A method for evaluating the spatial representativeness of ground vegetation phenological observation sites

Abstract

The present invention discloses a method for evaluating the spatial representativeness of ground vegetation phenological observation sites. Before verifying remote sensing phenological products, the proportion of the main vegetation coverage area MVP, the multi-date average relative sill coefficient MRCS, and the average instantaneous phase synchronization MIPS within the coarse pixel scale are calculated based on high-resolution remote sensing image data to evaluate the spatial representativeness of the ground vegetation phenological observation sites within the coarse pixel scale. Then, all ground phenological observation sites are classified, and the ground phenological observation sites with higher spatial representativeness are selected. The evaluation process of the present invention is scientific and reasonable, with high verification accuracy and reliability, simple calculation and low implementation difficulty. By classifying the ground vegetation phenological observation sites and selecting the ground vegetation phenological observation sites with higher spatial representativeness, it lays a data foundation for the authenticity verification of satellite remote sensing phenological products, making the verification results more reliable.

Description

Technical Field

[0001] The present invention relates to the field of authenticity verification of satellite remote sensing technology and products, and particularly relates to an algorithm for evaluating the spatial representativeness of ground vegetation phenology observation sites. Background Art

[0002] With the enhancement of the ability of remote sensing technology to detect land surface phenology (LSP, Land Surface Phenology) at regional, continental and global scales, it has become possible to study phenological changes related to climate change. Before using remote sensing phenology products for research, it is necessary to verify the authenticity of remote sensing phenology products to evaluate the quality and reliability of remote sensing phenology products. At present, the most commonly used evaluation method is the evaluation based on the observation data of ground phenology stations.

[0003] However, since remote sensing extracts phenology mainly based on the information of the vegetation canopy within pixels, while ground station phenology observations are of the phenological characteristics of single plants or specific species, there is a mismatch in the observation scales between the two. Therefore, before using ground phenology observation stations to verify satellite remote sensing phenology products, it is necessary to evaluate the spatial representativeness of the stations. "Spatial representativeness" refers to the extent to which the data obtained at a specific spatial scale can reflect the real situation at different spatial scales. Relevant research shows that in areas with stronger ground homogeneity of vegetation cover (such as deciduous forests), the phenological results of ground observations and satellite monitoring are generally basically the same, indicating that the landscape heterogeneity of ground stations is an important source of errors and uncertainties in verifying remote sensing phenology products. If the underlying surface of the ground phenology observation station is relatively homogeneous and has strong spatial representativeness, then the station can well verify satellite remote sensing phenology products and reduce the uncertainty in the process of verifying and evaluating satellite remote sensing phenology products. Therefore, it is crucial to evaluate the spatial representativeness of the commonly used ground phenology stations at present.

[0004] At present, the evaluation of spatial representativeness mainly focuses on directions such as surface albedo, land surface temperature (LST), leaf area index, etc., and mainly only focuses on the representativeness research at a single temporal phase spatial scale. There are few reports on the evaluation of the spatial representativeness of real ground phenology observation stations. The evaluation of ground phenology observation stations is more complex than other evaluations because it not only needs to consider the surface homogeneity and spatial variability at a certain time point, but also needs to judge whether the temporal changes of each pixel within the pixel scale of satellite remote sensing phenology are consistent with the temporal changes of the pixel where the station is located, that is, to judge whether the growth trends of the vegetation are consistent, so as to ensure representativeness throughout the vegetation growth process, rather than simply using the spatial variability at a certain moment for evaluation. Therefore, it is necessary to design an algorithm for evaluating the spatial representativeness of ground vegetation phenology observation stations for verifying remote sensing phenology products according to the mechanism of satellite remote sensing vegetation phenology monitoring. Summary of the Invention

[0005] The present invention provides a method for evaluating the spatial representativeness of ground vegetation phenological observation sites. This method can be used to evaluate the spatial representativeness of global ground vegetation phenological observation sites and classify them, and select sites with higher quality to verify global satellite remote sensing phenological products. It is an algorithm for quality evaluation and screening of sites before verifying satellite remote sensing phenological products.

