse rangeland with a density of ≤ 20%: SRD. The predicted probability of presence, with values ranging from 0 to 100, is depicted by colours.

    4. Discussion and Conclusion

    In this study, the habitat suitability and potential distribution map of Caspian cobra was predicted using the most effective variables based on an optimized modelling framework. In addition to considering environmental variables, other factors to increase model accuracy were addressed, including sampling bias in presence points,correct selection of backgrounds, and optimizing the input parameters. These factors have a significant role in the precision and accuracy of the ENMs. Considering that in most studies of species distribution modelling, recorded presence points are used from opportunistic observations,the bias in collecting and recording these points can affect the performance of prediction models (Fourcade et al.,2014). In this study, the spatial filtering method was used to eliminate the sampling bias. Studies show that reducing the number of presence points using spatial filtering is an effective technique in reducing sampling bias in presence points (Kramer-Schadt et al., 2013; Boria et al., 2014;Kittle et al., 2018). Furthermore, the results showed that the use of this method improves the final distribution model and reduces the probability of overfitting. Although the method used in this study to reduce the effect of sampling bias in presence points decreased the number of presence points, this method increased the reliability of the results.

    Ensemble models provide a better representation of species distribution patterns for many species. Until now, EMs have been used in many studies, including to assess the effects of climate changes, identify protected areas, and predict species distribution (Kujala et al.,2011; Carroll et al., 2010; Khosravi et al., 2018). Our results showed that the EM (AUC = 0.932) compared to MaxEnt model (AUC = 0.840) had more efficiency and power in predicting the distribution of Caspian Cobra in north and north-eastern Iran. This result was consistent with previous studies (Marmion et al., 2009;Rodriguez-Soto et al., 2011; Abade et al., 2014). In an EM, areas are predicted as suitable habitat only if all algorithms in the ensemble model have reached the same and stable prediction. Therefore, EMs in most cases show more efficiency compared to individual models(Zarco-Gonzalez et al., 2013). Generally, the area of suitable habitats predicted by EMs is smaller than the area predicted by a single algorithm (e.g. MaxEnt).Consequently, using EM in the case of dealing with data obtained from opportunistic observation (for example reptiles) is an effective approach to predict species distribution.

    Field studies show that Naja oxiana inhabits mostly stony and rocky habitats covered by bushes in dry and semi-arid climate and altitudes up to 3 000 meters (WCH Clinical Toxinology Resource, 2012). Therefore, it can be expected that annual precipitation and elevation play an important role in distribution and habitat suitability of Naja oxiana. The results of this study confirm the importance of these variables in the distribution and habitat suitability of the species. Based on the output of the EM, the behaviour pattern of the species to changes in habitat variables shows clearly a close correlation of the habitat selection behaviour of the species with its habitat elements. The study of the importance of variables based on EM determined that sparse rangeland with a density of ≤ 20%, average annual precipitation and topographic roughness have the greatest impact on the distribution of this species. Human variables in comparison with environmental variables showed less impact on species distribution.

    The probability of distribution and habitat suitability of the species increased with increasing distance to roads. However, the observations of Naja oxiana in areas affected by human activities imply the extension of its range beyond its preferred habitats (Kularatne and Senanayake, 2014). Since Caspian cobra prefers habitats with little human disturbance, constructing major roads with high vehicle traffic would decrease the habitat suitability for the species. Roads are often associated with high levels of noise pollution, which can have a negative effect on vibration-sensitive species, in addition to the high percentage of road killings of wildlife, including Caspian cobra and other reptile species. In addition,new roads can facilitate urban expansion, which leads to an increase in human interactions with habitats and consequently habitat destruction and fragmentation. For this reason, it is not surprising that such habitats have lower suitability. As for topographical roughness, changes in the response curve showed that an increase in surface roughness initially increased the habitat suitability until its effect became negligible, after which additional increases reduced suitability. A high increase in habitat suitability of Caspian cobra with a slight increase of surface roughness can be attributed to the dependence of the species on stony and rocky areas. Both elevation and surface roughness are variables with a high impact on the distribution of many species (Hof et al., 2012; Ahmadi et al., 2017, Khosravi et al., 2018). The combination of changes of two variables, sparse rangeland with a density of ≤ 20% and roughness, shows that stony-rocky areas with a low degree of roughness and sparse vegetation are one of the most suitable habitat types for Caspian cobra. Although Caspian cobra avoids areas with dense vegetation, regions with sparse vegetation or low plant density in rocky areas play an important role providing shelter and hiding patches for the species.

