A method for retrieving daily canopy water content by coupling optical and microwave remote sensing

Abstract

This invention discloses a method for retrieving diurnal canopy water content by coupling optical and microwave remote sensing, belonging to the field of ecological and environmental remote sensing technology. The method includes: retrieving canopy water content data from optical images based on a canopy radiative transfer model; calculating vegetation optical thickness data based on a zero-order microwave radiative transfer model and a multi-channel collaborative inversion algorithm; acquiring vegetation optical thickness data, vegetation type, vegetation index, and meteorological data, and constructing a sample dataset after unifying the resolution; using the canopy water content data from optical images as ground truth labels, and combining the cumulative effect of rainfall and the lag effect of vegetation optical thickness, constructing an extreme gradient boosting machine learning estimation model; using this model to estimate the canopy water content in areas missing from the optical images, and filling in the missing areas to obtain a global diurnal, spatiotemporally continuous vegetation canopy water content dataset. This invention, employing the above method, achieves the complementary advantages of optical and microwave remote sensing, providing reliable technical support for monitoring vegetation moisture status.

Description

Technical Field

[0001] This invention relates to the field of ecological and environmental remote sensing technology, and in particular to a method for inverting daily canopy water content by coupling optical and microwave remote sensing. Background Technology

[0002] Canopy water content is a crucial physiological and ecological parameter reflecting vegetation water status, transpiration processes, water stress, and photosynthetic efficiency. Timely monitoring of vegetation water status is therefore essential. Canopy water content can be expressed as the product of the leaf thickness (equal amount of water) and the leaf area index, and is also equivalent to the difference between the fresh mass and dry mass of the leaves. Traditional destructive ground sampling methods (drying and weighing) are time-consuming, labor-intensive, and difficult to implement on a large scale. Remote sensing technology, on the other hand, is a rapid, non-destructive, and multi-scale method for detecting the biophysical characteristics of vegetation. Therefore, utilizing remote sensing technology to achieve rapid, non-destructive, and large-area monitoring of vegetation water content has become a research hotspot in the field of remote sensing applications.

[0003] Estimating vegetation canopy water content based on satellite remote sensing mainly includes two types of methods: one is to invert vegetation canopy water content using statistical models or radiative transfer models based on optical imagery; the other is to invert vegetation optical thickness (VOD), an indicator of vegetation water content, based on microwave remote sensing and radiative transfer processes. These methods have certain advantages and disadvantages. Generally, optical imagery can invert the absolute value of canopy water content, directly indicating the vegetation water content situation. However, optical imagery is often affected by weather conditions such as clouds and rain, and although it has high spatial resolution, it suffers from significant data gaps, making daily-scale monitoring difficult. Microwave remote sensing belongs to the field of active remote sensing and has the advantage of high daily-scale data coverage. Its disadvantage is that it currently cannot directly invert canopy water content, but can only invert the exponential VOD, which indicates vegetation water content.

[0004] Therefore, there is an urgent need to develop a method for inverting vegetation canopy water content by coupling optical and microwave remote sensing, so as to achieve accurate estimation of canopy water content and provide reliable technical support for monitoring drought stress and studying vegetation physiological status. Summary of the Invention

[0005] The purpose of this invention is to provide a method for inverting daily canopy water content by coupling optical and microwave remote sensing, overcoming the dual limitations of optical remote sensing being susceptible to cloud and rain interference leading to data loss, and microwave remote sensing having high coverage but being unable to directly invert the absolute value of canopy water content.