[0006] An algorithm for evaluating the spatial representativeness of ground vegetation phenological observation sites used to verify remote sensing phenological products. Before verifying remote sensing phenological products, calculate the proportion of the main vegetation coverage area (MVP, Main Vegetation Proportion), the multi-date average relative sill coefficient (MRCS, Mean Relative Coefficient of Sill), and the mean instantaneous phase synchrony (MIPS, Mean Instantaneous Phase Synchrony) within the coarse pixel scale based on high-resolution remote sensing image data to evaluate the spatial representativeness of ground vegetation phenological observation sites within the coarse pixel scale. Then classify all ground phenological observation sites and select ground phenological observation sites with higher spatial representativeness, indicating good spatial representativeness within this coarse pixel scale, and then verify remote sensing phenological products.

[0007] Based on the mechanism of satellite remote sensing for monitoring vegetation phenology and considering the particularity of remotely sensed extraction of vegetation phenology, the present invention proposes evaluation indicators and systems suitable for evaluating the spatial representativeness of ground phenological observation sites, fully considering ground homogeneity, spatial heterogeneity, and pixel time synchrony, and evaluating and classifying ground phenological observation sites. The evaluation process of the present invention is scientific and reasonable, with high verification accuracy and reliability, simple calculation, and low implementation difficulty. Classify ground vegetation phenological observation sites, select ground vegetation phenological observation sites with higher spatial representativeness, lay a data foundation for the authenticity verification of satellite remote sensing phenological products, make the verification results more reliable, provide a guarantee for the improvement of satellite remote sensing phenological products, thereby promoting the production of more accurate land surface phenological products, and is of great significance for deeply understanding the land surface water and heat processes, carbon cycle processes under the background of global climate change, and predicting the spatio-temporal changes of terrestrial ecosystems.

[0008] Calculate the proportion of the main vegetation coverage area (MVP, Main Vegetation Proportion) through the following formula:

[0009]

[0010] In the formula, MVP (Main Vegetation Proportion) is the proportion of the largest vegetation area in the area within the site neighborhood. A(V) is the area of each type of vegetation within the site neighborhood, mainly including forest land, grassland, and farmland. n is the total number of land cover types within the neighborhood, and A(i) is the area of each type of land cover within the site neighborhood.

[0011] The multi-date average relative sill coefficient (MRCS, Mean Relative Coefficient of Sill) algorithm is calculated through the following process, which mainly includes:

[0012] 1.1) Obtain all pixel pairs within the area of the ground phenological observation site corresponding to the corresponding date from multi-temporal Landsat 8 remote sensing image data

[0013] 1.2) Fit the variogram values of all pixel pairs using the semi-variogram model respectively to obtain the key parameters reflecting the surface spatial heterogeneity: the sill value sill of the d-th date d

[0014] 1.3) Calculate the multi-date average relative sill coefficient (MRCS, Mean Relative Coefficient of Sill) according to the sill value. The specific formula is as follows:

[0015]

[0016] In the formula, MRCS is the multi-date average relative sill coefficient (Mean Relative Coefficient of Sill), n is the total number of dates, sill d is the sill value of the d-th date, is the mean value within the area of the site neighborhood on the d-th date.

[0017] The mean instantaneous phase synchrony (MIPS, Mean Instantaneous Phase Synchrony) algorithm is calculated through the following process, which mainly includes:

[0018] 2.1) Perform Hilbert transform on the original time series information obtained based on high-resolution remote sensing images to obtain the analytical information, and calculate the instantaneous phase of all times between the pixels within the area of the ground phenological observation site and the pixel where the ground phenological observation site is located. The calculation formula is:

[0019] z a (t) = z(t) + jΗ{z(t)}

[0020] where \(z(t)\) is the initial time series, \(\mathcal{H}\) represents the Hilbert transform applied to it, and \(z\) a (t) is a sequence constructed based on the initial time series, and this equation can also be expressed as follows:

[0021] z a (t)=a(t)e jφ(t)

[0022] where \(a(t)\) is the instantaneous envelope and \(\varphi(t)\) is the instantaneous phase.

