Method for automatically extracting lake ice phenological information based on ice water pixel proportion

Abstract

The invention discloses a lake ice phenological information automatic extraction method based on an ice water pixel proportion, which comprises the following steps: step 1, data preprocessing-image cutting: four images of two land surface temperature wavebands of a lake in daytime and nighttime are acquired from two data sets, step 2, QC waveband quality screening: quality evaluation is carried out on the images at the QC waveband, and the QC waveband is selected as the QC waveband; dividing the pixels into qualified pixels and unqualified pixels; step 3, data fusion: assigning an unqualified pixel as a null value, and taking the maximum value of the same pixel of the four images on the same day as a fusion numerical value; according to the method, the extraction and operation processes are realized through GEE, the degree of dependence of the extraction process on artificial experience is reduced, and the working efficiency is improved; the method disclosed by the invention can be used for accurately monitoring the whole freezing and melting process of the lake ice in large-scale and long-time series lake ice dynamic remote sensing monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing applications, and specifically to an automatic extraction method for lake ice phenological information based on the proportion of ice-water pixels. Background Technique

[0002] Lake ice phenology is a sensitive indicator of regional climate change and plays a crucial role in aspects such as lake environmental management, lake energy exchange, and human winter activities. Monitoring and studying lake ice phenology helps to study the significant impacts of climate system changes on natural ecosystems, is the theoretical basis for analyzing and predicting the impacts of climate change on the human living environment, and provides technical support for formulating response policies to global warming. Lake ice phenology refers to the complex phenomenon of the freezing, covering, and melting processes of lake ice as climate changes, including the start of ice formation, complete ice formation, start of ice melting, complete ice melting, complete freezing time, and lake ice duration. Therefore, monitoring the heat changes and phenological characteristics of lake ice during the ice-covered period is crucial for studying the feedback mechanism of lake ice in the climate system and the scientific management of the lake ecological environment during the ice-covered period.

[0003] However, the traditional extraction of lake ice phenological information has the following disadvantages:

[0004] The limitations of the traditional field environment and monitoring conditions for extracting lake ice phenological information pose significant challenges for long-term continuous observations. Summary of the Invention

[0005] The purpose of the present invention is to provide an automatic extraction method for lake ice phenological information based on the proportion of ice-water pixels to solve the problem that the traditional extraction of lake ice phenological information faces significant challenges for long-term continuous observations due to the limitations of the field environment and monitoring conditions in the above-mentioned background technique.

[0006] To achieve the above purpose, the present invention provides the following technical solution: An automatic extraction method for lake ice phenological information based on the proportion of ice-water pixels, including the following steps:

[0007] Step 1. Data preprocessing - Image cropping: Obtain four images of the two surface temperature bands of the lake during the day and at night from two datasets, and preprocess the images;

[0008] Step 2. QC band quality screening: The QC band evaluates the quality of the images and classifies the pixels into qualified pixels and unqualified pixels;

[0009] Step 3. Data fusion: Assign null values to the unqualified pixels, and take the maximum value of the same pixel in the four images of the same day as the fusion value;

[0010] Step 4, Pixel category determination: The LST of the current day is converted into LWST by means of band calculation, and the pixels are divided into water pixels and ice pixels according to the pixel temperature threshold.

[0011] Step 5, Time series generation - Extracting phenological information by threshold method: Calculate and statistically count the number of daily ice pixels and water pixels within the vector range of the lake respectively, and generate a time series of the daily lake coverage area ratio.

[0012] Step 6, Accuracy verification: Use the lake ice phenology provided by the measured data or publicly available remote sensing data as the true value, and the phenological value extracted by this algorithm as the test value, and calculate the determination coefficient R 2 , RMSE, MAE and p - value to evaluate the extraction accuracy.

[0013] As a preferred technical solution of the present invention, the image in Step 1 is specifically an image generated from the daytime and nighttime surface temperature bands of MOD11A1 and MYD11A1 in the MODIS LST product called on the GEE platform, with a spatial resolution of 1 km and a temporal resolution of daily. The lake vector boundary is derived from the Hydro LAKES global lake dataset.

[0014] As a preferred technical solution of the present invention, the QC band in Step 2 is a binary number on the image, and each bit represents different image quality attributes, which is used for evaluating the quality of remote sensing images.

[0015] As a preferred technical solution of the present invention, the data fusion in Step 3 is to fuse the pixel temperature values of the qualified images after screening the QC band and take the maximum temperature value of the same pixel as the LST of the current day to ensure the accuracy of the temperature value of that day.

[0016] As a preferred technical solution of the present invention, the mid - segment calculation in Step 4 is to convert the unit of the pixel temperature value from Fahrenheit K to Celsius °C.

