• <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 ?

    Spatial and Temporal Validation of In-Situ and Satellite Weather Data for the South West Agricultural Region of Australia

    2022-03-12 07:52:18TristanCAMPBELLandPeterFEARNS
    Journal of Meteorological Research 2022年1期
    關(guān)鍵詞:品種鑒定秈型雜交種

    Tristan CAMPBELL and Peter FEARNS

    1 Centre for Mine Site Restoration, Curtin University of Technology, WA 6102, Australia

    2 Remote Sensing and Satellite Research Group, Curtin University of Technology, WA 6102, Australia

    ABSTRACT

    Key words: Australian Water Availability Project, land surface temperature, satellite precipitation, validation

    1. Introduction

    Variations in weather conditions have a significant effect on crop yields, with simplistic crop yield models based on minimum and maximum temperature and precipitation data that are able to describe significant proportions of variances in crop yield (Lobell et al., 2007;Chen et al., 2014; Kheiri et al., 2017). Effects of weather on crop yields can be nonlinear and asymmetric, with differences in maximum temperature of as little as 2°C having a large impact on yields (Schlenker and Roberts,2009). This nonlinear and asymmetric relationship has also been demonstrated for non-cereal agricultural production, such as modeling the production of honey from mature native forests, with floral bud initiation having precise temperature thresholds and precipitation during flowering having a step-wise impact on honey harvest weight (Campbell et al., 2020).

    Due to these sensitivities, the accuracy of weather data is an important consideration in data selection for input into both qualitative and quantitative crop yield models.There are a variety of weather datasets commonly available, ranging from in-situ weather stations to satellitebased remote sensing products.

    In-situ weather stations are commonly regarded as a source of accurate data for their spatial location, but even scientific-grade weather station temperature gauges such as those used for national scale climate monitoring networks in the United States of America can have inherent instrument errors of ±0.2°C (Lin and Hubbard, 2004).Likewise, precipitation gauges have inherent limitations with, for example, wind having potentially significant effects on the volume of precipitation entering the gauge(F?rland et al., 1996). To minimize the effects of location and installation characteristics, it is important for weather station networks to have individual stations installed according to common standards, such as those defined by Canterford (1997).

    Efforts have been made to improve the ability of weather station networks to capture spatial variability. In Australia, for example, the Australian Water Availability Project (AWAP) developed national gridded datasets for rainfall, temperature, and vapour pressure that integrate topographic effects between weather stations to the interpolation process (Jones et al., 2009).

    Satellite-derived measurements of weather offer the advantage of spatially continuous datasets, which is particularly important when in-situ weather stations are sparse; however the performance can vary widely depending on the weather conditions, landform, and land use types (Mendelsohn et al., 2007).

    Retrieval of land surface temperature (LST) parameters from satellite remote sensing data primarily utilizes the thermal infrared (TIR) bands and correcting the data for atmospheric, angular, and emissivity effects (Dash et al., 2004). Validation of the Copernicus LST data, created and distributed by the European Space Agency(Martins et al., 2020) from theHimawari-8satellite operated by the Japanese Aerospace Exploration Agency(JAXA), established a root-mean-square error (RMSE)for the LST product of 2.37°C over Japan, with a bias of?0.55°C over a temperature range from ?5 to 33°C(Freitas et al., 2017). Validation studies of this product over a wider geographic region (from Japan to Australia)found similar accuracies, with an RMSE of 2.29°C and bias of ?0.70°C from Li et al. (2020) and an RMSE of 3.6°C and bias of ?0.1°C by Martins et al. (2018). This last study noted that there was an increasing positive bias for temperatures above 35°C (the temperature range in the study was from ?10 to 40°C). It is important to note that these studies relied primarily on validation against other satellite-derived LST products, with limited in-situ weather station validation.

    Satellite measurement of precipitation typically uses a combination of passive microwave (PMW) and infrared(IR) data. The Global Satellite Mapping of Precipitation(GSMaP) product produced by JAXA (Kubota et al.,2007) has used data from theHimawari-8satellite since 2015 and includes a suite of products, from near realtime (GSMaP-N), vector filtered data (GSMaP-M), and rain-gauge calibrated products (GSMaP-G). Generally,satellite precipitation products will overestimate precipitation at low rainfall rates and underestimate high rainfall rates (Habib et al., 2012; Ning et al., 2017). Rainfall type can also affect satellite precipitation accuracy, for example cold frontal rainfall events were measured less accurately in Australia than other rainfall types, which indicate a potential seasonal variance in accuracies (Chua et al., 2020).

    This study provides a detailed validation of the spatial and seasonal variations in accuracy of temperature and precipitation data from ground-based in-situ and gridded products and satellite products for the South West Agricultural Region of Australia (SWAR; Data WA, 2018).The region contains a weather station network operated by the national Bureau of Meteorology (BOM), which is used to create interpolated gridded products and to calibrate satellite products, as well as an independent weather station network operated by the state’s Department of Primary Industries and Regional Development (DPIRD,2018) to provide improved spatial coverage of weather stations for agricultural producers. While this region has been included in national-scale assessments of weather products (Chua et al., 2020), there has been limited assessment of the accuracy variability across this 350,000-km2area, which accounts for 95% of Western Australia’s$10 billion AUD in agricultural production (Australian Bureau of Statistics, 2020), despite covering less than 12% of the state’s land area.

    2. Materials and methods

    2.1 Study area

    The SWAR is the main agricultural region of Western Australia and extends almost 1000 km to the north from the south coast of the continent, and almost 1000 km to the east from the west coast (Fig. 1).

    The region covers five classifications of the K?ppen–Geiger climate classification system (Peel et al.,2007), from the hot, arid steppe environment (BSh) in the north to the warm temperate, fully humid, warm summer(Cfb) along the south coast to the cold, to the arid steppe environment (BSk) to the east. The regions are clarified by DPIRD into crop variety testing zones (CVTs) based on temperature and rainfall contours. In Fig. 1, the rainfall regions range from low in the inland areas (< 325-mm average annual rainfall, designated as “L” in Fig. 1)to very high near the coast in the southwestern quadrant of the region (> 750-mm average annual rainfall, designated as “VH” in Fig. 1). Temperature zones range from average annual temperature of greater than 28°C in the north (Zone 1) to less than 24°C in the south (Zone 5).

    While the general weather patterns for the region are warmer, drier summers and cooler, wetter winters, rainfall patterns since 1997 have seen a marked decrease in autumn, winter, and spring rainfall (Charles et al., 2010).Regional climate projections for the next few decades indicate a continued drying trend during these seasons,with increased localized summer rainfall events and overall increasing temperatures and increased likelihood of extreme temperature events (Andrys et al., 2017).

