The dataset contains phenological camera observation data collected at the Arou Superstation in the midstream of the Heihe integrated observatory network from June 13 to November 16, 2018. The instrument was developed with data processed by Beijing Normal University. The phenomenon camera integrates data acquisition and data transmission functions. The camera captures high-quality data with a resolution of 1280×720 by looking-downward. The calculation of the greenness index and phenology are following 3 steps: (1) calculate the relative greenness index (GCC, Green Chromatic Coordinate, calculated by GCC=G/(R+G+B)) according to the region of interest, (2) perform gap-filling for the invalid values, filtering and smoothing, and (3) determine the key phenological parameters according to the growth curve fitting (such as the growth season start date, Peak, growth season end, etc.) There are also 3 steps for coverage data processing: (1) select images with less intense illumination, (2) divide the image into vegetation and soil, and (3) calculate the proportion of vegetation pixels in each image in the calculation area. After the time series data is extracted, the original coverage data is smoothed and filtered according to the time window specified by the user, and the filtered result is the final time series coverage. This data set includes relative greenness index (Gcc). Please refer to Liu et al. (2018) for sites information in the Citation section.

This dataset contains the flux measurements from site No.16 eddy covariance system (EC) in the flux observation matrix from 6 June to 17 September, 2012. The site (100.36411° E, 38.84931° N) was located in a cropland (maize surface) in Daman irrigation district, which is near Zhangye, Gansu Province. The elevation is 1564.31 m. The EC was installed at a height of 4.9 m; the sampling rate was 10 Hz. The sonic anemometer faced north, and the separation distance between the sonic anemometer and the CO2/H2O gas analyzer (Gill&Li7500) was 0.2 m. Raw data acquired at 10 Hz were processed using the Eddypro post-processing software (Li-Cor Company, http://www.licor.com/env/products/ eddy_covariance/software.html), including spike detection, lag correction of H2O/CO2 relative to the vertical wind component, sonic virtual temperature correction, angle of attack correction, coordinate rotation (2-D rotation), corrections for density fluctuation (Webb-Pearman-Leuning correction), and frequency response correction. The EC data were subsequently averaged over 30 min periods. Moreover, the observation data quality was divided into three classes according to the quality assessment method of stationarity (Δst) and the integral turbulent characteristics test (ITC), which was proposed by Foken and Wichura [1996]: class 1 (level 0: Δst<30 and ITC<30), class 2 (level 1: Δst<100 and ITC<100), and class 3 (level 2: Δst>100 and ITC>100), representing high-, medium-, and low-quality data, respectively. In addition to the above processing steps, the half-hourly flux data were screened in a four-step procedure: (1) data from periods of sensor malfunction were rejected; (2) data before or after 1 h of precipitation were rejected; (3) incomplete 30 min data were rejected when the missing data constituted more than 3% of the 30 min raw record; and (4) data were rejected at night when the friction velocity (u*) was less than 0.1 m/s. There were 48 records per day; the missing data were replaced with -6999. Moreover, suspicious data were marked in red. The released data contained the following variables: data/time, wind direction (Wdir, °), wind speed (Wnd, m/s), the standard deviation of the lateral wind (Std_Uy, m/s), virtual temperature (Tv, ℃), H2O mass density (H2O, g/m^3), CO2 mass density (CO2, mg/m^3), friction velocity (ustar, m/s), stability (z/L), sensible heat flux (Hs, W/m^2), latent heat flux (LE, W/m^2), carbon dioxide flux (Fc, mg/ (m^2s)), quality assessment of the sensible heat flux (QA_Hs), quality assessment of the latent heat flux (QA_LE), and quality assessment of the carbon flux (QA_Fc). In this dataset, the time of 0:30 corresponds to the average data for the period between 0:00 and 0:30; the data were stored in *.xlsx format. For more information, please refer to Liu et al. (2016) (for multi-scale observation experiment or sites information), Xu et al. (2013) (for data processing) in the Citation section.