[0017] As a preferred technical solution of the present invention, the band calculation formula in Step 4 is:

[0018] LWST = LST×0.02 - 273.15,

[0019] In the formula: LWST is the lake surface temperature, with the unit of °C, and LST is the surface temperature value of MODIS, with the unit of K;

[0020] The pixel temperature threshold formula is:

[0021]

[0022] As a preferred technical solution of the present invention, the definition of the lake ice coverage area percentage in Step 5 is:

[0023]

[0024] where N ice and N total correspond to the number of ice pixels and the total number of lake pixels respectively. The four parameters for extracting lake ice phenology based on the threshold are: the start of lake freezing FUS, full ice formation FUE, start of ice melting BUS, and full ablation BUE. The difference between FUS and BUE is the ice cover duration ICD, and the difference between FUE and BUS is the full ice freezing time CFD.

[0025] The discriminant threshold formula for phenological parameters is:

[0026]

[0027] As a preferred technical solution of the present invention, the absolute coefficient R 2 is calculated by the formula:

[0028]

[0029] In the formula: y i is the true value, is the verification value, is the average value of the true values, and n is the number of samples; the calculation formula of RMSE is:

[0030]

[0031] In the formula: y i is the verification value, is the true value, and n is the number of samples; the calculation formula of MAE is:

[0032]

[0033] In the formula: y i is the verification value, is the true value, and n is the number of samples.

[0034] Compared with the prior art, the beneficial effects of the present invention are:

[0035] 1. The extraction and operation process of the present invention is realized by GEE, reducing the dependence on manual experience in the extraction process and improving work efficiency;

[0036] 2. The method of the present invention can accurately monitor the whole process of lake ice freezing and melting in large-scale and long-term lake ice dynamic remote sensing monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is the flowchart of the present invention;

[0038] Figure 2 Flow chart of the present invention

[0039] Figure 3 Spatial distribution map of the freezing process of Qinghai Lake according to the present invention

[0040] Figure 4 Map for extracting phenological parameters of lake ice in Qinghai Lake according to the present invention

[0041] Figure 5 Map for verifying the accuracy of the extraction results of lake ice phenology in Qinghai Lake according to the present invention Detailed implementation manners

[0042] Next, in combination with the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0043] Please refer to Figures 1-5 , the present invention provides an automatic extraction method for lake ice phenological information based on the proportion of ice-water pixels, including the following steps:

[0044] Step 1. Data preprocessing - image cropping: Obtain four images of two surface temperature bands, day and night, of the lake from two datasets, and preprocess the images;

[0045] Step 2. QC band quality screening: Use the QC band to evaluate the quality of the images and classify the pixels into qualified pixels and unqualified pixels;

[0046] Step 3. Data fusion: Assign null values to the unqualified pixels, and take the maximum value of the same pixel in the four images of the same day as the fusion value;

[0047] Step 4. Pixel category determination: Convert the LST of the same day into LWST by means of band calculation, and divide the pixels into water pixels and ice pixels according to the pixel temperature threshold;

[0048] Step 5. Time series generation - threshold method to extract phenological information: Calculate and statistically count the number of daily ice pixels and water pixels within the vector range of the lake respectively, and generate a time series of the daily lake coverage area ratio;

[0049] Step 6. Accuracy verification: Use the lake ice phenology provided by the measured data or publicly available remote sensing data as the true value, and the phenological value extracted by this algorithm as the test value, and calculate the determination coefficient R 2 , RMSE, MAE and p-value to evaluate the extraction accuracy.

[0050] In Step 1, the image is specifically an image generated from the daytime and nighttime land surface temperature bands of MOD11A1 and MYD11A1 in the MODIS LST product on the GEE platform, with a spatial resolution of 1 km and a temporal resolution of daily. The lake vector boundary is sourced from the Hydro LAKES global lake dataset.

[0051] In Step 2, the QC band is a binary number on the image, with each bit representing different image quality attributes and being used for evaluating the quality of remote sensing images.

[0052] In Step 3, data fusion is to fuse the temperature values of the qualified image pixels after screening the QC band and take the maximum temperature value of the same pixel as the LST for the day to ensure the accuracy of the temperature value for that day.

[0053] In Step 4, the mid - section calculation is to convert the unit of the pixel temperature value from Fahrenheit K to Celsius °C.

[0054] The band calculation formula in Step 4:

[0055] LWST = LST x 0.02 - 273.15,

[0056] In the formula: LWST is the lake surface temperature, with the unit of °C, and LST is the land surface temperature value of MODIS, with the unit of K;

[0057] The pixel temperature threshold formula is:

[0058]

[0059] In Step 5, the percentage of lake ice coverage area is defined as:

[0060]

[0061] Where N ice and N total correspond to the number of ice pixels and the total number of lake pixels respectively. The four parameters for extracting lake ice phenology according to the threshold are: the start of lake freezing FUS, full ice formation FUE, start of ice melting BUS, and full ablation BUE. The difference between FUS and BUE is the lake ice duration ICD, and the difference between FUE and BUS is the lake ice full - freezing time CFD.