    2.2 Datasets

    The temperature and precipitation datasets for the data validation are selected based on the following characteristics:

    1) Data availability, such that users can freely access both current and historical data;2) Dataset longevity:

    ? Historic data available over a sufficient period to assess the accuracy (the period 2015–2019 for this study);

    ? Data provided by government entities with historic,current, and future data anticipated to be freely available for future users;

    3) Spatial resolution of maximum 10-km cell size for detailed analysis for individual farm use for medium to large landholders (rather than broader regional use).

    2.2.1Ground-based measurements

    Both BOM and DPIRD have weather station networks in the region with online portals for searching and retrieval of data (Bureau of Meteorology, 2021a; Campbell, 2021), thus meeting the three criteria above. With the BOM weather stations being used for the calibration of several satellite datasets, as well as the generation of inter-station gridded products, the DPIRD weather station network is selected as the baseline for the validation process for this study as the BOM weather station network is utilized in calibration of almost all the other datasets.

    2.2.2Ground-based interpolation dataset

    As mentioned in the Introduction, BOM has developed an approach for interpolation of in-situ temperature and precipitation data across Australia that incorporates some of the factors that affect these measurements between stations, such as topography (Jones et al., 2009).Grids of interpolated weather data are available for download from Bureau of Meteorology (2021b).

    Cross-validation of these datasets across Australia by Jones et al. (2009) calculated RMSEs of monthly maximum and minimum temperature between 0.5 and 1.0°C for the SWAR (national average RMSE 1.6 and 1.5°C, respectively). Cross-validation of the monthly rainfall data across the SWAR gave an RMSE between 10 and 25 mm, against a national average RMSE of 24.7 mm.

    2.2.3Satellite-derived LST

    While daily maximum and minimum LST data are available globally on scales of 1 km or less from platforms such as the Moderate Resolution Imaging Spectroradiometer (MODIS; Zhu et al., 2017), these measurements are only recorded a few times per day. As a result,they are not reliable measures of the actual daily maximum or minimum as the time of these maxima and minima do not coincide with the time of satellite overpass. Therefore, the hourly Copernicus LST dataset was used for LST (Martins et al., 2020), with the daily maxima and minima extracted and used to calculate the monthly average maximum and minimum temperature. While the spatial resolution of this dataset is coarser than products such as MODIS MOD11 LST product (5 km compared to 1 km, respectively), the significantly higher temporal resolution of the hourly Copernicus LST more reliably captures the maximum and minimum temperature. The Copernicus hourly LST data are available from European Commission Joint Research (2021).

    While validation of the Copernicus LST product for the coverage of theHimawari-8satellite gives an RMSE of less than 3°C, it has been noted by Martins et al.(2018) that there is an increasing positive bias above 35°C but this is not quantified. It is therefore expected that for this work a bias-correction would need to be applied to the LST maximum temperature data, particularly in the summer months.

    2.2.4Satellite-derived precipitation

    While there are a number of global precipitation products derived from satellite data, such as the Climate Prediction Center morphing method (CMORPH; Joyce et al., 2004) and Indian National Satellite (INSAT) Multispectral Rainfall Algorithm (IMSRA), these have resolutions of 0.25°, which are quite coarse when considering local variations in rainfall. Accordingly, the GSMaP products from the JAXA were selected for this study(Kubota et al., 2007). The GSMaP data are available from JAXA (2021). Hereafter, the near real-time GSMaP product is referred to as “GSMaP-N” and the rain gaugecorrected product as “GSMaP-G” if not qualified as the gauge—corrected product.

    2.3 Methods

    For validation of the products, the statistical methods described in Table 1 were used to compare each of the weather datasets against the DPIRD weather station data as the verification dataset, with all datasets aggregated to monthly values over the period January 2015–December 2019 inclusive. Based on the proximity principle, where a single gauge station can reflect the same rainfall pattern in the surrounding area (Wang et al., 2021), and the relatively high spatial resolution of the gridded and satellite temperature products (5-km grid/pixel size), nearest neighbor analysis was used to compare the weather datasets with the relevant DPIRD weather station data. As the objective of assessing both the BOM weather station data and the BOM gridded weather products was to determine the effectiveness of the gridding process, nearest neighbor analysis was also used for comparing the DPIRD and BOM weather stations.

    The validation was performed initially on the aggregate of all monthly data over the 2015–2019 period. The same validation statistics were generated from all monthly data over the period for each DPIRD station to assess spatial variability of accuracy for each weather dataset. The same analysis (aggregate and spatial) was also performed for each month of the year for each DPIRD station to assess seasonal variations in accuracy.

    Due to the expected mean bias (MB) of the Copernicus LST product for temperatures greater than 35°C(Martins et al., 2018), a bias-corrected Copernicus LST dataset was calculated by applying a linear bias correction. The bias correction was derived from linear regression model between the daily error of the Copernicus maximum temperature and nearest DPIRD station, and the daily maximum Copernicus temperature (see Fig. 2).An assessment of the bias for the daily minimum temperature was also performed. This assessment gave a significantly lower correlation coefficient compared to the daily maximum temperature but overall showing a strong negative bias. Accordingly, a linear bias-corrected dataset was also calculated for the Copernicus LST daily minimum temperature data. Hereafter, the bias-corrected Copernicus LST data are referred to as “Copernicus_LST_BC,” if not explicitly described as such.

    3. Results

    3.1 Statistical comparisons of accuracy

    The comparisons of statistical analysis of the temperature and precipitation products are shown in Tables 2 and 3, respectively. These summaries show that the BOM gridded products generally had the lowest errors for the three weather parameters assessed (monthly average minimum and maximum temperature, and monthly precipitation). The BOM weather stations generally performed second best out of the four data sources for each weather property, with the gridding process used byBOM reducing the MAE, RMSE, and median error in all datasets.

    Table 1. Definition of statistical methods used for validation. Note that Ei represents the estimated value at a point i, Oi is the observed value at the nearest DPIRD weather station, and N is the number of samples

    Fig. 2. Error between daily (a) maximum and (b) minimum Copernicus LST compared to DPIRD weather stations. Data are colored in a “heatmap” fashion by the concentration of datapoints.

    The uncorrected satellite derived temperature and precipitation products resulted in the highest errors for both temperature and precipitation. However, the bias correction for the Copernicus LST data applied in this study and the precipitation gauge calibration for the GSMaP data provided by JAXA reduced the errors considerably(RMSE improvements of 82% for monthly average maximum temperature, 58% for minimum temperature, and 38% for monthly precipitation). This result highlights the limitation of weather data derived purely from remote sensing data and the need for corrections using groundbased data sources.

