CN-121723409-B - Mountain land air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing

CN121723409BCN 121723409 BCN121723409 BCN 121723409BCN-121723409-B

Abstract

The invention belongs to the technical field of meteorological remote sensing and geographic information, in particular to a mountain land air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing, which comprises the steps of obtaining multi-source satellite earth surface temperature product data, meteorological site observation data and multidimensional auxiliary factor data; the method comprises the steps of constructing an air temperature estimation model based on different observation scenes of multi-source satellite earth surface temperature product data, constructing the air temperature estimation model based on a random forest regression model, training the air temperature estimation model by using weather site observation data to generate a plurality of sets of air temperature estimation initial value products, carrying out weighted fusion on the plurality of sets of air temperature estimation initial value products according to deviation of each set of air temperature estimation initial value products relative to weather site observation data to generate fusion air temperature data, and carrying out space-time continuity interpolation filling on vacancies existing in the fusion air temperature data to obtain a mountain air temperature data set with space-time continuity.

Inventors

XIE XINYAO
ZHAO WEI
WANG YONG
ZHAO JUNLI

Assignees

中国科学院、水利部成都山地灾害与环境研究所

Dates

Publication Date: 20260508
Application Date: 20260224

Claims (7)

1. The mountain land air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing is characterized by comprising the following steps of: Acquiring multi-source satellite earth surface temperature product data, meteorological site observation data and multi-dimensional auxiliary factor data; the multidimensional cofactor data comprises vegetation index data and topography factor data; The vegetation index data are obtained by calculating a normalized vegetation index, an enhanced vegetation index, a soil adjustment vegetation index and a difference vegetation index; the topographic factor data is obtained by extracting altitude, gradient, slope direction and topographic relief information in the digital elevation model data; Constructing an air temperature estimation model based on different observation scenes of the multi-source satellite earth surface temperature product data, wherein the air temperature estimation model is constructed based on a random forest regression model, and the multi-dimensional auxiliary factor data is input into the air temperature estimation model; Training the air temperature estimation model by utilizing the meteorological site observation data to generate a plurality of sets of air temperature estimation initial value products; According to the deviation of each set of air temperature estimation initial value products relative to the observation data of the meteorological site, weighting and fusing the sets of air temperature estimation initial value products to generate fused air temperature data; and filling the gaps existing in the fused air temperature data by space-time continuous interpolation to obtain a space-time continuous mountain air temperature data set.
2. The mountain land air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing according to claim 1, wherein the obtaining of multi-source satellite earth surface temperature product data comprises: Obtaining at least two surface temperature products from ERS-2, ENVISAT, TERRA MODIS, AQUA MODIS, sentinel-3A, and multisensor fusion products; and uniformly resampling the acquired earth surface temperature products to target spatial resolution, and acquiring the data of the earth surface temperature products of the multi-source satellite.
3. The method of claim 1, wherein the obtaining vegetation index data by calculating a normalized vegetation index, an enhanced vegetation index, a soil-adjusted vegetation index, and a differential vegetation index comprises: calculating the normalized vegetation index NDVI includes: ; Calculating the enhanced vegetation index EVI includes: ; Wherein G is a gain coefficient, and C 1 、C 2 is a coefficient of the reflectivity values of the red wave band and the blue wave band respectively; Calculating the soil conditioning vegetation index SAVI includes: ; Wherein, the Is a soil regulating factor; calculating the differential vegetation index DVI comprises: ; Wherein, the 、、 Respectively the reflectivity values of the near infrared band, the red band and the blue band.
4. The method for estimating the long time sequence high resolution of the mountain land air temperature based on the multi-source satellite remote sensing according to claim 1, wherein the obtaining of the topography factor data by extracting the altitude, gradient, slope direction and topography relief information in the digital elevation model data comprises: extracting the Slope includes: ; Extracting the slope Aspect includes: ; Wherein, the Indicating the rate of change of elevation z in the east-west direction x, Indicating the rate of change of elevation z in the north-south direction y, Representing the rate of change of elevation y in the east-west direction x; extracting the relief of topography TRI comprises: ; Wherein, the For an elevation of eight neighborhood pixels, Is the elevation of the center pixel.
5. The mountain land air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing according to claim 1, wherein the random forest regression model is independently constructed in different seasons, and model parameters of spring, summer, autumn and winter are respectively established.
6. The mountain land air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing according to claim 5, wherein constructing an air temperature estimation model based on different observation scenes of the multi-source satellite earth surface temperature product data comprises: The different observation scenes comprise a day and night double full observation scene, a night observation scene and a day observation scene; constructing an air temperature estimation model which simultaneously utilizes night ground surface temperature and daytime ground surface temperature for the day and night double-full observation scene; constructing an air temperature estimation model only using night ground surface temperature for the night observation scene; And constructing an air temperature estimation model only using the daytime surface temperature for the daytime observation scene.
7. The mountain air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing according to claim 6, wherein the weighting fusion of the sets of air temperature estimation initial value products comprises: the weighted fusion adopts an air temperature estimation initial value product corresponding to a day and night double full observation scene; when the products of the night and day double-full observation scene are missing, estimating an initial value product by adopting the air temperature corresponding to the night observation scene; when products of the day and night double-full observation scene and the night observation scene are both absent, the initial value product is estimated by adopting the air temperature corresponding to the day observation scene.