    The results of the study can play a remarkable role in identifying Caspian cobra habitat patches as well as creating connectivity between identified patches.Evaluating the ecological and habitat needs of the species is the first step of conservation management plans, and thus plays an essential role in the species survival. In the other words, creating landscape connectivity between identified patches and protecting these habitat patches are important in facilitating the gene flow between the remaining populations and thereby decreasing the genetic threats caused by population's disconnection.

    AcknowledgementsOur sincere thanks go to Raheleh DEZFOULIAN and Ben GOTTESMAN for their great assistance on editing and constructive criticism of the manuscript. This study was permitted by the Iranian Department of Environment under permits No. 94/43027 and 94/17701.

    黑人欧美特级aaaaaa片| 三级国产精品欧美在线观看| 草草在线视频免费看| 深夜精品福利| 亚洲18禁久久av| 国产成人系列免费观看| 伊人久久大香线蕉亚洲五| 悠悠久久av| 欧美三级亚洲精品| 国产一级毛片七仙女欲春2| av在线蜜桃| 中文字幕av成人在线电影| av中文乱码字幕在线| 免费看a级黄色片| 美女免费视频网站| 欧美午夜高清在线| 亚洲av成人av| 十八禁网站免费在线| 色精品久久人妻99蜜桃| 久久久久精品国产欧美久久久| 欧美日韩国产亚洲二区| 日韩欧美精品v在线| 亚洲熟妇中文字幕五十中出| 麻豆成人午夜福利视频| 亚洲av中文字字幕乱码综合| 日韩欧美国产一区二区入口| 亚洲无线在线观看| 夜夜夜夜夜久久久久| 18+在线观看网站| 国内少妇人妻偷人精品xxx网站| 久久中文看片网| 色播亚洲综合网| 老司机福利观看| 国产精品乱码一区二三区的特点| 欧美av亚洲av综合av国产av| 最近最新中文字幕大全免费视频| 精品午夜福利视频在线观看一区| 又黄又爽又免费观看的视频| 国产野战对白在线观看| 亚洲欧美一区二区三区黑人| 日韩欧美在线二视频| 成人国产综合亚洲| 婷婷六月久久综合丁香| 亚洲精品影视一区二区三区av| 特大巨黑吊av在线直播| 国语自产精品视频在线第100页| 成人三级黄色视频| 男女之事视频高清在线观看| 99久久综合精品五月天人人| 一个人看视频在线观看www免费 | 欧美在线一区亚洲| 日本三级黄在线观看| 老司机在亚洲福利影院| 久久精品国产清高在天天线| 真人一进一出gif抽搐免费| 国产色爽女视频免费观看| 欧美区成人在线视频| 精品国产美女av久久久久小说| 免费人成视频x8x8入口观看| 中文字幕人妻熟人妻熟丝袜美 | 成年人黄色毛片网站| 最近最新中文字幕大全电影3| 三级男女做爰猛烈吃奶摸视频| 国内揄拍国产精品人妻在线| www.www免费av| 