[0062] The phenological parameter discrimination threshold formula is:

[0063]

[0064] In Step 6, the calculation formula for the absolute coefficient R 2 is:

[0065]

[0066] where: y i is the true value, is the verification value, is the average value of the true values, and n is the number of samples; The calculation formula of RMSE is:

[0067]

[0068] where: y i is the verification value, is the true value, and n is the number of samples; The calculation formula of MAE is:

[0069]

[0070] where: y i is the verification value, is the true value, and n is the number of samples.

[0071] Example 1:

[0072] In the present invention, in the GEE platform, the Day and Night surface temperature bands of MOD11A1 and MYD11A1 in the MODIS LST product are called. The preprocessing of the image includes image cropping, radiometric correction, atmospheric correction, geometric correction, and georegistration. The MODIS LST product is the Day and Night bands in MOD11A1 and MYD11A1, with a spatial resolution of 1 km and a temporal resolution of daily. The lake vector is sourced from the Hydro LAKES global lake dataset. Lake ice phenology provides the monitoring time range for lake management and water quality monitoring, serves the dynamic monitoring of the risk assessment of algal growth and the pollutant diffusion process; lake ice phenology provides the time range for ice activities and winter fishing, ensuring travel safety and the quality of passenger experience; provides data support for winter fishery under-ice seedling raising and spring fishery resource restoration; The QC band in MODIS is a band used to evaluate the quality of the image. The QC band is a binary number on the image, and each bit represents different image quality attributes, which is used to evaluate remote sensing images. It has a total of four identifiers, namely 1&0, 3&2, 5&4, and 7&6. The determination conditions for each identifier value are shown in the following table. Select the 1&0, 3&2, and 7&6 identifiers to evaluate the pixels. If the identifier value is 0, it is marked as a qualified pixel, otherwise it is unqualified and excluded. The four different band images obtained after screening are MOD_Day_qualified, MYD_Day_qualified, MOD_Night_qualified, and MOD_Night_qualified;

[0073]

[0074]

[0075] After the pixels pass the quality screening, they are divided into two categories: qualified pixels and unqualified pixels. The unqualified pixels are assigned null values and do not participate in subsequent calculations. For the pixels with qualified quality, the maximum value of the same pixel in the four images on the same day is taken as the fusion value. Data fusion is to fuse the pixel temperature values of the qualified images after screening the QC band and take the maximum temperature value of the same pixel as the LST of the day to ensure the accuracy of the temperature value of that day. The LST of the day is converted into LWST by means of band calculation. According to the pixel temperature threshold, the pixels are divided into water pixels and ice pixels. The band calculation formula is:

[0076] LWST = LST × 0.02 - 273.15,

[0077] where: LWST is the lake surface temperature, with the unit of °C, and LST is the land surface temperature value of MODIS, with the unit of K;

[0078] The pixel temperature threshold formula is:

[0079]

[0080] Calculate and count the number of ice pixels and water pixels in the daily lake vector range respectively, and generate a time series of the daily lake coverage area ratio. Among them, the lake ice coverage percentage ICF is defined as

[0081] where N ice and N total correspond to the number of ice pixels and the total number of lake pixels respectively. The four parameters of lake ice phenology extracted according to the threshold are: the start of lake freezing FUS, complete ice formation FUE, start of ice melting BUS, and complete ablation BUE. Among them, the difference between FUS and BUE is the lake ice duration ICD, and the difference between FUE and BUS is the lake ice complete freezing time CFD.

[0082] The phenological parameter discrimination threshold formula is:

[0083]

[0084] Using the lake ice phenology provided by measured data or publicly available remote sensing data as the true value and the phenological value extracted by this algorithm as the verification value, calculate the coefficient of determination R 2 , RMSE, MAE and p-value to evaluate the extraction accuracy. The calculation formula of the absolute coefficient R 2 is:

[0085]

[0086] where: y i is the true value, is the verification value, is the true value average, and n is the number of samples; the calculation formula of RMSE is:

[0087]

[0088] In the formula: y i is the verification value, is the true value, and n is the number of samples; the calculation formula of MAE is:

[0089]

[0090] In the formula: y i is the verification value, is the true value, and n is the number of samples.