Description

Mountain land air temperature long time sequence high resolution estimation method based on multi-source satellite remote sensing Technical Field The invention belongs to the technical field of meteorological remote sensing and geographic information, and particularly relates to a mountain land air temperature long time sequence high-resolution estimation method based on multi-source satellite remote sensing. Background Mountainous areas are an important component of the earth's system, being an important pool of fresh water resources and hot spots of biodiversity, and are highly sensitive to climate change. Near-surface air Temperature (TA) is a fundamental variable for understanding the climate influence of the mountain ecosystem. However, due to the complex mountain terrain, existing high altitude meteorological sites are sparse and unevenly distributed, and it is difficult to capture strong microclimate changes within short distances. While reanalyzed products (e.g., ERA5-Land, MERRA-2) provide global coverage, their spatial resolution (typically 0.1℃or greater) is difficult to meet with the demands of mountain area fine research. The main means for acquiring the mountain air temperature at present and the problems are as follows: and (3) observing weather sites, namely sparsely observing the weather sites in high-altitude areas, and difficultly capturing complex space thermodynamic changes. The analysis data, such as ERA5, is that the product covers the world, but the spatial resolution is coarse (usually more than 10 km), and the mountain microclimate cannot be finely described. Satellite surface temperature (LST) based estimates that have high spatial resolution (e.g., 1 km) are effective proxy variables for estimating air temperature. The existing LST-TA conversion method mainly comprises the following steps: Temperature-vegetation index (TVX) assuming that the canopy temperature of an infinitely thick vegetation is close to air temperature. The defect is that in a vegetation sparse area (such as a mountain bare land), the vegetation sparse area is greatly disturbed by soil humidity, and the precision is reduced. The surface energy balance physical model needs a large amount of parameter calibration and is difficult to apply in mountain areas with insufficient data. Establishing LST-TA relationships based on statistical regression or machine learning is the dominant method of generating large scale air temperature products. However, the prior art has the following major drawbacks: depending on a single data source, the existing products usually depend on a single satellite sensor (such as MODIS), are seriously affected by cloud coverage, cause a large number of space-time deletions in cloudy mountain regions, and cannot provide space-time continuous data. There is a lack of a multisource fusion mechanism-although long-time-series LST observations from different satellite platforms (e.g., ERS-2, ENVISAT, sentinel-3, etc.) currently exist, there is a systematic bias in the LST-air temperature relationship of the different source data due to significant differences in transit time, observation geometry, and sensitivity to ground-air coupling of the different sensors. The prior art lacks an effective strategy for integrating complementary multi-source observation information, and a global mountain long-time sequence air temperature data set capable of effectively combining multi-source satellite observation is not available at present. Under-utilization of day-night difference, the existing method is easy to use day-night average LST, the characteristic that night LST is higher than daytime LST and air temperature is usually considered, and a hierarchical processing mechanism is not designed for the condition of single day/night data missing. Therefore, a global mountain air temperature estimation method capable of integrating multi-source satellite observation data, solving the problem of cloud coverage deficiency through a priority weighted fusion strategy, generating long time sequence, high resolution (1 km) and space-time continuity is needed. Disclosure of Invention In order to solve the technical problems, the invention provides a mountain air temperature long time sequence high-resolution estimation method based on multi-source satellite remote sensing, which aims to realize continuous estimation of global mountain long time sequence, high precision and high spatial resolution month average air temperature by constructing nonlinear relation models under different observation scenes (double night and day, night and day only), introducing NDVI, EVI, SAVI, DVI and other vegetation indexes and fine terrain factors such as altitude, gradient and slope direction and combining a weighting fusion strategy based on priority with a 3D space-time filling technology. In order to achieve the above purpose, the present invention provides a mountain land air temperature long time sequence high resolution estim