[0006] To achieve the above objectives, this invention provides a method for inverting diurnal canopy water content by coupling optical and microwave remote sensing, comprising the following steps: S1. Training samples are generated by forward simulation based on the radiative transfer model. Equivalent spectra are obtained by spectral convolution. The optimal feature combination and machine learning algorithm suitable for different vegetation types are determined through evaluation to obtain optical image canopy water content data. S2. Based on the zero-order microwave radiative transfer model, the single albedo of vegetation is optimized pixel by pixel, and the optical thickness data of vegetation is calculated using a multi-channel collaborative inversion algorithm. S3. Acquire and preprocess vegetation optical thickness data, vegetation type, vegetation index and meteorological data, and unify the resolution to 0.1°×0.1° to generate a raster dataset of input factors; S4. Based on the preprocessed vegetation optical thickness data, a sample dataset is formed by combining meteorological data, vegetation index, and vegetation type. Considering the cumulative effect of rainfall and the lag effect of vegetation optical thickness, an extreme gradient boosting machine learning estimation model is constructed. S5. For the raster dataset of input factors obtained in step S3, use the extreme gradient boosting machine learning estimation model in step S4 to estimate the daily vegetation canopy water content data. S6. Using the vegetation canopy water content data obtained in step S5, fill in the missing areas in the optical image canopy water content data obtained in step S1, thereby obtaining global daily-scale vegetation canopy water content data.

[0007] Preferably, step S1 specifically includes: S11. The radiative transfer model is used to simulate the canopy reflectivity in a forward manner. The PROSAIL model and PROGeoSail model are used for forest types, and the PROSAIL model is used for farmland types. S12. The simulated canopy reflectance of different vegetation types is converted into satellite equivalent spectra using the spectral response function, and Gaussian noise is added. The specific formula is as follows: = ; In the formula, It is the simulated equivalent reflectance. It is the original spectral reflectance. It is the number of spectral bands within each band. Indicates the first The original spectral reflectance of each spectral band. Indicates the first Spectral response function of each spectral band; S13. Conduct exhaustive trials for different combinations of feature variables to construct multiple candidate models, and use the coefficient of determination (R²) to determine the model. 2 The final optical image canopy water content data, taking into account vegetation type, are determined using the root mean square error (RMSE).

[0008] Preferably, the calculation formula for the zero-order microwave radiative transfer model in step S2 is as follows: ; in, Indicates the observed brightness temperature. Indicates the emissivity of rough soil. and These are vegetation emission and transmission terms, respectively, which characterize the emission and transmission behavior of microwave signals in vegetation.

[0009] Preferably, the meteorological data in step S3 includes rainfall and water vapor pressure deficit, and the vegetation index includes normalized vegetation index and enhanced vegetation index.

[0010] Preferably, step S3 specifically includes: S31. Calculate the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI). S32. Obtain the ERA5 reanalysis dataset, calculate the hourly vapor pressure deficit (VPD) based on the hourly 2-meter temperature data and dew point temperature data of the ERA5 reanalysis dataset, and synthesize the daily average value of the hourly vapor pressure deficit. S33. Obtain rainfall data from the MSWEP dataset, and calculate the cumulative rainfall over 1 to 30 days within the target period, with a daily time step. S34. Select vegetation function type data; S35. Unify the raster data to a resolution of 0.1°×0.1°. For categorical raster data, use the mode method to determine the category of the raster cell. For numerical data, perform grid mean aggregation with 0.1° as the target grid size.

[0011] Preferably, in step S35, the specific calculation method for grid mean aggregation is as follows: ; in, For a 0.1° grid cell value, This represents the number of raster cells. These are the original raster cell values. Indicates the first Each raster pixel Indicates the row number of the raster cell. This indicates the column number of the raster cell.

[0012] Preferably, step S4 specifically includes: S41. Take the canopy water content data of the optical image obtained in step S1 as the true value, and perform quality screening to retain the areas where the leaf area index is greater than 0 and the rainfall on that day is 0. S42. Based on the canopy water content data of the optical image in step S1, the vegetation optical thickness data in step S2 is extracted and combined with rainfall, water vapor pressure deficit, normalized vegetation index, enhanced vegetation index, and vegetation function type to form a sample dataset, taking into account the cumulative effect of rainfall and the lag effect of vegetation optical thickness. S43. Divide the sample dataset into a test set and a training set, and use spatial partitioning to avoid feature leakage. Use water vapor pressure deficit, normalized vegetation index, enhanced vegetation index, and vegetation function type as fixed features. Combine rainfall with vegetation optical thickness with different cumulative time lengths and different lag times in pairs. Use five-fold cross-validation to determine the optimal feature combination based on the coefficient of determination and root mean square error, and construct an extreme gradient boosting machine learning estimation model.