    Table 2. Comparison of BOM and Copernicus LST monthly maximum and minimum temperature product accuracies (relative to DPIRD in-situ data). Monthly statistics from January 2015 to December 2019 inclusive. The equivalent statistics from the Australian analysis of the BOM products by Jones et al. (2009) and the hemispheric analysis of Copernicus LST by Martins et al. (2018) are provided in brackets where applicable

    Table 3. Comparison of BOM and JAXA monthly rainfall product accuracies. Monthly statistics from January 2015 to December 2019 inclusive. The equivalent statistics from the Australian analysis of the BOM products by Jones et al. (2009) and GSMaP-G by Chua et al. (2020) are provided in brackets where applicable. Note that for Chua et al. (2020), as the analysis was performed on daily precipitation, the statistics have been scaled to monthly numbers as appropriate

    The relative accuracies of the BOM and satellite products are also apparent in cross plots of these datasets against the nearest DPIRD stations, shown in Fig. 3. In all cases, the BOM gridded products have the highest proportion of datapoints on a 1 : 1 trendline compared to DPIRD data, with the BOM weather station data having the next highest proportion on this trendline. The relative performance of the satellite products is less clear in terms of the density of datapoints. The improvements of the Copernicus LST once the linear bias corrections are applied are apparent but the improvements to the GSMaP product once gauge calibrations are applied are only apparent in the summary statistics in Table 3.

    3.2 Spatial patterns of accuracy

    The average weather conditions over the 5-yr period 2015–2019 for each dataset presented in Fig. 4 show that every dataset captures the broad climatic trends across the study area as described by the K?ppen–Geiger climate zones and DPIRD CVTs. Even at this broad scale,the bias of the uncorrected Copernicus LST data is obvious, with the monthly maximum temperature overestimated and monthly minimum temperature underestimated.The lower accuracy of the non-gauge corrected GSMaP product is evident but not as apparent as the Copernicus LST temperature bias.

    The BOM gridded temperature and precipitation products and the GSMaP-G product are designed to reduce errors in weather data interpolated between weather stations. Figure 5 shows the annual RMSE for each DPIRD validation point plotted against the distance between the BOM weather station and the DPIRD validation point. There is a clear improvement in the average RMSE between the BOM weather station and BOM gridded products evidenced by the lower spread in data points and lower overall RMSE, as well as from the uncorrected to corrected Copernicus LST and GSMaP satellite products. While there is a general increase in RMSE of the BOM weather station temperature data with increasing distance from the nearest DPIRD weather station, there is not a strong correlation (Rvalues between 0.28 and 0.46). This is only a strong relationship at distances greater than 50 km. Over this distance, it is not reasonable to expect a weather station to be a reasonable measure of weather conditions without some form of interpolation.

    Fig. 3. Crossplots of BOM and satellite weather data against DPIRD station data. Monthly statistics from January 2015 to December 2019 inclusive. Data are coloured by concentration of data points. A 1 : 1 ratio line is presented on each subplot for reference.

    The gridding algorithm BOM using does lessen the errors and reduce the correlation between RMSE and distance, reducing theRvalues between 0.17 and 0.28 compared to the weather station data. It is therefore reasonable to assume that the interpolation process is reducing errors resulting from localized weather variations between weather stations.

    An assessment of the spatial variability of RMSE and bias of the weather products against the DPIRD validation weather stations shows strong spatial patterns of weather data accuracy (see Figs. 6 and 7 for these data).As expected, all BOM station metrics are improved when the data are gridded and all satellite data metrics are improved when bias correction or precipitation gauge calibration is used (for the Copernicus LST and GSMaP products, respectively).

    Fig. 4. Annual average weather data for the validation dataset (DPIRD) and the selected weather datasets. Monthly statistics from January 2015 to December 2019 inclusive.

    The spatial patterns of RMSE for the BOM weather stations and gridded products are similar, with temperature RMSE’s highest in the northern and eastern edges of the study area (where Fig. 1 shows a lower density of BOM temperature weather stations) and the precipitation less accurate in the south–west quadrant. The bias-corrected LST data show the greatest errors over and around the capital city of Perth, on the west coast of the study area (32.0°S, 115.8°E), which may indicate issues with the ability of the original LST data to adequately account for the urban heat island effect (Price, 1979). The gauge-calibrated GSMaP product has markedly higher RMSE in the south–west corner of the study area,roughly correlating with the “Very High (VH)” rainfall region as classified by DPIRD CVT zones.

    The bias patterns for the BOM products are less spatially coherent than the RMSE results. The BOM maximum temperature bias displays no coherent pattern with regards to climatic zones or land use types, although the minimum temperature bias is generally higher in areas of lower spatial density of BOM temperature weather stations. The BOM rainfall products have several “bullseye”features in the south–west quadrant, and higher bias in the eastern extent of the study area, but again no clear relationship to climatic zones.

    The spatial patterns of bias are clearer for the satellite products, with the maximum temperature generally underestimated along the Darling Range [a significant geological and topographical feature running from north of Perth to the south coast (Nemchin and Pidgeon, 1997)]and overestimated east of this. The minimum temperature is overestimated near Perth, and the gauge-corrected precipitation is generally underestimated along the western coastal area.

    3.3 Seasonal patterns of accuracy

    Fig. 5. Weather data RMSE against distance from the nearest BOM weather station. Note the different scales across different plots. Each data point represents a DPIRD station.

    The accuracy of spatial temperature products is often dependent on temperature gradients (Jones et al., 2009)and the accuracy of spatial precipitation products is often dependent on precipitation event type (Chua et al.,2020). To assess how summer versus winter weather conditions affect the spatial variability of the RMSE for each product, the RMSE was generated on a monthly basis for the study area and study period. The results for January and July are shown in Figs. 8 and 9 respectively to assess the effects during the warmest-driest–coolestwettest months of the study area. A full set of all months is presented in the supplementary data (Campbell, 2021).

    For the month of January, the BOM temperature products showed similar spatial RMSE patterns to the annual RMSE data; RMSE values generally higher in the northern and eastern extents and the gridded product having a lower RMSE than the weather station data. With the annual and January RMSE maps plotted on the same color scale, it is apparent that the maximum temperatures in January have higher RMSE in the northern and eastern zones of the study area than the annual RMSE.Areas of higher RMSE for BOM precipitation products for January are generally bullseye features. This indicates that the weather station network is unable to adequately capture localized, high intensity rainfall events,which are the primary cause of higher rainfall totals for the month of January (Sudmeyer et al., 2016).

    The bias-corrected Copernicus LST maximum January temperature product performed similarly to the annual RMSE for the northern, inland, and eastern regions of the study area, but exhibited high errors along the Darling Scarp (identified as an area of higher bias for this product in the annual data). The Copernicus LST minimum temperature product has a lower RMSE on average for January compared with annual data.