[0023] 2.2) Calculate the instantaneous phase difference at all times between the pixels within the neighborhood of the ground phenological observation site and the pixel where the ground phenological observation site is located. The specific formula is as follows:

[0024] \(\Delta\varphi(t)=\varphi\) p (t)-\(\varphi\) s (t)

[0025] 2.3) Represent the synchrony using the sine value of the instantaneous phase difference between the pixels within the neighborhood of the phenological observation site and the pixel where the ground phenological observation site is located, and calculate the IPS ps (t) at time \(t\) for the time series curve of the pixels within the neighborhood and the time series curve of the pixel where the site is located. The calculation formula is:

[0026] IPS ps (t)=1 - abs(sin(\(\varphi\) p (t)-\(\varphi\) s (t)))

[0027] where IPS ps (t) is the instantaneous phase synchrony value at time \(t\) between the pixel where the site is located and a certain pixel within the neighborhood, \(\varphi\) p (t) is the instantaneous phase of a certain pixel within the neighborhood at time \(t\), and \(\varphi\) s (t) is the instantaneous phase of the pixel where the site is located at time \(t\). When the phases of the two time series curves are the same at a certain moment, IPS ps (t)=1.

[0028] 2.4) Calculate the average value of the instantaneous phase synchrony (IPS, Instantaneous Phase Synchrony) at all times between all pixels within the neighborhood and the pixel where the site is located to evaluate the consistency of the time series within the neighborhood. The specific formula is as follows:

[0029]

[0030] where MIPS psIt is the Mean Instantaneous Phase Synchrony (MIPS). m is the number of pixels within the neighborhood range around the site, and n is the number of valid dates, which represents the temporal synchrony between all pixels within the neighborhood range of the site and the pixel where the site is located.

[0031] The present invention mainly includes the following steps:

[0032] Step 1) Data acquisition: Acquire ground vegetation phenology observation sites to be evaluated, remote sensing phenology products, high-spatial-resolution land cover fine products, and Landsat 8 remote sensing image data.

[0033] Step 2) Based on the existing high-spatial-resolution land cover fine products, calculate the proportion of the main vegetation coverage area (MVP, Main Vegetation Proportion) of the ground phenology observation site within the scale of the remote sensing phenology product. Since even in areas with uniform land cover, there may be significant spatial variation characteristics in phenology, calculate the Mean Relative Coefficient of Sill (MRCS) index based on Landsat 8 remote sensing image data to measure the overall heterogeneity in the spatial dimension of the ground phenology observation site. For sites with a relatively high degree of heterogeneity, calculate the Mean Instantaneous Phase Synchrony (MIPS, Mean Instantaneous Phase Synchrony) between the pixels within its neighborhood range and the pixel where the site is located to further evaluate the spatial representativeness of the site.

[0034] Step 3) Use the natural break method to determine the thresholds of the relevant indicators in Step 2) and divide the sites, and classify all ground phenology observation sites. Then use several indicators such as the Pearson correlation coefficient (including significance test), slope, RMSE, BIAS, Abs.BIAS, etc. to analyze the site division results to evaluate the effect of site level division.

[0035] In the described Step 1), select the commonly used dataset of ground vegetation phenology observation sites. For the remote sensing phenology products, select pixels with relatively high confidence after strict quality control. The spatial resolution of the land cover fine product is 30m. The spatial resolution of the Landsat 8 remote sensing image data is 30m, and calculate vegetation indices such as NDVI, EVI, EVI2, etc. to calculate the indicators for evaluating the variation.