This dataset contains the flux measurements from site No.7 eddy covariance system (EC) in the flux observation matrix from 29 May to 18 September, 2012. The site (100.36521° E, 38.87676° N) was located in a cropland (maize surface) in Yingke irrigation district, which is near Zhangye, Gansu Province. The elevation is 1556.39 m. The EC was installed at a height of 3.8 m; the sampling rate was 10 Hz. The sonic anemometer faced north, and the separation distance between the sonic anemometer and the CO2/H2O gas analyzer (CSAT3&Li7500A) was 0.15 m. Raw data acquired at 10 Hz were processed using the Edire post-processing software (University of Edinburgh, http://www.geos.ed.ac.uk/abs/research/micromet/EdiRe/), including spike detection, lag correction of H2O/CO2 relative to the vertical wind component, sonic virtual temperature correction, coordinate rotation (2-D rotation), corrections for density fluctuation (Webb-Pearman-Leuning correction), and frequency response correction. The EC data were subsequently averaged over 30 min periods. Moreover, the observation data quality was divided into three classes according to the quality assessment method of stationarity (Δst) and the integral turbulent characteristics test (ITC), which was proposed by Foken and Wichura [1996]: class 1 (level 0: Δst<30 and ITC<30), class 2 (level 1: Δst<100 and ITC<100), and class 3 (level 2: Δst>100 and ITC>100), representing high-, medium-, and low-quality data, respectively. In addition to the above processing steps, the half-hourly flux data were screened in a four-step procedure: (1) data from periods of sensor malfunction were rejected; (2) data before or after 1 h of precipitation were rejected; (3) incomplete 30 min data were rejected when the missing data constituted more than 3% of the 30 min raw record; and (4) data were rejected at night when the friction velocity (u*) was less than 0.1 m/s. There were 48 records per day; the missing data were replaced with -6999. Moreover, suspicious data were marked in red. The released data contained the following variables: data/time, wind direction (Wdir, °), wind speed (Wnd, m/s), the standard deviation of the lateral wind (Std_Uy, m/s), virtual temperature (Tv, ℃), H2O mass density (H2O, g/m^3), CO2 mass density (CO2, mg/m^3), friction velocity (ustar, m/s), stability (z/L), sensible heat flux (Hs, W/m^2), latent heat flux (LE, W/m^2), carbon dioxide flux (Fc, mg/ (m^2s)), quality assessment of the sensible heat flux (QA_Hs), quality assessment of the latent heat flux (QA_LE), and quality assessment of the carbon flux (QA_Fc). In this dataset, the time of 0:30 corresponds to the average data for the period between 0:00 and 0:30; the data were stored in *.xlsx format. For more information, please refer to Liu et al. (2016) (for multi-scale observation experiment or sites information), Xu et al. (2013) (for data processing) in the Citation section.

Snow is a significant component of the ecosystem and water resources in high-mountain Asia (HMA). Therefore, accurate, continuous, and long-term snow monitoring is indispensable for the water resources management and economic development. The present study improves the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard Terra and Aqua satellites 8 d (“d” denotes “day”) composite snow cover Collection 6 (C6) products, named MOD10A2.006 (Terra) and MYD10A2.006 (Aqua), for HMA with a multistep approach. The primary purpose of this study was to reduce uncertainty in the Terra–Aqua MODIS snow cover products and generate a combined snow cover product. For reducing underestimation mainly caused by cloud cover, we used seasonal, temporal, and spatial filters. For reducing overestimation caused by MODIS sensors, we combined Terra and Aqua MODIS snow cover products, considering snow only if a pixel represents snow in both the products; otherwise it is classified as no snow, unlike some previous studies which consider snow if any of the Terra or Aqua product identifies snow. Our methodology generates a new product which removes a significant amount of uncertainty in Terra and Aqua MODIS 8 d composite C6 products comprising 46 % overestimation and 3.66 % underestimation, mainly caused by sensor limitations and cloud cover, respectively. The results were validated using Landsat 8 data, both for winter and summer at 20 well-distributed sites in the study area. Our validated adopted methodology improved accuracy by 10 % on average, compared to Landsat data. The final product covers the period from 2002 to 2018, comprising a combination of snow and glaciers created by merging Randolph Glacier Inventory version 6.0 (RGI 6.0) separated as debris-covered and debris-free with the final snow product MOYDGL06*. We have processed approximately 746 images of both Terra and Aqua MODIS snow containing approximately 100 000 satellite individual images. Furthermore, this product can serve as a valuable input dataset for hydrological and glaciological modelling to assess the melt contribution of snow-covered areas. The data, which can be used in various climatological and water-related studies, are available for end users at https://doi.org/10.1594/PANGAEA.901821 (Muhammad and Thapa, 2019).