    鑫兩優(yōu)212是合肥市蜀香種子有限公司利用抗旱性較強(qiáng)的兩系不育系蜀鑫1S與抗高溫恢復(fù)系鑫恢212 2009年配組,2014年通過安徽省審定的雜交水稻新品種(審定編號(hào):皖稻2014022),經(jīng)上海市農(nóng)業(yè)生物基因中心2012年抗旱性鑒定,抗旱性達(dá)1級(jí),是國(guó)內(nèi)選育的首個(gè)雜交秈型旱稻品種[13]。筆者擬利用農(nóng)業(yè)部頒布的行業(yè)標(biāo)準(zhǔn)《水稻品種鑒定技術(shù)規(guī)程SSR標(biāo)記法》(NY/T 1433—2014)中推薦的48個(gè)SSR標(biāo)記,篩選適合的特異引物用于該雜交種純度的快速鑒定。

    The GSMaP-G product performed better in the VH rainfall region in the south–west corner but exhibited bullseye patterns of higher errors elsewhere (similar to the BOM products). Interestingly, there are some areas where the GSMaP-N product performed better than the gauge-calibrated product (along 30°S for example). This may be due to the GSMaP-G product being calibrated to weather stations that are not capturing localized precipitation events that the product is measuring, or extending localized precipitation events at weather stations.

    Fig. 6. Comparison of BOM and satellite weather data RMSE against DPIRD stations. Monthly statistics from January 2015 to December 2019 inclusive.

    The spatial pattern of errors was noticeably different in July compared to January for all weather products.BOM weather station and gridded temperature products had lower RMSE for both minimum and maximum temperature than the annual RMSE across almost the whole study area. The only exceptions to this are a few areas in the eastern region where the minimum temperature RMSE is slightly higher for July than the annual RMSE.BOM precipitation RMSE for July was higher than the annual RMSE for the very high rainfall area, and some of the high rainfall areas. There is less evidence of bullseye errors from localized events than were prevalent in the January RMSE.

    The Copernicus LST maximum temperature RMSE in July was similar to the annual RMSE for the southern region. July RMSE was lower for Perth and higher for the northern region compared with the annual RMSE. There is significantly reduced influence of the Darling Range on the RMSEs for July compared with January. Copernicus LST minimum temperature for July showed similar RMSE compared to the annual and July data.

    Fig. 7. Comparison of BOM and satellite weather data bias against DPIRD stations. Monthly statistics from January 2015 to December 2019 inclusive.

    The GSMaP-G product showed significantly higher errors in the high and very high rainfall regions of the south–west and south–east coastal regions. This is not unexpected with satellite precipitation generally underestimating higher rainfall events and less accurate for coldfrontal based precipitation (Chua et al., 2020).

    3.4 Comparative spatial and seasonal accuracy

    To directly compare the accuracy of the BOM gridded products and the calibrated/corrected satellite products (the best performing datasets from ground and satellite-derived data respectively), the relative RMSE maps in Fig. 10 were generated. These maps show the relative accuracy of the BOM grids compared to the calibrated satellite data with the DPIRD weather station network as a baseline in both cases.

    The RMSE maps are colored such that green areas show where the BOM grids have a better RMSE than the satellite data, yellow areas here where both datasets have similar RMSE and red where the satellite data have better RMSE. Temperature data were considered better where the RMSE was less than 0.25°C. Rainfall data were considered better where the RMSE was less than 5 mm. This analysis was done on an annual basis and for each month in the year. The annual, January, and July data are shown in Fig. 10 and the full set of figures is provided in the supplementary data (Campbell, 2021).

    Overall, the BOM gridded products perform better but there are some exceptions where the satellite data are more accurate, primarily for minimum and maximum temperature in the northern and eastern edges of the study area, and minimum temperature in the south–west corner and isolated areas along the south coast of the study area. These areas are all locations with relatively sparse BOM temperature weather stations (Fig. 1).

    Fig. 8. Comparison of BOM and satellite weather data RMSE against DPIRD stations for January (austral summer). January statistics from January 2015 to January 2019 inclusive.

    Apart from some very localized points, the BOM gridded rainfall product performs similarly to better than the GSMaP-G product across the study area.

    4. Discussion

    4.1 Copernicus LST bias correction

    With several validation studies completed for the Copernicus LST product in-situ weather station data, mostly showing a consistent RMSE of 2–3°C (Freitas et al.,2017; Martins et al., 2018, 2020), the size of the positive bias of this dataset at higher temperature was an unexpected result (bias corrections in the order of 20°C for weather station temperature in the order of 40°C). The effectiveness of the linear bias correction for higher temperature (82% improvement in RMSE for average monthly maximum temperature) was likewise unexpected.

    Fig. 9. Comparison of BOM and satellite weather data RMSE against DPIRD stations for July (austral winter). July statistics from July 2015 to July 2019 inclusive.

    The Copernicus LST product is validated against both in-situ weather stations (Martins et al., 2018), and for regional and biome assessment, the MODIS LST product(Wan, 2015). While globally the LST product has a strong linear 1 : 1 relationship for temperature ranging from ?30 to 50°C for both in-situ weather stations and the MODIS LST product, the comparison with the in-situ Tateno weather station in Japan (the only in-situ validation point for theHimawari-8-based LST) shows a clear increasing positive bias for in-situ temperature measurements above approximately 30°C (Martins et al., 2018).This positive bias trend was not present when comparing the Copernicus LST data to the MODIS LST data in the same validation study (Martins et al., 2018). However,multiple comparisons of MODIS daytime LST and insitu weather station data have found that MODIS daytime LST can be overestimated by 5–20°C in some ecosystems (Vancutsem et al., 2010; Lin et al., 2012; Assiri,2017; Phan et al., 2019). Therefore, the MODIS daytime LST dataset may not be a reliable product for validation of the Copernicus LST dataset at higher temperature.

    To assess the degree and consistency of bias of the Copernicus LST product at higher temperature in different climatic zones and ecosystems, Copernicus LST data were compared to BOM weather station temperature data for a series of locations in the northern (warmer) half of Australia. The result of this process, shown alongside the selected BOM weather stations in Fig. 11, again showed an almost linear relationship between increasing temperature and increasing positive bias for the Copernicus LST data. While theR2value is lower, at 0.73, for this analysis versus 0.94 for the main study area of this paper, the same overall pattern is present.

    Fig. 10. Performance of BOM gridded products and satellite products versus DPIRD data for annual, January, and July. Green areas are where only the BOM gridded product has an RMSE at least 0.25°C or 5 mm lower than the equivalent satellite product; red areas are where only the satellite product has an RMSE at least 0.25°C or 5 mm lower than the equivalent; and yellow areas are where the RMSEs of both the BOM gridded product and the relevant satellite product have a difference of less than 0.25°C or 5 mm.

    Applying the correction factor shown in Fig. 11 to the Copernicus northern Australian LST data improves the RMSE from 15.7 to 2.8°C (an improvement of 82%, almost exactly the same improvement ratio for the SWAR). Applying the SWAR correction factor to the northern Australian data reduced the RMSE to 3.0°C. Although an improved result, the difference in RMSE for the different areas indicates that the bias correction may be improved with localized adjustments. Creating a bias correction factor for each weather station in the northern Australian region resulted in a further slight improvement to an RMSE of 2.4°C (an 84% reduction in RMSE compared to the original Copernicus LST product).