[0036] In step 2), based on the existing high-spatial-resolution land cover fine products, evaluate the land cover homogeneity of the site within the scale of the remote sensing phenology product, and calculate the proportion of the main vegetation cover area (MVP, Main Vegetation Proportion) of the ground phenology observation site within the scale of the remote sensing phenology product. The calculation formula of MVP is as follows:

[0037]

[0038] MVP (Main Vegetation Proportion) is the proportion of the vegetation area with the largest area within the neighborhood of the site. A(V) is the area of each type of vegetation within the neighborhood of the site, mainly including forest land, grassland, and farmland. n is the total number of land cover types within the neighborhood, and A(i) is the area of each type of land cover within the neighborhood of the site. The larger the MVP, the more homogeneous the surface within the neighborhood of the site. Select sites with MVP greater than 0.6 for classification.

[0039] Since even in areas with homogeneous land cover, there may be significant spatial variation characteristics in phenology, by using the semivariogram model to fit the variogram values between all pixel pairs, key parameters reflecting the surface spatial heterogeneity are obtained. Commonly used semivariogram models mainly include the spherical model, exponential model, and Gaussian model, etc. Generally, the spherical model is selected to calculate the sill value. As the distance between pixel pairs increases, the variogram value continuously increases. When it reaches a stable level, the variogram value is a relatively stable constant, and this constant is the sill value, which is used to evaluate the magnitude of spatial variability within the neighborhood of the site. Since the observed values of different sites at different times may be different, the multi-date Mean Relative Coefficient of Sill (MRCS) is proposed to reflect the magnitude of relative spatial heterogeneity. The calculation formula is as follows:

[0040]

[0041] In the formula, MRCS is the multi-date Mean Relative Coefficient of Sill, n is the total number of dates, sill d is the sill value of the d-th date, is the mean value within the neighborhood of the site on the d-th date. The smaller the MRCS, the smaller the degree of spatial heterogeneity within the neighborhood of the site, and vice versa.

[0042] Analyzing from the mechanism of extracting phenology from satellite remote sensing, in the process of using remote sensing to extract phenology, the method of determining phenological indicators based on time series is greatly affected by the peak, slope, curvature, etc. of the time series. Therefore, before using ground stations to verify satellite remote sensing phenology products, it is necessary to consider the synchrony of the time series of the pixel where the station is located and other pixels in the neighborhood range. Therefore, the Mean Instantaneous Phase Synchrony (MIPS) index is adopted to reflect whether the increase or decrease of the time series of the vegetation index is synchronous, so as to further evaluate the synchrony of all pixels in the neighborhood range of the station and the pixel where the station is located in terms of time series.

[0043] First, perform the Hilbert transform on the real time series signal to obtain the analytical information.

[0044] z a (t) = z(t) + jΗ{z(t)}

[0045] In the formula, z(t) is the initial time series, Η represents performing the Hilbert transform on it, and z a (t) is the sequence constructed based on the initial time series, and this formula can also be expressed as follows:

[0046] z a (t) = a(t)e jφ(t)

[0047] a(t) is the instantaneous envelope, and φ(t) is the instantaneous phase.

[0048] According to the above two formulas, the instantaneous phase at a specific moment can be calculated. The instantaneous phase difference between two time series at time t can be expressed as Δφ(t) = φ p (t) - φ s (t), and the synchrony is represented by the sine value of the phase difference. The consistency between the time series curves of the pixels in the neighborhood range and the time series curve of the pixel where the station is located at time t can be expressed by the following formula:

[0049] IPS ps (t) = 1 - abs(sin(φ p (t) - φ s (t)))

[0050] In the formula, IPS ps (t) is the instantaneous phase synchrony value between the pixel where the station is located and a certain pixel in the neighborhood range at time t, φ p (t) is the instantaneous phase of a certain pixel in the neighborhood range at time t, and φ s (t) is the instantaneous phase of the pixel where the station is located at time t. When the phases of the two time series curves are consistent at a certain moment, IPS ps (t) = 1.

[0051] Evaluate the consistency of the time series within the neighborhood range using the average IPS of all pixels within the neighborhood range and all moments of the pixel where the site is located:

[0052]

[0053] In the formula, MIPS ps is the Mean Instantaneous Phase Synchrony (MIPS), m is the number of pixels within the neighborhood range around the site, and n is the number of valid dates, which represents the temporal synchrony of all pixels within the neighborhood range of the site and the pixel where the site is located.