亚洲天堂国产精品一区在线| 三级国产精品欧美在线观看| 国产精品日韩av在线免费观看| 国产淫片久久久久久久久 | 香蕉久久夜色| www日本黄色视频网| 岛国在线免费视频观看| 丰满人妻熟妇乱又伦精品不卡| 欧美日韩精品网址| 色吧在线观看| 18禁黄网站禁片午夜丰满| 国产精品电影一区二区三区| 色在线成人网| 男女视频在线观看网站免费| 国产高清三级在线| 日本黄大片高清| 国产成年人精品一区二区| 成年人黄色毛片网站| 免费人成在线观看视频色| 91久久精品电影网| 国产精品女同一区二区软件 | av女优亚洲男人天堂| 国产伦精品一区二区三区四那| 国产高清视频在线播放一区| 俄罗斯特黄特色一大片| 成人18禁在线播放| av在线蜜桃| 久久久久久久午夜电影| 免费看日本二区| 久久久久国产精品人妻aⅴ院| 3wmmmm亚洲av在线观看| 午夜免费激情av| 国内精品久久久久精免费| 老熟妇仑乱视频hdxx| 久久精品综合一区二区三区| 中出人妻视频一区二区| 精品福利观看| 国产国拍精品亚洲av在线观看 | 亚洲中文字幕日韩| 一个人免费在线观看的高清视频| 99久国产av精品| 黄色日韩在线| АⅤ资源中文在线天堂| 国产精品亚洲av一区麻豆| 亚洲国产精品合色在线| 高清日韩中文字幕在线| 国产亚洲欧美在线一区二区| 亚洲欧美激情综合另类| 亚洲色图av天堂| 他把我摸到了高潮在线观看| 青草久久国产| 国产精品亚洲美女久久久| 男女下面进入的视频免费午夜| 一个人观看的视频www高清免费观看| 国产精品久久电影中文字幕| 人妻夜夜爽99麻豆av| 久久久久久国产a免费观看| 国产亚洲欧美98| 欧美精品啪啪一区二区三区| 亚洲精品亚洲一区二区| 亚洲精品美女久久久久99蜜臀| 两个人的视频大全免费| 成年女人永久免费观看视频| 性欧美人与动物交配| 免费无遮挡裸体视频| 香蕉av资源在线| 久久性视频一级片| 香蕉久久夜色| 久久久久久久久久黄片| 一进一出抽搐动态| 黄色日韩在线| 国产一区二区三区在线臀色熟女| 国产精品久久电影中文字幕| 亚洲专区中文字幕在线| 丝袜美腿在线中文| 丁香六月欧美| av欧美777| 男插女下体视频免费在线播放| 高潮久久久久久久久久久不卡| 在线观看av片永久免费下载| 亚洲aⅴ乱码一区二区在线播放| 国产精品 欧美亚洲| 身体一侧抽搐| 51午夜福利影视在线观看| 国产成人啪精品午夜网站| 动漫黄色视频在线观看| 免费大片18禁| 亚洲人成网站高清观看| 五月伊人婷婷丁香| 可以在线观看的亚洲视频| 欧美日韩国产亚洲二区| 中文字幕人妻丝袜一区二区| 很黄的视频免费| 亚洲最大成人中文| 88av欧美| 成熟少妇高潮喷水视频| 大型黄色视频在线免费观看| 老鸭窝网址在线观看| 亚洲av成人精品一区久久| 成人国产一区最新在线观看| 岛国在线免费视频观看| 看片在线看免费视频| 最好的美女福利视频网| 99久久无色码亚洲精品果冻| 窝窝影院91人妻| 村上凉子中文字幕在线| 欧美乱妇无乱码| 99久久99久久久精品蜜桃| av欧美777| 亚洲国产色片| 国产精品乱码一区二三区的特点| 99在线人妻在线中文字幕| 日本黄色片子视频| 身体一侧抽搐| 亚洲七黄色美女视频| 最新美女视频免费是黄的| 国产精品久久久久久久电影 | 老司机午夜福利在线观看视频| 在线观看舔阴道视频| 亚洲片人在线观看| 在线免费观看不下载黄p国产 | 欧美高清成人免费视频www| 国产精品香港三级国产av潘金莲| 欧美一区二区亚洲| 岛国在线免费视频观看| 国产高潮美女av| 香蕉av资源在线| 一边摸一边抽搐一进一小说| 欧美国产日韩亚洲一区| 成人永久免费在线观看视频| 三级男女做爰猛烈吃奶摸视频| 此物有八面人人有两片| 深夜精品福利| 极品教师在线免费播放| 中文字幕熟女人妻在线| 夜夜爽天天搞| 少妇的逼好多水| 亚洲狠狠婷婷综合久久图片| 午夜免费激情av| 中文字幕精品亚洲无线码一区| 桃色一区二区三区在线观看| 