    4.2 Effect of land use type and landscape heterogeneity

    With land use type and heterogeneity reported to affect both satellite-derived LST (Martins et al., 2018) and satellite-derived precipitation (Chua et al., 2020), the average RMSE for the monthly data at each DPIRD validation point was compared against land use data for the biascorrected Copernicus LST data and gauge-corrected GSMaP product.

    Land use classifications by ABARES (2001) are extracted for a radius of 2.5 km around each DPIRD validation point (corresponding to a similar area to the Copernicus LST and GSMaP pixels) in Fig. 12. Land use heterogeneity is calculated from the number of different land use classes within this region and the land use type from the land use type covering the largest area within this region. Of the 126 DPIRD validation points, 2 are in urban areas, 4 in conservation regions, 18 in managed forest regions, 26 in pasture grazing regions, and 76 in broadacre cropping regions.

    The results from this process, shown in Fig. 13, indicate that land use heterogeneity has no bearing on the Copernicus LST_BC accuracy or the GSMaP-G product.However, land use type does impact on monthly temperature and precipitation accuracy.

    Fig. 11. Comparison of Copernicus LST and BOM weather station daily maximum temperature data for northern Australia. (a) Locations of BOM weather stations over the K?ppen–Geiger climate zone and (b) LST errors versus Copernicus LST temperatures (equivalent plot to Fig. 2).

    While most of the errors for the LST_BC maximum and minimum temperature products had RMSEs between 0.5 and 1.5°C, managed forest and urban land use types all had errors in some cases higher than the cropping and grazing land use types. More than 90% of the grazing and cropping land use locations had an RMSE of less than 1.5°C for the maximum temperature and 1.0°C for the minimum temperature. While land use heterogeneity based on land use classes was not correlated to temperature accuracy (scatter plots), the variability in vegetation cover for the forest land uses and multiple land cover types within the urban land use classification likely reduces the accuracy of the Copernicus LST_BC product in these regions.

    Precipitation accuracy was similar for all land use types except conservation, which had no locations with an average RMSE better than 15 mm month?1. Precipitation accuracies for grazing and cropping land use types were slightly better than for managed forest land.

    5. Conclusions

    From the accuracy analysis of both ground and satellite-based weather products for the SWAR in this study from January 2015 to December 2019, the BOM gridded temperature and rainfall products provide the most accurate weather data in most regions and months (average monthly temperature RMSE of 0.9°C and monthly precipitation RMSE of 14.0 mm), followed by the BOM weather station data (RMSEs of 1.1°C and 20.0 mm).Calibrated or bias-corrected satellite data generally performed third best (RMSEs of 1.4°C and 23.8 mm) and uncalibrated or uncorrected satellite data were generally the least accurate (RMSEs of 6.0°C and 38.5 mm). These average accuracies are similar to those found in other studies of these weather products in larger and other regions.

    A strong linear bias was found for the daily Copernicus LST product. This bias ranged from approximately?5°C at daily maximum LST of approximately 10°C to over 20°C at LST of 60°C. A linear correction of bias based on the daily maximum LST greatly improved the accuracy of the daily maximum LST from an RMSE of 9.25 to 1.7°C (an improvement of 82%). An assessment of the bias correction from other regions in Australia outside the SWAR found that this linear bias pattern was present in multiple climatic regions although the relative scale of the bias did vary to some degree. For the weather stations outside of the SWAR used for this analysis, applying the SWAR correction factor improved the RMSE from 15.7 to 3.0°C. Creating a bias correction factor for each weather station resulted in a further improvement to an RMSE of 2.4°C (an 84% reduction in RMSE compared with the original Copernicus LST product). The success of this relatively simple bias correction process indicates that a bias-corrected Copernicus LST product is a feasible option, similar to the precipitation gauge corrected GSMaP product.

    Fig. 12. DPIRD weather stations in the SWAR overlaid on land use types.

    Fig. 13. Scatter plots: Land use heterogeneity (number of different land use types within a 25-pixel region) versus satellite weather product accuracy (LST RMSE and GSMaP-G rainfall). Higher number of pixels implies more heterogeneous land use. Histograms: Cumulative proportion of DPIRD sites for increasing RMSE for the five different primary land use classifications.

    The accuracy of the weather products showed both spatial and seasonal varibility. The BOM weather station and gridded products were less accurate in summer at the northern and eastern extents of the study area (where the density of BOM weather stations is markedly reduced).While the bias-corrected satellite temperatures were more accurate in these regions in summer, this product had a marked increase in RMSE in summer over Darling Fault,the main topographic feature of the study area. Minimum temperature was generally better represented by the BOM weather products with the exception of a region along the southern coastal region of the study area.

    The BOM precipitation products were less accurate in the very high rainfall zones (> 750 mm yr?1) in the south–west corner of the study area. While the satellite precipitation products did achieve similar accuracies to the BOM products in some areas and seasons, there was no location or month where the satellite precipitation products were markedly more accurate than the BOM precipitation products.

    In addition to mapping the spatial and seasonal accuracy of each weather product, a series of maps have been generated to highlight regions and months where the BOM gridded products are more or less accurate than the satellite weather products. This allows users of weather data to both understand the accuracy limitations of weather data for different locations as well as select the best performing weather product based on their location and time of year.

    6. Data availability

    Monthly DPIRD weather station data used for this analysis are available in Comma Separated Variable format from https://doi.org/10.25917/fry7-nx79 (Campbell, 2021). Maps of statistical analysis on an aggregated annual and monthly basis are also available from this location.