[0054] In step 3) described above, according to the natural break method and the verification effect of the above indicators, a hierarchical method is adopted to classify all ground phenological observation sites. The specific classification process is as follows: When the MVP within the scale of the satellite remote sensing phenology product of the site is equal to 1, it indicates that the land cover around the site is uniform. The MRCS index is used for evaluation. If the MRCS is small, it indicates that the ground space of the site is homogeneous, and the spatial representativeness of this type of site is good, and it is classified as a first-level site. For sites with a larger MRCS, the MIPS index is calculated to judge the temporal synchrony between the pixel where the site is located and the pixels within the neighborhood range. The closer the MIPS is to 1, the more consistent the vegetation growth of the pixel where the site is located and the pixels within the neighborhood range. This type of site is classified as a second-level site, while for sites with a lower MIPS value, they are classified as third-level sites. When the MVP within the scale of the satellite remote sensing phenology product of the site is not equal to 1, directly eliminate the sites with too small MVP. For some sites with MVP not equal to 1 but also having a relatively large value, spatial heterogeneity analysis is carried out. If the MRCS is small, it indicates that the surface spatial variation is small, and it can be considered to have spatial representativeness. This type of site is classified as a fourth-level site. For sites with MVP not equal to 1 and a large MRCS, the surface of these sites is heterogeneous and the spatial variation is large, and they do not have spatial representativeness, and they are classified as fifth-level sites. Then, several indicators such as RMSE, BIAS, Abs.BIAS, slope, and Pearson correlation coefficient are used to analyze the site classification results to evaluate the effect of site level classification. The calculation formulas are as follows:

[0055]

[0056] In the formula, L i is the pixel value of the phenology product where the site is located, G i is the ground observation value of the site, and n is the number of ground phenological observation sites.

[0057] The present invention relates to an evaluation method for the spatial representativeness of ground vegetation phenological observation stations for verifying remote sensing phenological products. Based on the principle of satellite remote sensing for monitoring vegetation phenology and the method of station spatial representativeness, MVP, MRCS, and MIPS ps evaluation indicators are proposed to analyze the spatial representativeness of ground phenological observation stations within the pixel scale of satellite remote sensing phenological products, classify the stations, select ground phenological observation stations with higher quality, lay a data foundation for the authenticity verification of satellite remote sensing phenological products, make the verification results more reliable, provide guarantee for the improvement of satellite remote sensing phenological products, thus promoting the production of land surface phenological products with higher accuracy, and have important significance for deeply understanding the land surface water and heat processes, carbon cycle processes under the background of global climate change and predicting the spatio-temporal changes of terrestrial ecosystems.

[0058] Compared with the prior art, the present invention has the following advantages:

[0059] Relying on the mechanism of satellite remote sensing for monitoring vegetation phenology, in view of the particularity of remote sensing extraction of vegetation phenology, the present invention proposes evaluation indicators and systems suitable for evaluating the spatial representativeness of ground phenological observation stations, fully considers ground homogeneity, spatial heterogeneity and pixel time synchronization, evaluates and classifies ground phenological observation stations, and lays a data foundation for the authenticity verification of satellite remote sensing phenological products.

[0060] The evaluation process of the present invention is scientific and reasonable, with high verification accuracy and reliability, simple calculation and low implementation difficulty. It classifies ground vegetation phenological observation stations, selects ground vegetation phenological observation stations with higher spatial representativeness, lays a data foundation for the authenticity verification of satellite remote sensing phenological products, makes the verification results more reliable, provides guarantee for the improvement of satellite remote sensing phenological products, thus promoting the production of land surface phenological products with higher accuracy, and has important significance for deeply understanding the land surface water and heat processes, carbon cycle processes under the background of global climate change and predicting the spatio-temporal changes of terrestrial ecosystems. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 Shows a technical flow chart of an evaluation method for the spatial representativeness of ground vegetation phenological observation stations for verifying remote sensing phenological products of the present invention;