亚洲欧美精品综合久久99| 久久伊人香网站| 国产精品野战在线观看| 国产精品一区二区三区四区免费观看 | 亚洲欧美激情综合另类| 首页视频小说图片口味搜索| 欧美中文日本在线观看视频| 国产精品久久久久久久久免 | 亚洲国产欧美人成| 亚洲av熟女| 久久久久国内视频| 老司机福利观看| 欧美激情久久久久久爽电影| 搡老岳熟女国产| 99国产精品一区二区蜜桃av| 亚洲av中文字字幕乱码综合| 黄片小视频在线播放| 色视频www国产| 亚洲 欧美 日韩 在线 免费| 国产精品久久久久久人妻精品电影| 人妻丰满熟妇av一区二区三区| 人人妻,人人澡人人爽秒播| 18禁裸乳无遮挡免费网站照片| 在线播放无遮挡| 91麻豆精品激情在线观看国产| 国产日本99.免费观看| 欧美区成人在线视频| 叶爱在线成人免费视频播放| 精品久久久久久久人妻蜜臀av| 搡女人真爽免费视频火全软件 | 国产极品精品免费视频能看的| 嫩草影视91久久| 亚洲av电影在线进入| 天堂影院成人在线观看| 国产一区二区在线观看日韩 | av福利片在线观看| 搡老妇女老女人老熟妇| 国产熟女xx| 日本精品一区二区三区蜜桃| 日本在线视频免费播放| 人人妻人人澡欧美一区二区| 无人区码免费观看不卡| 12—13女人毛片做爰片一| 嫩草影院入口| 成人国产综合亚洲| 精品免费久久久久久久清纯| 韩国av一区二区三区四区| 在线免费观看的www视频| 欧美一级毛片孕妇| 免费看a级黄色片| 一本一本综合久久| 日本成人三级电影网站| 最近最新中文字幕大全免费视频| 韩国av一区二区三区四区| 久久久久久久久大av| 色视频www国产| 亚洲中文字幕日韩| 成人无遮挡网站| 搞女人的毛片| 国产精品自产拍在线观看55亚洲| 国产在视频线在精品| 搡老岳熟女国产| 热99re8久久精品国产| 九九在线视频观看精品| 亚洲七黄色美女视频| 国内精品久久久久精免费| 国产精品av视频在线免费观看| 亚洲人成伊人成综合网2020| 国产精品综合久久久久久久免费| 十八禁人妻一区二区| 99精品在免费线老司机午夜| 老熟妇仑乱视频hdxx| 国产亚洲欧美在线一区二区| 国产色婷婷99| 欧美黑人欧美精品刺激| 日本在线视频免费播放| 亚洲片人在线观看| 女人被狂操c到高潮| 在线观看av片永久免费下载| 欧美日韩一级在线毛片| 国产高清视频在线播放一区| 最后的刺客免费高清国语| 国产精品乱码一区二三区的特点| 日韩欧美免费精品| 欧美日韩国产亚洲二区| 欧美性感艳星| 亚洲成av人片在线播放无| 少妇的逼好多水| 久久国产精品人妻蜜桃| 国产欧美日韩一区二区三| 国产精品久久久久久久久免 | 欧美在线黄色| 亚洲第一欧美日韩一区二区三区| 国产av在哪里看| 精品一区二区三区视频在线 | 女警被强在线播放| 每晚都被弄得嗷嗷叫到高潮| 亚洲av免费高清在线观看| 亚洲av二区三区四区| 制服丝袜大香蕉在线| 亚洲中文日韩欧美视频| 成年女人永久免费观看视频| a级一级毛片免费在线观看| 亚洲,欧美精品.| www.熟女人妻精品国产| 国产精品久久久久久精品电影| 狂野欧美激情性xxxx| 国产三级中文精品| 午夜激情福利司机影院| 国产精品影院久久| 97超级碰碰碰精品色视频在线观看| 一进一出抽搐gif免费好疼| a在线观看视频网站| 亚洲va日本ⅴa欧美va伊人久久| 丰满乱子伦码专区| 午夜日韩欧美国产| 乱人视频在线观看| 欧美日本视频| 俄罗斯特黄特色一大片| 欧美乱色亚洲激情| 一夜夜www| 韩国av一区二区三区四区| 亚洲无线在线观看| 免费av观看视频| 国产三级黄色录像| 亚洲中文日韩欧美视频| 噜噜噜噜噜久久久久久91| 免费一级毛片在线播放高清视频| 国产一区二区激情短视频| 亚洲欧美日韩高清在线视频| 十八禁网站免费在线| 亚洲人成网站高清观看| 长腿黑丝高跟| 日日夜夜操网爽| 亚洲av成人av| 日日摸夜夜添夜夜添小说| www.