This dataset is the Fractional Vegetation Cover observation in the artificial oasis experimental region of the middle stream of the Heihe River Basin. The observations lasted for a vegetation growth cycle from May 2012 to September 2012 (UTC+8). Instruments and measurement method: Digital photography measurement is implemented to measure the FVC. Plot positions, photographic method and data processing method are dedicatedly designed. Details are described in the following: 0. In field measurements, a long stick with the camera mounted on one end is beneficial to conveniently measure various species of vegetation, enabling a larger area to be photographed with a smaller field of view. The stick can be used to change the camera height; a fixed-focus camera can be placed at the end of the instrument platform at the front end of the support bar, and the camera can be operated by remote control. 1. For row crop like corn, the plot is set to be 10×10 m2, and for the orchard, plot scale is 30×30 m2. Shoot 9 times along two perpendicularly crossed rectangular-belt transects. The picture generated of each time is used to calculate a FVC value. “True FVC” of the plot is then acquired as the average of these 9 FVC values. 2. The photographic method used depends on the species of vegetation and planting pattern: Low crops (<2 m) in rows in a situation with a small field of view (<30 ), rows of more than two cycles should be included in the field of view, and the side length of the image should be parallel to the row. If there are no more than two complete cycles, then information regarding row spacing and plant spacing are required. The FVC of the entire cycle, that is, the FVC of the quadrat, can be obtained from the number of rows included in the field of view. 3. High vegetation in rows (>2 m) Through the top-down photography of the low vegetation underneath the crown and the bottom-up photography beneath the tree crown, the FVC within the crown projection area can be obtained by weighting the FVC obtained from the two images. Next, the low vegetation between the trees is photographed, and the FVC that does not lie within the crown projection area is calculated. Finally, the average area of the tree crown is obtained using the tree crown projection method. The ratio of the crown projection area to the area outside the projection is calculated based on row spacing, and the FVC of the quadrat is obtained by weighting. 4. FVC extraction from the classification of digital images. Many methods are available to extract the FVC from digital images, and the degree of automation and the precision of identification are important factors that affect the efficiency of field measurements. This method, which is proposed by the authors, has the advantages of a simple algorithm, a high degree of automation and high precision, as well as ease of operation.

This data set includes a monthly composite of 30 m × 30 m surface vegetation coverage products in the Qilian Mountain Area in 2019. In this paper, the maximum value composition (MVC) method is used to synthesize monthly NDVI products and calculate FVC by using the reflectance data of Landsat 8 and sentinel 2 red and near infrared channels. The data is monthly synthesized by Google Earth engine cloud platform, and the index is calculated by the model. The missing pixels are interpolated with good quality, which can be used in environmental change monitoring and other fields.

Zhangye basin mainly includes 20 irrigation areas. Under the restriction of water diversion, the surface water consumption of the irrigation area is under control, but the groundwater exploitation is increased, resulting in the groundwater level drop in the middle reaches, resulting in potential ecological environment risks. Due to the complex and frequent exchange of surface water and groundwater in the study area, it is possible to realize the overall water resource saving by optimizing the utilization ratio of surface water and groundwater in each irrigation area. In this project, on the premise of not changing the water demand of the middle reaches irrigation area, the two problems of maximizing the outflow of Zhengyi Gorge (given groundwater reserve constraint) and maximizing the outflow of Zhengyi Gorge (given groundwater reserve constraint) are studied.

ALOS PRISM dataset includes 13 scenes; one covers the A'rou foci experimental area on Mar. 19, 2008, one covers the Haichaoba on Mar. 19, 2008, one covers the Biandukou foci experimental area on Apr. 17, 2008, and one covers the Linze grassland and Linze station foci experimental areas on Apr. 22, 2008. The data version is LB2, which was released after radiometric correction and geometric correction.

This data is from the central station of environmental monitoring in gansu province. The data includes three observation elements, namely sulfur dioxide, nitrogen dioxide and inhalable particles, which are published on the network. The data format is a text file. The first column is the city name, the second column is sulfur dioxide, the third column is nitrogen dioxide, the fourth column is pm10, and the fifth column is the observation date. The data included lanzhou, jiayuguan, jinchang, baiyin, tianshui, qingyang, pingliang, dingxi, longnan, wuwei, zhangye, jiuquan and linxia. This data will be updated automatically and continuously according to the data source.