[0013] Preferably, the specific implementation method of step S6 is as follows: calculate the global daily-scale vegetation canopy water content data using an extreme gradient boosting machine learning estimation model, remove pixels with leaf area index less than 0, and merge them with the optical image canopy water content data of step S1 to obtain the filled global daily-scale vegetation canopy water content data.

[0014] Therefore, the present invention employs the above-mentioned method for inverting diurnal canopy water content using coupled optics and microwave remote sensing, which has the following beneficial effects: (1) Realize continuous monitoring at the daily scale: effectively fill the data gap caused by cloud and rain interference in optical images, and realize the inversion of vegetation canopy water content at the global daily scale and in time and space for the first time; (2) Integrating the advantages of multi-source information: Combining the high-precision canopy water content inversion of optical remote sensing with the all-weather coverage capability of microwave remote sensing, giving full play to the advantages of both; (3) Combining physical constraints with machine learning: Using the inversion results of the radiative transfer model as training labels, the machine learning model is given physical interpretability, which improves the estimation accuracy and robustness; (4) High precision and high coverage: Experimental results show that this method can fill in more than 50% of the missing optical data areas, and the model accuracy R 2 It can reach 0.63, with an RMSE of 0.06, demonstrating good estimation ability and spatial coverage.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method for retrieving daily canopy water content by coupling optics and microwave remote sensing according to the present invention. Figure 2This is an example of the daily-scale canopy water content inversion method of the present invention, which combines optical and microwave remote sensing. The accuracy map of the canopy water content model is constructed based on VOD, meteorological parameters and vegetation index. Detailed Implementation

[0017] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0018] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0019] Example 1 The present invention will be further illustrated below using optical images and microwave data acquired by the Fengyun (FY) 3D satellite. The implementation examples are based on the technical solutions of the present invention. It should be understood that these examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0020] like Figure 1 As shown, a method for inverting diurnal canopy water content by coupling optics and microwave remote sensing is described, with the following specific steps: S1. Training samples are generated based on forward simulation of the radiative transfer model. Equivalent spectra are obtained through spectral convolution. The optimal feature combination and machine learning algorithm suitable for different vegetation types are determined through evaluation to obtain optical image canopy water content data.

[0021] Based on the PROSAIL radiative transfer model, canopy reflectance spectra of various vegetation types under different biophysical conditions were simulated. The PROSAIL and PROGeoSail models were used for forest types, while the PROSAIL model was used for farmland types. The spectral response function was then used to convert the reflectance to an equivalent reflectance consistent with the FY-3D sensor, and Gaussian noise was added. The specific formula is as follows: = ; In the formula, It is the simulated equivalent reflectance. It is the original spectral reflectance. It is the number of spectral bands within each band. Indicates the first The original spectral reflectance of each spectral band. Indicates the first Spectral response function of each spectral band; Based on this, equivalent reflectance, optical vegetation parameters constructed from it, and corresponding canopy water content were used as input and output variables. Exhaustive experiments were conducted on different combinations of feature variables to construct multiple candidate models. The final optical image canopy water content inversion model considering vegetation type was determined by the coefficient of determination and root mean square error. Global canopy water content data at the astronomical scale for some months in 2023 were generated and quality-screened, retaining pixels with a leaf area index greater than 0 and a daily rainfall of 0. These pixels were then rasterized and used as ground truth labels for subsequent machine learning.

[0022] S2. Based on the zero-order microwave radiative transfer model, the single albedo of vegetation is optimized pixel by pixel, and the optical thickness data of vegetation is calculated using a multi-channel collaborative inversion algorithm.