[0062] Figure 2 Shows a spatial distribution map of global ground vegetation phenological observation stations evaluated in an embodiment of the present invention;

[0063] Figure 3 Shows the evaluation of the spatial representativeness verification results of ground phenological observation stations at different levels in an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0064] To make the objectives, technical processes, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. It should be noted that the following description is exemplary and is for the purpose of fully understanding the present invention, not for limiting the scope of application of the present invention. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below. The following refers to Figures 1 to 3 describe the present invention:

[0065] As Figure 1 shown, a method and process for evaluating the spatial representativeness of ground vegetation phenological observation sites for verifying remote sensing phenological products. Specific embodiments:

[0067] Step 1) Data acquisition: Obtain global ground phenological observation sites to be evaluated, including CERN, USA-NPN, PEP725, Canada plant watch (CNW), Harvard Forest Data (HF), CPON, Hubbard Brook Experimental Northern Research Station of the US Department of Agriculture Forest Service, UK_Nature_Calendar. The observation years and quantities of the sites are shown in Table 1, and the spatial distribution map of the sites is shown in Figure 2 . Download the remote sensing phenological product MCD12Q2 through Google Earth Engine (GEE), with a spatial resolution of 500 m. In this embodiment, the start of season (SOS) of vegetation growth in 2015 is used to evaluate the site classification results. Obtain the high-spatial-resolution land cover fine product GLC_FCS30_2015 and FROM_GLC_30m_2015. The GLC_FCS30_2015 data is produced by the research group of Liu Liangyun, and the overall accuracy of the product is 81.4%. The FROM_GLC_30m_2015 is produced by the Tsinghua University team, and the overall accuracy is 77.3%. Download the Landsat8 remote sensing image data in 2015 through GEE, with a spatial resolution of 30 m, and calculate the two-band enhanced vegetation index (EVI2, 2-band Enhanced Vegetation Index).

[0068] Table 1 Global ground vegetation phenological observation sites

[0069]

[0070] Step 2) Based on the existing high-spatial-resolution land cover fine product GLC_FCS30_2015 and FROM_GLC_30m_2015, calculate the proportion of the main vegetation cover area (MVP, Main Vegetation Proportion) within the scale of the remote sensing phenology product for the ground phenology observation sites. To reduce the impact of classification errors in the land cover fine product on the evaluation results, in this embodiment, sites that meet the MVP indicators of both land cover fine products are selected.

[0071] The calculation formula of MVP is as follows:

[0072]

[0073] MVP (Main Vegetation Proportion) is the proportion of the largest vegetation area within the site neighborhood range. A(V) is the area of each type of vegetation within the site neighborhood range, mainly including forest land, grassland, and farmland. n is the total number of land cover types within the pixel scale, and A(i) is the area of each type of land cover within the site neighborhood range.

[0074] Since even if the land cover is uniform, there may be significant spatial variation characteristics in its vegetation phenology. Based on the EVI2 data of Landsat8 remote sensing images, calculate the multi-date average relative sill coefficient (MRCS, Mean Relative Coefficient of Sill) index to measure the overall heterogeneity in the spatial dimension of the ground phenology observation sites. The calculation formula is as follows:

[0075]

[0076] In the formula, MRCS is the multi-date average relative sill coefficient (Mean Relative Coefficient of Sill), n is the total number of dates, sill d is the sill value of the d-th date, and z d (x) is the mean value within the site neighborhood range of the d-th date. The smaller the MRCS, the smaller the degree of spatial heterogeneity within the site neighborhood range, and vice versa.

[0077] To characterize whether the increase or decrease of the time series of the vegetation index is synchronous, and further evaluate the temporal synchrony between all pixels within the site neighborhood range and the pixel where the site is located, calculate the mean instantaneous phase synchrony (MIPS) index. The detailed calculation steps are as follows:

[0078] First, perform a Hilbert transform on the real time series signal to obtain the analytic information.