    猜你喜歡
    品種鑒定秈型雜交種
    廣適秈型兩系不育系7011S
    早秈型香稻三系不育系禾5A的選育與應(yīng)用
    多舉措促玉米雜交種制提質(zhì)增產(chǎn)
    提高桑蠶一代雜交種雜交率的方法和措施
    雜交種子為什么不能留種
    植物DNA條形碼鑒定研究進(jìn)展
    國(guó)家2015年審定通過的水稻新品種(上)
    廣西壯族自治區(qū)2015年審定通過的水稻新品種(上)
    藏藥鑒定及質(zhì)量控制研究現(xiàn)狀
    NA指紋在水稻品種鑒定中的應(yīng)用與展望
    国内精品久久久久精免费| 一个人看的www免费观看视频| 亚洲自偷自拍三级| 免费av观看视频| 午夜老司机福利剧场| 国产一区二区在线观看日韩| 变态另类成人亚洲欧美熟女| 又爽又黄无遮挡网站| 久久精品影院6| 国产成人freesex在线| 秋霞在线观看毛片| 欧美一区二区国产精品久久精品| 插逼视频在线观看| 精品午夜福利在线看| 国产精品,欧美在线| 又爽又黄a免费视频| or卡值多少钱| 久久久久性生活片| 天堂√8在线中文| 日韩国内少妇激情av| 简卡轻食公司| 国产精品日韩av在线免费观看| 久久精品91蜜桃| 午夜精品国产一区二区电影 | 国产一区二区亚洲精品在线观看| 国产精品乱码一区二三区的特点| 欧美另类亚洲清纯唯美| 亚洲内射少妇av| 亚洲欧美日韩东京热| 久久久国产成人精品二区| 久久久精品94久久精品| 国产精品一区二区性色av| 国产精品久久久久久久久免| 日韩强制内射视频| 国产成人精品久久久久久| 国产乱人偷精品视频| 搡老妇女老女人老熟妇| 日本五十路高清| 成人欧美大片| 麻豆国产av国片精品| av免费在线看不卡| 人体艺术视频欧美日本| 别揉我奶头 嗯啊视频| 色播亚洲综合网| 日本色播在线视频| 免费看光身美女| 一边亲一边摸免费视频| 国产免费一级a男人的天堂| 波多野结衣巨乳人妻| 国产精品美女特级片免费视频播放器| 国产精品人妻久久久影院| 国产极品精品免费视频能看的| 天堂影院成人在线观看| 国产一区二区激情短视频| 国产高潮美女av| 亚洲欧美精品综合久久99| 变态另类丝袜制服| 亚洲第一区二区三区不卡| www.av在线官网国产| 亚洲av不卡在线观看| 男女视频在线观看网站免费| 爱豆传媒免费全集在线观看| 日韩欧美 国产精品| 赤兔流量卡办理| 国产精品伦人一区二区| 我要看日韩黄色一级片| 亚洲成人久久性| 亚洲人成网站在线观看播放| 青春草国产在线视频 | 精品人妻偷拍中文字幕| 蜜桃久久精品国产亚洲av| 亚洲婷婷狠狠爱综合网| 久久久色成人| 久久精品夜夜夜夜夜久久蜜豆| 黄色日韩在线| 中文字幕人妻熟人妻熟丝袜美| 国产片特级美女逼逼视频| 色哟哟·www| 日日啪夜夜撸| 人妻夜夜爽99麻豆av| 欧美一级a爱片免费观看看| 国产麻豆成人av免费视频| 国产一区二区三区在线臀色熟女| 男女下面进入的视频免费午夜| 夫妻性生交免费视频一级片| 国产精品永久免费网站| 毛片女人毛片| 中国国产av一级| 久久精品国产亚洲av香蕉五月| 国内精品美女久久久久久| 欧美一级a爱片免费观看看| 国产精品日韩av在线免费观看| 免费看a级黄色片| 国产高清不卡午夜福利| 欧美一级a爱片免费观看看| 亚洲av电影不卡..在线观看| 国产精品久久久久久av不卡| 又粗又硬又长又爽又黄的视频 | 婷婷亚洲欧美| 欧美最黄视频在线播放免费| 久久人人爽人人片av| 国产精品国产三级国产av玫瑰| 欧美日韩综合久久久久久| 成人综合一区亚洲| 天堂影院成人在线观看| 国产精品久久久久久久久免| 一夜夜www| 男人舔女人下体高潮全视频| 麻豆成人av视频| 麻豆精品久久久久久蜜桃| 乱人视频在线观看| 久久99蜜桃精品久久| 不卡一级毛片| 久久人妻av系列| 欧美zozozo另类| 99久久中文字幕三级久久日本| 少妇丰满av| 波野结衣二区三区在线| 日韩,欧美,国产一区二区三区 | 免费观看a级毛片全部| 欧美三级亚洲精品| 寂寞人妻少妇视频99o| 国产精品一区二区性色av| .国产精品久久| 伦理电影大哥的女人| 免费看光身美女| 国产伦在线观看视频一区| 99热精品在线国产| 99热这里只有是精品50| 中出人妻视频一区二区| 日日摸夜夜添夜夜爱| 色综合亚洲欧美另类图片| 亚洲电影在线观看av| 可以在线观看毛片的网站| 日韩高清综合在线| 日韩三级伦理在线观看| 国产高清不卡午夜福利| 精品人妻一区二区三区麻豆| 色综合站精品国产| av天堂在线播放| 不卡一级毛片| 亚洲自拍偷在线| 五月伊人婷婷丁香| 91av网一区二区| 大型黄色视频在线免费观看| 日韩欧美 国产精品| 日本成人三级电影网站| 乱码一卡2卡4卡精品| 亚洲久久久久久中文字幕| 国产久久久一区二区三区| 久久精品国产99精品国产亚洲性色| 亚洲精品影视一区二区三区av| 天美传媒精品一区二区| 免费大片18禁| 亚洲av一区综合| 中国国产av一级| 中文欧美无线码| 国产伦精品一区二区三区四那| 亚洲欧美精品综合久久99| 中文字幕制服av| 一级av片app| 欧美色视频一区免费| 女人十人毛片免费观看3o分钟| 69av精品久久久久久| 免费人成视频x8x8入口观看| 亚洲七黄色美女视频| АⅤ资源中文在线天堂| 亚洲熟妇中文字幕五十中出| 免费不卡的大黄色大毛片视频在线观看 | 亚洲av熟女| 天天躁日日操中文字幕| av免费在线看不卡| 免费不卡的大黄色大毛片视频在线观看 | 亚洲精品久久久久久婷婷小说 | 欧美一区二区精品小视频在线| 国产综合懂色| 麻豆国产av国片精品| 毛片一级片免费看久久久久| 亚洲乱码一区二区免费版| 天堂av国产一区二区熟女人妻| 成年av动漫网址| 亚洲第一电影网av| 一级av片app| 人妻久久中文字幕网| 桃色一区二区三区在线观看| 性欧美人与动物交配| 亚洲av不卡在线观看| 国产av一区在线观看免费| 国产成人精品一,二区 | 岛国毛片在线播放| 免费无遮挡裸体视频| 波野结衣二区三区在线| 日韩欧美 国产精品| 