色视频.com| 亚洲avbb在线观看| 一区福利在线观看| 日韩欧美在线二视频| 午夜a级毛片| 精品久久久久久成人av| 国产主播在线观看一区二区| 精品久久久久久久久久久久久| 日本 av在线| 99热这里只有精品一区| 国产免费av片在线观看野外av| 熟女少妇亚洲综合色aaa.| 美女大奶头视频| e午夜精品久久久久久久| 午夜福利在线在线| 欧美黄色淫秽网站| 3wmmmm亚洲av在线观看| 国产精品 欧美亚洲| 草草在线视频免费看| 亚洲不卡免费看| 国产高清激情床上av| 亚洲久久久久久中文字幕| 欧美bdsm另类| 欧美国产日韩亚洲一区| 亚洲,欧美精品.| 日本一本二区三区精品| 少妇的逼好多水| 99精品在免费线老司机午夜| 综合色av麻豆| 欧美乱码精品一区二区三区| 免费观看人在逋| 欧美乱色亚洲激情| 丝袜美腿在线中文| 精品不卡国产一区二区三区| 男插女下体视频免费在线播放| 亚洲成人免费电影在线观看| 给我免费播放毛片高清在线观看| 熟女电影av网| 午夜精品在线福利| 亚洲 国产 在线| 99热6这里只有精品| 女人高潮潮喷娇喘18禁视频| 99久国产av精品| 在线观看一区二区三区| 国产精品永久免费网站| 国产成年人精品一区二区| 午夜精品一区二区三区免费看| 天堂动漫精品| 成人精品一区二区免费| 99国产综合亚洲精品| 在线播放无遮挡| 成人特级av手机在线观看| 国产精品永久免费网站| 精品久久久久久,| 欧美日韩一级在线毛片| 久久久久亚洲av毛片大全| aaaaa片日本免费| 国产午夜精品久久久久久一区二区三区 | www.熟女人妻精品国产| 国产精品免费一区二区三区在线| 精品一区二区三区av网在线观看| 国产一区二区在线av高清观看| 国产精品野战在线观看| 国内毛片毛片毛片毛片毛片| 欧美日韩瑟瑟在线播放| 久久天躁狠狠躁夜夜2o2o| 全区人妻精品视频| 精品一区二区三区视频在线 | 国产v大片淫在线免费观看| 色噜噜av男人的天堂激情| 欧美国产日韩亚洲一区| 亚洲精品美女久久久久99蜜臀| 免费观看精品视频网站| 国产一区二区三区在线臀色熟女| 伊人久久精品亚洲午夜| 欧美xxxx黑人xx丫x性爽| 亚洲电影在线观看av| 好男人在线观看高清免费视频| 国产精品综合久久久久久久免费| 日本撒尿小便嘘嘘汇集6| 中文字幕人妻丝袜一区二区| 成人欧美大片| 听说在线观看完整版免费高清| 91九色精品人成在线观看| 日本a在线网址| 国产乱人伦免费视频| 久久久久九九精品影院| 一进一出抽搐gif免费好疼| 99热6这里只有精品| 亚洲国产精品sss在线观看| 亚洲av电影不卡..在线观看| 午夜免费观看网址| 99久久精品一区二区三区| www.999成人在线观看| 窝窝影院91人妻| 极品教师在线免费播放| 老司机在亚洲福利影院| 国产精品嫩草影院av在线观看 | 欧美高清成人免费视频www| 午夜亚洲福利在线播放| 在线观看66精品国产| 国产亚洲精品久久久com| 午夜福利在线观看免费完整高清在 | 国产欧美日韩一区二区精品| 午夜福利在线在线| 亚洲男人的天堂狠狠| 亚洲激情在线av| 脱女人内裤的视频| 午夜亚洲福利在线播放| 九九在线视频观看精品| 狂野欧美白嫩少妇大欣赏| 亚洲欧美日韩无卡精品| 天堂影院成人在线观看| 久久久久久久午夜电影| 欧美日韩瑟瑟在线播放| 免费观看的影片在线观看| 伊人久久大香线蕉亚洲五| 99精品欧美一区二区三区四区| 国产色爽女视频免费观看| 国产国拍精品亚洲av在线观看 | 亚洲国产欧洲综合997久久,| 91九色精品人成在线观看| 免费搜索国产男女视频| 亚洲 欧美 日韩 在线 免费| 免费在线观看成人毛片| 一a级毛片在线观看| 久久人妻av系列| 丰满乱子伦码专区| 欧美不卡视频在线免费观看| 一本一本综合久久| 无遮挡黄片免费观看| 色尼玛亚洲综合影院| 国产精品一区二区三区四区久久| 日韩欧美在线乱码| 欧美黑人巨大hd| 国产免费一级a男人的天堂| e午夜精品久久久久久久| 