Data Overview: Zhangye's channels are divided into five levels: dry, branch, bucket, agricultural and Mao channels, of which the agricultural channels are generally unlined. Mao channels are field projects, so the three levels of dry, branch and bucket channels and a small part of agricultural channels are mainly collected. The irrigation canal system data includes 2 main canals (involving multiple irrigation districts), 157 main canals (within a single irrigation district), 782 branch canals and 5315 dou canals, with a total length of 8, 745.0km. Data acquisition process: remote sensing interpretation and GPS field measurement are adopted for data acquisition of irrigation canal system. Direct GPS acquisition channel is the most effective method, but the workload of GPS acquisition channel is too large, and we only verify the measurement in some irrigation areas. The main method is to first collect the manual maps of irrigation districts drawn by each water pipe. Most of these maps have no location, only some irrigation districts such as Daman and Shangsan have been located based on topographic maps, and some irrigation districts in Gaotai County have used GPS to locate some channels. Referring to the schematic diagram of the irrigation district, channel spatial positioning is carried out based on Quikbird, ASTER, TM remote sensing images and 1: 50000 topographic maps. For the main canal and branch canal, due to the obvious linear features on remote sensing images and the general signs on topographic maps, it can be located more accurately. For Douqu, areas with high-resolution images can be located more accurately, while other areas can only be roughly located according to fuzzy linear features of images and prompt information of irrigation district staff, with low positioning accuracy. Each water management office simultaneously provides channel attribute data, which is one-to-one corresponding to spatial data. After the first draft of the channel distribution map is completed, it is submitted twice to the personnel familiar with the channel distribution of each water pipe for correction. The first time is mainly to eliminate duplication and leak, and the second time is mainly to correct the position and perfect the attribute data. Description of data content: The fields in the attribute table include code, district and county name, irrigation area name, channel whole process, channel name, channel type, location, total length, lined, design flow, design farmland, design forest and grass, real irrigation farmland, real irrigation forest and grass, water right area, and remarks. Code example: G06G02Z15D01, where the first letter represents the county name, the 2nd and 3rd numbers represent the county (district) number, the 4th to 6th characters represent the trunk canal code, the 7th to 9th characters represent the branch canal code, and the 10th to 12th characters represent the dou canal code.

In 2007 and 2008, Landsat data set 49 scenes, covering the entire black river basin. The acquisition time is:2007-08-12, 2007-09-23, 2008-01-05, 2008-02-06, 2008-03-17, 2008-03-25, 2008-05-10, 2008-05-19, 2008-05-28, 2008-06-04, 2008-07-07, 2008-07-15, 2008-07-22, 2008-07-23, 2008-08-16, 2008-08-30,2008-09-08, 2008-09-15, 2008-09-17, 2008-10-01, 2008-10-10, 2008-10-19, 2008-10-26, 2008-11-02, 2008-11-04, 2008-11-18, 2008-11-20, 2008-11-27, 2008-12-06, 2008-12-13, 2008-12-14. The product is class L1 and has been geometrically corrected.It includes 4 scenes of TM image and 45 scenes of ETM+ image. The Landsat satellite remote sensing data set of heihe integrated remote sensing joint experiment was obtained through free download.

The Landsat TM Mosaic Image of the Heihe River Basin can be effectively applied to monitoring land-use change of the basin, which reflects the current situation of the Heihe River Basin in 2010, and provides a reliable basis for ecological planning and restoration. This mosaic image collected the TM images released by the USGS for free in 2010 (data from July to September 2010, totally 21 scenes, the maximum cloud amount is less than 10%), and the preprocessed images were geometrically registered by topographic maps(polynomial geometry correction method), then a geometrically-corrected digital mosaic map was generated, which was of high quality after a certain accuracy evaluation. The images were stored in ERDAS IMG format, and the most abundant bands 5, 4 and 3 combination, with three colors: red, green, and blue were selected to generate a color composite image. The combined composite image not only is similar to natural color, which is more in accordance with people's visual habits, but also can fully display the differences in image features because of the rich amount of information.