[0079] z a z̃(t) = z(t) + jΗ{z(t)}

[0080] where z(t) is the initial time series, Η represents the Hilbert transform performed on it, and z a z̃(t) is the sequence constructed based on the initial time series, and this equation can also be expressed as follows:

[0081] z a z̃(t) = a(t)e jφ(t)

[0082] a(t) is the instantaneous envelope, and φ(t) is the instantaneous phase.

[0083] Based on the above two formulas, the instantaneous phase at a specific moment can be calculated. The instantaneous phase difference between two time series at time t can be expressed as Δφ(t) = φ p (t) - φ s (t), and the synchrony is represented by the sine value of the phase difference.

[0084] The consistency between the time series curve of the pixel within the neighborhood range and the time series curve of the pixel where the site is located at time t can be expressed by the following formula:

[0085] IPS ps (t) = 1 - abs(sin(φ p (t) - φ s (t)))

[0086] where IPS ps (t) is the instantaneous phase synchrony value between the pixel where the site is located and a certain pixel within the neighborhood range at time t, φ p (t) is the instantaneous phase of a certain pixel within the neighborhood range at time t, and φ s (t) is the instantaneous phase of the pixel where the site is located at time t. When the phases of the two time series curves are consistent at a certain moment, IPS ps (t) = 1.

[0087] The consistency of the time series within the neighborhood range is evaluated using the average value of IPS for all pixels within the neighborhood range and all moments with the pixel where the site is located:

[0088]

[0089] where MIPS ps is the Mean Instantaneous Phase Synchrony (MIPS), m is the number of pixels within the neighborhood range around the site, and n is the number of valid dates, representing the time synchrony between all pixels within the neighborhood range of the site and the pixel where the site is located.

[0090] Step 3), verify the effect according to the natural breaks method and the above indicators, and use the hierarchical classification method to classify all ground phenological observation stations. The specific classification process is as follows: When the MVP within the scale of the satellite remote sensing phenological product of the station is equal to 1, it indicates that the land cover around the station is uniform. The MRCS indicator is used for evaluation. If the MRCS is small, it indicates that the ground space of the station is homogeneous, and the spatial representativeness of this type of station is good, and it is classified as a first-level station. For stations with a large MRCS, the MIPS indicator is calculated to judge the temporal synchronization between the pixel where the station is located and the pixels within the neighborhood range. The closer the MIPS is to 1, the more consistent the vegetation growth of the pixel where the station is located and the pixels within the neighborhood range. This type of station is classified as a second-level station, while for stations with a low MIPS value, they are classified as third-level stations. When the MVP within the scale of the satellite remote sensing phenological product of the station is not equal to 1, directly eliminate the stations with too small MVP. For some stations with MVP greater than or equal to 0.6, perform spatial heterogeneity analysis. If the MRCS is small, it indicates that the surface spatial variation is small and it can be considered to have spatial representativeness. This type of station is classified as a fourth-level station. For stations with MVP not equal to 1 and a large MRCS, the surface of these stations is heterogeneous and the spatial variation is large, and they do not have spatial representativeness, and they are classified as fifth-level stations.

[0091] Then, use several indicators such as RMSE, BIAS, Abs.BIAS, slope, and Pearson correlation coefficient to analyze the station classification results to evaluate the effect of station level classification. The calculation formulas are as follows:

[0092]

[0093]

[0094] In the formula, L i is the pixel value of the phenological product where the station is located, G i is the ground observation value of the station, and n is the number of ground phenological observation stations.

[0095] After the calculation and station level classification of the above steps, in this embodiment, the results are shown in Table 1 and Figure 3 as shown. From Table 2 and Figure 3 it can be seen that the higher the level (level 1) of the station, the smaller the root mean square error, bias, and absolute deviation from the satellite remote sensing phenological product, and they all have correlations.