热99在线观看视频| 看免费av毛片| 一个人免费在线观看电影| 精品不卡国产一区二区三区| 午夜激情福利司机影院| 久久这里只有精品中国| 99热精品在线国产| 51午夜福利影视在线观看| 亚洲中文字幕日韩| xxxwww97欧美| 黄色女人牲交| 成人精品一区二区免费| 啪啪无遮挡十八禁网站| 免费在线观看影片大全网站| 欧美3d第一页| 97超视频在线观看视频| netflix在线观看网站| 国产精品久久久久久人妻精品电影| 精品一区二区三区视频在线观看免费| 在线免费观看不下载黄p国产 | 2021天堂中文幕一二区在线观| 国产成人啪精品午夜网站| 亚洲熟妇熟女久久| 757午夜福利合集在线观看| 青草久久国产| 日韩成人在线观看一区二区三区| 国产 一区 欧美 日韩| 中文字幕熟女人妻在线| 麻豆成人av在线观看| 热99re8久久精品国产| 亚洲片人在线观看| a级一级毛片免费在线观看| 久久久久久久精品吃奶| 一进一出抽搐gif免费好疼| 精品午夜福利视频在线观看一区| 天堂动漫精品| 成人高潮视频无遮挡免费网站| 男人的好看免费观看在线视频| 欧美+亚洲+日韩+国产| 一卡2卡三卡四卡精品乱码亚洲| 97超视频在线观看视频| 欧美+日韩+精品| 久久性视频一级片| 成人三级黄色视频| 高潮久久久久久久久久久不卡| 日韩欧美免费精品| 国产极品精品免费视频能看的| 色视频www国产| 精品乱码久久久久久99久播| 好看av亚洲va欧美ⅴa在| 高清毛片免费观看视频网站| 午夜福利18| 国产成人a区在线观看| 观看美女的网站| 动漫黄色视频在线观看| 老司机午夜福利在线观看视频| 亚洲av美国av| 欧美乱色亚洲激情| 日韩亚洲欧美综合| 无遮挡黄片免费观看| 亚洲在线自拍视频| 国产黄色小视频在线观看| 亚洲一区二区三区色噜噜| 又爽又黄无遮挡网站| 波多野结衣高清作品| 757午夜福利合集在线观看| av女优亚洲男人天堂| 最新中文字幕久久久久| 国产精品野战在线观看| 精品久久久久久久末码| 亚洲精品一区av在线观看| 观看免费一级毛片| 我要搜黄色片| 麻豆久久精品国产亚洲av| 中出人妻视频一区二区| 哪里可以看免费的av片| 国产乱人伦免费视频| 国内精品久久久久精免费| 天天添夜夜摸| 国产精品久久久久久久电影 | 久久精品国产清高在天天线| 国产伦一二天堂av在线观看| 18禁裸乳无遮挡免费网站照片| 人人妻人人澡欧美一区二区| 国产成人a区在线观看| 最近最新中文字幕大全免费视频| 一区二区三区高清视频在线| 亚洲avbb在线观看| 成人特级av手机在线观看| 在线观看免费视频日本深夜| av片东京热男人的天堂| 亚洲激情在线av| 熟妇人妻久久中文字幕3abv| tocl精华| 久久精品国产自在天天线| 波多野结衣巨乳人妻| 18禁裸乳无遮挡免费网站照片| 亚洲美女黄片视频| 免费观看精品视频网站| 亚洲狠狠婷婷综合久久图片| 欧美不卡视频在线免费观看| 日韩有码中文字幕| 男人舔女人下体高潮全视频| 亚洲人成网站高清观看| 亚洲av日韩精品久久久久久密| 精品一区二区三区av网在线观看| 国产探花极品一区二区| 老汉色∧v一级毛片| 免费在线观看亚洲国产| 九九久久精品国产亚洲av麻豆| 九色国产91popny在线| 19禁男女啪啪无遮挡网站| 91在线观看av| 嫩草影院入口| 在线观看av片永久免费下载| 人妻久久中文字幕网| 亚洲熟妇中文字幕五十中出| 欧美乱色亚洲激情| 亚洲五月婷婷丁香| 亚洲专区中文字幕在线| 日本黄色片子视频| 婷婷六月久久综合丁香| 看片在线看免费视频| av视频在线观看入口| 亚洲精品一卡2卡三卡4卡5卡| 亚洲国产精品成人综合色| 国内精品美女久久久久久| 亚洲人成电影免费在线| 国产av在哪里看| 国内久久婷婷六月综合欲色啪| 丰满人妻一区二区三区视频av | 国产精品永久免费网站| 五月玫瑰六月丁香| 日韩国内少妇激情av| 国产欧美日韩一区二区精品| 一级a爱片免费观看的视频|