[0023] Based on the zero-order microwave radiative transfer model, the single albedo of vegetation is optimized pixel by pixel, and a multi-channel collaborative inversion algorithm is used to calculate global VOD data. The calculation formula for the zero-order microwave radiative transfer model is: ; in, Indicates the observed brightness temperature. Indicates the emissivity of rough soil. and These are vegetation emission and transmission terms, representing the emission and transmission behavior of microwave signals in vegetation, respectively. First, brightness temperature observation data from six dual-polarization channels (10H, 10V, 18H, 18V, 23H, 23V) of the FY-3D satellite were preprocessed, and channel 10H was identified as the core channel for inversion. Second, position parameters were initialized. ,in, It is based on day data obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) model, using multi-year averages. The main method involves determining the brightness of different vegetation types using auxiliary files from the SMOS SM inversion algorithm; then, the brightness of the core channel is determined based on the relationship between the brightness temperatures of each channel. The simulated brightness temperature of the cooperative channel is calculated. Then, a cost function is established, and the parameters are updated by minimizing the difference between the observed brightness temperature and the simulated brightness temperature. This process is repeated iteratively until the cost function converges, yielding the final parameter estimates. Finally, based on the corrected model, the VOD value for each channel is calculated.

[0024] S3. Acquire and preprocess vegetation optical thickness data, vegetation type, vegetation index (normalized vegetation index and enhanced vegetation index) and meteorological data (rainfall and water vapor pressure deficit), and unify the resolution to 0.1°×0.1° to generate a raster dataset of input factors.

[0025] The Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) were calculated based on the MODIS surface reflectance dataset (MOD09GA), and quality screening was performed to remove invalid values ​​affected by aerosols, sensors, and other factors.

[0026] Obtain the ERA5 reanalysis dataset, calculate the hourly vapor pressure deficit (VPD) based on the hourly 2-meter temperature data and dew point temperature data of the dataset, and synthesize the daily average value of the hourly vapor pressure deficit.

[0027] Obtain rainfall data from the MSWEP dataset, and calculate the cumulative rainfall over 1 day, 2 days, ..., 30 days within the target time period, using daily as the time step.

[0028] Obtain the vegetation function type dataset (MCD12Q1) from MODIS, select the vegetation function type data, remove non-vegetated areas, i.e. retain type codes 1-8.

[0029] The raster data was standardized to a resolution of 0.1° × 0.1°, resulting in a 3600 × 1800 raster image set. For categorical raster data, the mode method was used to determine the category of each raster cell. For numerical data, grid mean aggregation was performed with a target grid size of 0.1°. The specific calculation method is as follows: ; in, For a 0.1° grid cell value, This represents the number of raster cells. These are the original raster cell values. Indicates the first Each raster pixel Indicates the row number of the raster cell. This indicates the column number of the raster cell.

[0030] S4. Based on the preprocessed vegetation optical thickness data, a sample dataset is formed by combining meteorological data, vegetation indices, and vegetation types. Considering the cumulative effect of rainfall and the lag effect of vegetation optical thickness, an extreme gradient boosting machine learning estimation model is constructed. For example... Figure 2 As shown, based on VOD, NDVI, EVI, PFT, VPD, and rainfall data, and considering the lag effect of VOD (3 to 18 days) and the cumulative effect of rainfall (2 to 30 days), an extreme gradient improvement machine learning estimation model with high accuracy R is established. 2The value is 0.63, and the RMSE is 0.06.

[0031] The canopy water content data obtained in step S1 is used as the true value and quality screening is performed to retain areas with a leaf area index greater than 0 and a rainfall of 0 on that day.

[0032] Based on the canopy water content data from optical imagery obtained in step S1, the vegetation optical thickness data from step S2 is extracted and combined with rainfall, vapor pressure deficit, normalized difference vegetation index (NDVI), enhanced vegetation index (EDI), and vegetation functional type to form a sample dataset. Rainfall is calculated on a daily time step, with cumulative rainfall over 2 days, ..., 30 days within the target time period to quantify the impact of cumulative rainfall effects at different time scales on canopy water content. The lag between VOD and leaf growth is also considered, with lags of 3 days, 6 days, ..., 18 days.