[0096] Table 2 Station classification conditions and evaluation results under each level in the embodiment

[0097]

[0098] The results show that the spatial representativeness of the sites has a great impact on the validation of satellite remote sensing phenology products. When sites with better quality are selected to validate satellite remote sensing phenology products, the validation results are more reliable.

Claims

1. A method for evaluating the spatial representativeness of ground vegetation phenological observation sites, characterized in that, before verifying remote sensing phenological products, calculate the proportion of the main vegetation coverage area MVP, multi-date average relative sill coefficient MRCS, and average instantaneous phase synchronization MIPS within the coarse pixel scale based on high-resolution remote sensing image data to evaluate the spatial representativeness of ground vegetation phenological observation sites within the coarse pixel scale, classify all ground phenological observation sites, and select ground phenological observation sites with higher spatial representativeness; The specific calculation of the multi-date average relative sill coefficient MRCS includes: 1.1) Obtain all pixel pairs within the neighborhood range of the ground phenological observation site corresponding to the date from multi-temporal Landsat8 remote sensing image data; 1.2) By using the semivariogram model respectively to fit the variogram values of all pixel pairs, the key parameters reflecting the spatial heterogeneity of the surface are obtained: the sill value of the d-th date d ; 1.3) Calculate the multi-date average relative sill coefficient MRCS according to the sill value, and the specific formula is as follows: Wherein, MRCS is the average relative sill coefficient for multiple dates, n is the total number of dates, sill d is the sill value of the d-th date, is the mean value within the neighborhood of the site for the d-th date; The calculation of the average instantaneous phase synchronization MIPS specifically includes: 2.1) Perform Hilbert transform on the original time series information obtained based on high-resolution remote sensing images to obtain analytical information, and calculate the instantaneous phase of all times of the pixels within the neighborhood range of the ground phenological observation site and the pixel where the ground phenological observation site is located. The calculation formula is: z a (t) = z(t) + jΗ{z(t)} where \(z(t)\) is the initial time series, \(\mathcal{H}\) represents the Hilbert transform applied to it, and \(z a (t)\) is the sequence constructed based on the initial time series and is expressed as follows: z a (t) = a(t)e jφ(t) a(t) is the instantaneous envelope, and φ(t) is the instantaneous phase; 2.2) Calculate the instantaneous phase difference between the pixels within the neighborhood range of the ground phenological observation site and the pixel where the ground phenological observation site is located at all times. The specific formula is as follows: Δφ(t) = φ p (t) - φ s (t) 2.3) The synchrony is represented by the sine value of the instantaneous phase difference between the pixels within the neighborhood of the phenological observation site and the pixel where the ground phenological observation site is located, and the IPS ps (t) at time t between the time series curve of the pixels within the neighborhood and the time series curve of the pixel where the site is located is calculated. The calculation formula is as follows: IPS ps (t) = 1 - abs(sin(φ p (t) - φ s (t))) where, IPS ps (t) is the instantaneous phase synchronization value of the pixel where the site is located and a certain pixel within the neighborhood at time t, and φ p (t) is the instantaneous phase of a certain pixel within the neighborhood at time t, and φ s (t) is the instantaneous phase of the pixel where the site is located at time t. When the phases of the two time series curves are consistent at a certain moment, IPS ps (t) = 1; 2.4) Calculate the average value of the instantaneous phase synchronization of all pixels within the neighborhood range and the pixel where the site is located at all times to evaluate the consistency of the time series within the neighborhood range. The specific formula is as follows: where MIPS ps is the average instantaneous phase synchronization, m is the number of pixels within the neighborhood around the site, and n is the number of valid dates.

2. The method for evaluating the spatial representativeness of ground vegetation phenological observation sites according to claim 1, characterized in that, The calculation of the proportion of the main vegetation coverage area MVP is as follows: In the formula, MVP is the proportion of the largest vegetation area within the neighborhood range of the site, A(V) is the area of each type of vegetation within the neighborhood range of the site, n is the total number of land cover types within the neighborhood range, and A(i) is the area of each type of land cover type within the neighborhood range of the site.

Images