人妻夜夜爽99麻豆av| 在线观看免费视频日本深夜| 亚洲精品日韩av片在线观看| av在线观看视频网站免费| 人体艺术视频欧美日本| 伦理电影大哥的女人| 一个人看的www免费观看视频| 亚洲av二区三区四区| 成人特级黄色片久久久久久久| av福利片在线观看| 12—13女人毛片做爰片一| 亚洲色图av天堂| 一个人看视频在线观看www免费| 国产成人一区二区在线| 亚洲av中文字字幕乱码综合| 亚洲三级黄色毛片| 最近最新中文字幕大全电影3| 国产精品永久免费网站| 国产v大片淫在线免费观看| 美女高潮的动态| 91精品国产九色| 女的被弄到高潮叫床怎么办| 一边摸一边抽搐一进一小说| 最近手机中文字幕大全| 亚洲最大成人手机在线| 日韩欧美精品免费久久| 久久午夜亚洲精品久久| 午夜激情福利司机影院| 亚洲人与动物交配视频| 男人和女人高潮做爰伦理| 99国产精品一区二区蜜桃av| 国产精品爽爽va在线观看网站| 欧美性感艳星| 国产成人a∨麻豆精品| 国产精品嫩草影院av在线观看| 国产成人aa在线观看| 最近2019中文字幕mv第一页| 久久人人爽人人片av| 中文字幕精品亚洲无线码一区| 日韩高清综合在线| 日本爱情动作片www.在线观看| 男女做爰动态图高潮gif福利片| 亚洲精品成人久久久久久| 黄片wwwwww| 黄片wwwwww| 中文字幕av成人在线电影| 黄色欧美视频在线观看| 日本av手机在线免费观看| 又粗又硬又长又爽又黄的视频 | 联通29元200g的流量卡| 日韩在线高清观看一区二区三区| 99热精品在线国产| 日日摸夜夜添夜夜添av毛片| 国产白丝娇喘喷水9色精品| 在线a可以看的网站| 日韩在线高清观看一区二区三区| 国产在线精品亚洲第一网站| 亚洲欧美日韩高清专用| 亚洲精品国产av成人精品| 国产视频首页在线观看| 久久99蜜桃精品久久| 男的添女的下面高潮视频| 一级黄色大片毛片| 日韩一区二区视频免费看| 亚洲av电影不卡..在线观看| 国内精品久久久久精免费| 久久久久久久亚洲中文字幕| 免费av观看视频| 老熟妇乱子伦视频在线观看| 你懂的网址亚洲精品在线观看 | 最近的中文字幕免费完整| 婷婷六月久久综合丁香| 欧美极品一区二区三区四区| 午夜免费激情av| 国产真实乱freesex| 一本久久精品| 欧美zozozo另类| 校园人妻丝袜中文字幕| 激情 狠狠 欧美| 激情 狠狠 欧美| 久久久久久久久久黄片| 久久这里只有精品中国| 男人舔女人下体高潮全视频| 边亲边吃奶的免费视频| 国产成人午夜福利电影在线观看| 免费观看精品视频网站| 欧美日韩国产亚洲二区| 欧美三级亚洲精品| 国产三级在线视频| 日韩中字成人| 久久久久久久午夜电影| а√天堂www在线а√下载| 久久精品夜色国产| av在线老鸭窝| 我的老师免费观看完整版| 国产成人91sexporn| 男插女下体视频免费在线播放| 97热精品久久久久久| 国产黄色视频一区二区在线观看 | 国产黄色小视频在线观看| 一级黄片播放器| 99久国产av精品| 日韩高清综合在线| 午夜老司机福利剧场| 国产高清有码在线观看视频| 欧美成人一区二区免费高清观看| 麻豆国产av国片精品| 91精品国产九色| 美女高潮的动态| 欧美成人精品欧美一级黄| 51国产日韩欧美| 乱系列少妇在线播放| 欧美性猛交╳xxx乱大交人| 久久精品国产清高在天天线| 神马国产精品三级电影在线观看| 国产三级中文精品| 深夜a级毛片| 99久久中文字幕三级久久日本| 97在线视频观看| 久久6这里有精品| 日韩亚洲欧美综合| 国产69精品久久久久777片| 欧美色欧美亚洲另类二区| 在线观看午夜福利视频| 日本免费a在线| АⅤ资源中文在线天堂| 青春草亚洲视频在线观看| 春色校园在线视频观看| 午夜激情福利司机影院| 久久人妻av系列| 美女脱内裤让男人舔精品视频 | 欧美高清性xxxxhd video| a级一级毛片免费在线观看| 亚洲av一区综合| 狂野欧美白嫩少妇大欣赏| 啦啦啦韩国在线观看视频| 1000部很黄的大片| www日本黄色视频网| 午夜亚洲福利在线播放| 日本撒尿小便嘘嘘汇集6| 国产在视频线在精品| 麻豆av噜噜一区二区三区| 听说在线观看完整版免费高清| 91久久精品国产一区二区三区| 欧美日韩乱码在线| 国产精华一区二区三区| 成年av动漫网址| 亚洲性久久影院| 中文字幕av成人在线电影| 丰满的人妻完整版| 97在线视频观看| 亚洲欧美清纯卡通| 国产精品久久电影中文字幕| 久久久久久久久中文| 中文亚洲av片在线观看爽| 麻豆久久精品国产亚洲av| 99久久精品热视频| 听说在线观看完整版免费高清| 九九热线精品视视频播放| 男女下面进入的视频免费午夜| 26uuu在线亚洲综合色| 搡女人真爽免费视频火全软件| 久久人妻av系列| 国产成人精品一,二区 | 夜夜夜夜夜久久久久| 欧美成人免费av一区二区三区| 国产探花在线观看一区二区| 国产成人午夜福利电影在线观看| 别揉我奶头 嗯啊视频| 亚洲av中文字字幕乱码综合| 少妇熟女aⅴ在线视频| 人妻夜夜爽99麻豆av| 一级二级三级毛片免费看| 国产av在哪里看| 国产白丝娇喘喷水9色精品| 看非洲黑人一级黄片| 国产成人a∨麻豆精品| 亚洲第一电影网av| 亚洲第一区二区三区不卡| 国产男人的电影天堂91| 亚洲成人久久爱视频| 99热这里只有是精品50| 国产精品久久久久久亚洲av鲁大| 国产精品女同一区二区软件| 国产久久久一区二区三区| 寂寞人妻少妇视频99o| 一级毛片久久久久久久久女| 人妻久久中文字幕网| 人体艺术视频欧美日本| 成人漫画全彩无遮挡| 高清在线视频一区二区三区 | av在线亚洲专区| 欧美极品一区二区三区四区| 尾随美女入室| 91aial.com中文字幕在线观看| av卡一久久| 久久久a久久爽久久v久久| 精品人妻偷拍中文字幕| 国产高清有码在线观看视频| 国产成人精品一,二区 | 夜夜爽天天搞| 国内精品久久久久精免费| 国产精品.