[0033] The sample dataset was divided into a test set and a training set, and a 100km test set was used. A 100km spatial partitioning method was used to avoid feature leakage and ensure a uniform distribution of vegetation types within the training and testing sets. Water vapor pressure deficit, normalized difference in vegetation index, enhanced vegetation index, and vegetation function type were used as fixed features. Rainfall with different cumulative durations and vegetation optical thickness with different lag durations were paired, and the optimal feature combination was determined based on the coefficient of determination and root mean square error using five-fold cross-validation. An extreme gradient boosting machine learning estimation model was then constructed.

[0034] S5. For the raster dataset of input factors obtained in step S3, use the extreme gradient boosting machine learning estimation model in step S4 to estimate the daily vegetation canopy water content data.

[0035] Using the extreme gradient boosting machine learning estimation model constructed in step S4, pixel-by-pixel estimation is performed on the input factor raster dataset generated in step S3 to obtain daily vegetation canopy water content data.

[0036] S6. Using the vegetation canopy water content data obtained in step S5, fill in the missing areas in the optical image canopy water content data obtained in step S1, thereby obtaining global daily-scale vegetation canopy water content data.

[0037] The global daily-scale vegetation canopy water content data is calculated using an extreme gradient boosting machine learning estimation model. Pixels with a leaf area index less than 0 are removed and merged with the optical image canopy water content data from step S1 to obtain the incomplete global daily-scale vegetation canopy water content data.

[0038] Statistical results show that the number of pixels estimating the vegetation canopy water content based on VOD is 293,000, and the number of missing pixels is 576,800, accounting for 50.80% of the missing area on that day.

[0039] Therefore, this invention employs a diurnal canopy water content inversion method that couples optical and microwave remote sensing, achieving diurnal and spatiotemporally continuous canopy water content inversion. This effectively solves the data loss problem caused by cloud and rain interference in optical imagery, while also compensating for the limitation of microwave remote sensing in directly obtaining the absolute value of canopy water content. This method fully leverages the complementary advantages of high optical inversion accuracy and good microwave coverage. Using the inversion results from the physical radiative transfer model as training labels, it endows the machine learning model with good physical interpretability and extrapolation capabilities. Furthermore, by combining rainfall accumulation and the lag characteristics of vegetation optical thickness, it enhances the model's ability to characterize dynamic changes in vegetation moisture, thereby obtaining canopy water content products with both high spatiotemporal resolution and high reliability.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for inverting diurnal canopy water content using coupled optical and microwave remote sensing, characterized in that, Includes the following steps: S1. Training samples are generated by forward simulation based on the radiative transfer model. Equivalent spectra are obtained by spectral convolution. The optimal feature combination and machine learning algorithm suitable for different vegetation types are determined through evaluation to obtain optical image canopy water content data. S2. Based on the zero-order microwave radiative transfer model, the single albedo of vegetation is optimized pixel by pixel, and the optical thickness data of vegetation is calculated using a multi-channel collaborative inversion algorithm. S3. Acquire and preprocess vegetation optical thickness data, vegetation type, vegetation index and meteorological data, and unify the resolution to 0.1°×0.1° to generate a raster dataset of input factors; S4. Based on the preprocessed vegetation optical thickness data, a sample dataset is formed by combining meteorological data, vegetation index, and vegetation type. Considering the cumulative effect of rainfall and the lag effect of vegetation optical thickness, an extreme gradient boosting machine learning estimation model is constructed. S5. For the raster dataset of input factors obtained in step S3, use the extreme gradient boosting machine learning estimation model in step S4 to estimate the daily vegetation canopy water content data. S6. Using the vegetation canopy water content data obtained in step S5, fill in the missing areas in the optical image canopy water content data obtained in step S1, thereby obtaining global daily-scale vegetation canopy water content data.