久久久| 婷婷色综合大香蕉| 亚洲欧美日韩卡通动漫| 久久久精品大字幕| 久久99精品国语久久久| 国产成人影院久久av| 国产av麻豆久久久久久久| 国产一区二区在线观看日韩| 亚洲五月天丁香| 久久久a久久爽久久v久久| 国产免费一级a男人的天堂| a级毛片a级免费在线| 欧美成人免费av一区二区三区| 全区人妻精品视频| 青春草亚洲视频在线观看| 久久久a久久爽久久v久久| 成人毛片a级毛片在线播放| 国产精品国产三级国产av玫瑰| 久久久国产成人免费| 欧美潮喷喷水| 国产午夜精品久久久久久一区二区三区| 亚洲av第一区精品v没综合| 久久精品国产亚洲av天美| 天堂中文最新版在线下载 | 国产精品一区二区在线观看99 | 毛片一级片免费看久久久久| 亚洲av成人精品一区久久| 国产极品天堂在线| 精品久久久久久久末码| 久久亚洲国产成人精品v| 日韩三级伦理在线观看| 国产69精品久久久久777片| 久久精品91蜜桃| 日韩强制内射视频| 日本一本二区三区精品| 午夜福利在线在线| 99九九线精品视频在线观看视频| av女优亚洲男人天堂| 精品久久久久久久久久免费视频| 大型黄色视频在线免费观看| 欧美一区二区亚洲| 精品人妻熟女av久视频| 久久久久免费精品人妻一区二区| 91aial.com中文字幕在线观看| 国产亚洲av嫩草精品影院| 日韩国内少妇激情av| 又粗又硬又长又爽又黄的视频 | 亚洲精品乱码久久久v下载方式| 女人十人毛片免费观看3o分钟| 麻豆国产av国片精品| 中文资源天堂在线| 免费看a级黄色片| 嫩草影院新地址| 校园春色视频在线观看| 久久久久国产网址| 国产高清激情床上av| 精品久久久久久久末码| 级片在线观看| 亚洲欧美精品综合久久99| 亚洲经典国产精华液单| av在线天堂中文字幕| 免费人成在线观看视频色| 久久久久久伊人网av| 99热全是精品| 悠悠久久av| 高清毛片免费观看视频网站| 久久久久久久久久成人| 欧美日本亚洲视频在线播放| 欧美色视频一区免费| av天堂中文字幕网| 免费人成视频x8x8入口观看| 亚洲,欧美,日韩| 久久久精品94久久精品| 男女那种视频在线观看| 蜜臀久久99精品久久宅男| 亚洲四区av| 成人亚洲欧美一区二区av| 男女视频在线观看网站免费| 99热精品在线国产| 成人三级黄色视频| 在线观看午夜福利视频| 国产伦一二天堂av在线观看| 丰满的人妻完整版| 三级毛片av免费| 亚洲色图av天堂| 精品久久国产蜜桃| 国内久久婷婷六月综合欲色啪| 午夜福利在线在线| 日本一二三区视频观看| 色5月婷婷丁香| 一区福利在线观看| 秋霞在线观看毛片| 精品熟女少妇av免费看| 国产黄片视频在线免费观看| 欧美潮喷喷水| 国产蜜桃级精品一区二区三区| 亚洲一区高清亚洲精品| 欧美日韩一区二区视频在线观看视频在线 | 蜜桃亚洲精品一区二区三区| 欧美精品一区二区大全| 高清毛片免费观看视频网站| 欧美不卡视频在线免费观看| 国产成人精品婷婷| 国产亚洲欧美98| 日本黄色视频三级网站网址| 一个人免费在线观看电影| 午夜精品一区二区三区免费看| 精品一区二区免费观看| 国产单亲对白刺激| 日本五十路高清| 国产亚洲精品久久久com| 亚洲色图av天堂| 嫩草影院新地址| 天堂av国产一区二区熟女人妻| 久久婷婷人人爽人人干人人爱| 欧美色欧美亚洲另类二区| 久久精品91蜜桃| 日本爱情动作片www.在线观看| 91精品一卡2卡3卡4卡| 麻豆国产97在线/欧美| 又黄又爽又刺激的免费视频.| 国产成人91sexporn| 九九爱精品视频在线观看| 日本五十路高清| 一边摸一边抽搐一进一小说| 亚洲在久久综合| 一级毛片电影观看 | 能在线免费看毛片的网站| 夫妻性生交免费视频一级片| 能在线免费看毛片的网站| 亚洲av二区三区四区| 欧美在线一区亚洲| 亚洲性久久影院| 一级毛片电影观看 | 国产精品久久久久久av不卡| 久久久欧美国产精品| 五月玫瑰六月丁香| 欧洲精品卡2卡3卡4卡5卡区| 少妇人妻一区二区三区视频| 99热6这里只有精品| 欧美精品一区二区大全| 午夜福利视频1000在线观看| av天堂中文字幕网| 日本免费a在线| 日日摸夜夜添夜夜添av毛片| 最近视频中文字幕2019在线8| 亚洲三级黄色毛片| 亚州av有码| 久久人妻av系列| 一区福利在线观看| 最近2019中文字幕mv第一页| 国产精品综合久久久久久久免费| 久久九九热精品免费| 国内精品美女久久久久久| 欧美色视频一区免费| 国产一级毛片七仙女欲春2| 欧美成人一区二区免费高清观看| 婷婷精品国产亚洲av| 一个人观看的视频www高清免费观看| 国产精品不卡视频一区二区| 五月玫瑰六月丁香| 国产综合懂色| 亚洲精品成人久久久久久| 亚洲欧美精品综合久久99| 国产极品天堂在线| 晚上一个人看的免费电影| 午夜a级毛片| 蜜桃亚洲精品一区二区三区| 深夜a级毛片| 婷婷色av中文字幕| 久久精品久久久久久噜噜老黄 | 久久草成人影院| 此物有八面人人有两片| 久久人人爽人人片av| 日韩欧美精品v在线| 直男gayav资源| 97超视频在线观看视频| 日本一本二区三区精品| 亚洲国产欧美在线一区| 欧美一级a爱片免费观看看| 变态另类丝袜制服| 久久久欧美国产精品| 偷拍熟女少妇极品色| 欧美+日韩+精品| 狠狠狠狠99中文字幕| 白带黄色成豆腐渣| 亚洲四区av| 国产精品99久久久久久久久| 91在线精品国自产拍蜜月| 免费av不卡在线播放| 美女cb高潮喷水在线观看| 免费观看在线日韩| 免费观看的影片在线观看| 亚洲电影在线观看av| 日本一二三区视频观看| 欧美区成人在线视频| 亚洲va在线va天堂va国产| 99视频精品全部免费 在线| 18禁裸乳无遮挡免费网站照片| 夜夜看夜夜爽夜夜摸| 美女cb高潮喷水在线观看| 亚洲电影在线观看av| 久久九九热精品免费| 日本黄大片高清| 亚洲欧美中文字幕日韩二区| 午夜福利高清视频| 久久精品夜色国产| 蜜臀久久99精品久久宅男| 又爽又黄无遮挡网站| 成人三级黄色视频| 人妻久久中文字幕网| 国内揄拍国产精品人妻在线| 久久久久久久久久久丰满| 日日撸夜夜添| 亚洲在线自拍视频| 99在线人妻在线中文字幕| 免费一级毛片在线播放高清视频| 淫秽高清视频在线观看| 欧美高清成人免费视频www| 18禁在线播放成人免费|