2. The method for inverting diurnal canopy water content by coupling optics and microwave remote sensing according to claim 1, characterized in that, Step S1 specifically includes: S11. The radiative transfer model is used to simulate the canopy reflectivity in a forward manner. The PROSAIL model and PROGeoSail model are used for forest types, and the PROSAIL model is used for farmland types. S12. The simulated canopy reflectance of different vegetation types is converted into satellite equivalent spectra using the spectral response function, and Gaussian noise is added. The specific formula is as follows: = ； In the formula, It is the simulated equivalent reflectance. It is the original spectral reflectance. It is the number of spectral bands within each band. Indicates the first The original spectral reflectance of each spectral band. Indicates the first Spectral response function of each spectral band; S13. Conduct exhaustive experiments on different combinations of characteristic variables, construct multiple candidate models, and determine the final optical image canopy water content data considering vegetation type by using the coefficient of determination and root mean square error.

3. The method for inverting diurnal canopy water content by coupling optics and microwave remote sensing according to claim 1, characterized in that, The calculation formula for the zero-order microwave radiative transfer model in step S2 is as follows: ； in, Indicates the observed brightness temperature. Indicates the emissivity of rough soil. and These are vegetation emission and transmission terms, respectively, which characterize the emission and transmission behavior of microwave signals in vegetation.

4. The method for inverting diurnal canopy water content by coupling optical and microwave remote sensing according to claim 1, characterized in that, The meteorological data in step S3 include rainfall and water vapor pressure deficit, and the vegetation indices include normalized vegetation index and enhanced vegetation index.

5. The method for inverting diurnal canopy water content by coupling optics and microwave remote sensing according to claim 4, characterized in that, Step S3 specifically includes: S31. Calculate the normalized vegetation index and the enhanced vegetation index; S32. Obtain the ERA5 reanalysis dataset, calculate the hourly water vapor pressure deficit based on the hourly 2-meter temperature data and dew point temperature data of the ERA5 reanalysis dataset, and synthesize the daily average value of the hourly water vapor pressure deficit. S33. Obtain rainfall data from the MSWEP dataset, and calculate the cumulative rainfall over 1 to 30 days within the target period, with a daily time step. S34. Select vegetation function type data; S35. Unify the raster data to a resolution of 0.1°×0.1°. For categorical raster data, use the mode method to determine the category of the raster cell. For numerical data, perform grid mean aggregation with 0.1° as the target grid size.

6. The method for inverting diurnal canopy water content by coupling optics and microwave remote sensing according to claim 5, characterized in that, In step S35, the specific calculation method for grid mean aggregation is as follows: ； in, For a 0.1° grid cell value, This represents the number of raster cells. These are the original raster cell values. Indicates the first Each raster pixel Indicates the row number of the raster cell. This indicates the column number of the raster cell.

7. The method for inverting diurnal canopy water content by coupling optical and microwave remote sensing according to claim 5, characterized in that, Step S4 specifically includes: S41. Take the canopy water content data of the optical image obtained in step S1 as the true value, and perform quality screening to retain the areas where the leaf area index is greater than 0 and the rainfall on that day is 0. S42. Based on the canopy water content data of the optical image in step S1, the vegetation optical thickness data in step S2 is extracted and combined with rainfall, water vapor pressure deficit, normalized vegetation index, enhanced vegetation index, and vegetation function type to form a sample dataset, taking into account the cumulative effect of rainfall and the lag effect of vegetation optical thickness. S43. Divide the sample dataset into a test set and a training set, and use spatial partitioning to avoid feature leakage. Use water vapor pressure deficit, normalized vegetation index, enhanced vegetation index, and vegetation function type as fixed features. Combine rainfall with vegetation optical thickness with different cumulative time lengths and different lag times in pairs. Use five-fold cross-validation to determine the optimal feature combination based on the coefficient of determination and root mean square error, and construct an extreme gradient boosting machine learning estimation model.

8. The method for inverting diurnal canopy water content by coupling optics and microwave remote sensing according to claim 1, characterized in that, The specific implementation method of step S6 is as follows: use the extreme gradient boosting machine learning estimation model to calculate the global daily-scale vegetation canopy water content data, remove pixels with leaf area index less than 0, and merge with the optical image canopy water content data of step S1 to obtain the filled global daily-scale vegetation canopy water content data.

Images