[0091] Example 2:

[0092] In the present invention, the difference between this embodiment and the specific Example 1 is that the MODIS LST products are the Day and Night bands in MOD11A1 and MYD11A1, with a spatial resolution of 1 km and a temporal resolution of daily. The lake vector is derived from the Hydro LAKES global lake dataset. The lake ice phenology provides a monitoring time range for lake management and water quality monitoring, serves for the assessment of the risk of algal growth and the dynamic monitoring of the pollutant diffusion process; the lake ice phenology provides a time range for ice activities and winter fishing, ensuring travel safety and the quality of the passenger experience; and provides data support for winter fishery under-ice seedling raising and spring fishery resource restoration.

[0093] Example 3:

[0094] In the present invention, the difference between this embodiment and the specific Example 1 or Example 2 is that the QC band is a binary number on the image, and each bit represents different image quality attributes, which is used to evaluate remote sensing images.

[0095] Example 4:

[0096] In the present invention, the difference between this embodiment and the specific Example 1 is that data fusion is to fuse the qualified image pixel temperature values after screening the QC band and take the maximum temperature value of the same pixel as the LST of the day to ensure the accuracy of the temperature value of that day.

[0097] Example 5:

[0098] In the present invention, the difference between this embodiment and the specific Example 1 is that the band calculation is a common calculation method for converting the unit of the pixel temperature value from Fahrenheit K to Celsius °C.

[0099] Example 6:

[0100] In the present invention, in this embodiment, the MODIS remote sensing images of Qinghai Lake from 2003 to 2004 are used as example data provided by the Google Earth Engine cloud computing platform. The spatial resolution of the images is 1 km, the temporal resolution is daily, and the vector of the Qinghai Lake area is provided by the Hydro LAKES global lake dataset;

[0101] Combined with Figure 2 The specific operation steps of this embodiment are as follows:

[0102] First step: Upload the lake vector of the Qinghai Lake study area, use the GEE platform to call the Day and Night bands of the MODIS LST products MOD11A1 and MYD11A1 and perform preprocessing work, and use the uploaded vector to clip the images to obtain four images within the scope of Qinghai Lake, namely MOD_Day, MYD_Day, MOD_Night, and MYD_Night;

[0103] Second step: Use the QC band to perform quality screening on the extracted images. There are four identifiers in the QC band, namely 1&0, 3&2, 5&4, and 7&6. The determination conditions for each identifier value are as follows. Select the 1&0, 3&2, and 7&6 identifiers to evaluate the pixels. If the identifier value is 0, it is marked as a qualified pixel, otherwise it is unqualified and excluded. The four different band images after screening are MOD_Day_qualified, MYD_Day_qualified, MOD_Night_qualified, and MOD_Night_qualified;

[0104]

[0105] The pixel temperature threshold divides the pixels into water pixels and ice pixels. Combined with the attached Figure 3 of the specification, the band calculation formula:

[0106] LWST = LST x 0.02 - 273.15,

[0107] Where: LWST is the lake surface temperature, unit is °C, and LST is the MODIS land surface temperature value, unit is K;

[0108] The pixel temperature threshold formula is:

[0109]

[0110] Calculate and statistically count the daily number of ice pixels and water pixels within the lake vector range respectively, and generate a time series of the daily lake coverage area ratio. Among them, the lake ice coverage percentage ICF is defined as

[0111] where N ice and N total correspond to the number of ice pixels and the total number of lake pixels respectively. The four parameters for extracting lake ice phenology according to the threshold are: the start of lake freezing FUS, complete ice formation FUE, start of ice melting BUS, and complete ablation BUE. The difference between FUS and BUE is the ice cover duration ICD, and the difference between FUE and BUS is the complete ice freezing time CFD. The ecological disappearance state of the lake is determined based on the daily change of ICF. The area thresholds are 90% and 10%, that is, when the lake ice coverage exceeds 90%, the lake surface is completely covered by ice; when the lake ice coverage is less than 10%, the lake surface is open water.

[0112] The discriminant threshold formula for phenology parameters is:

[0113]

[0114] Among them, it includes the start date of lake freezing FUS, the complete ice formation date FUE, the start date of ice melting BUS, and the complete ablation date BUE. Then, calculate the difference between FUS and BUE as the ice cover duration ICD, and the difference between FUE and BUS as the complete ice freezing time CFD, determine 4 lake ice freezing / ice formation dates and 2 lake ice periods, and then guide the time window for human ice activities and lake monitoring;

[0115] Using the lake ice phenology parameter results of the published Qinghai-Tibet Plateau Lake Ice Phenology Dataset 1978 - 2016 for comparative analysis, the accuracy verification graph is as Figures 4-5 shown in Table 1:

[0116]

[0117] Using the lake ice phenology provided by the measured data or publicly available remote sensing data as the true value, and the phenology value extracted by this algorithm as the test value, calculate the determination coefficient R 2 , RMSE, MAE, and p-value to evaluate the extraction accuracy.

[0118] Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for automatically extracting lake ice phenological information based on ice-water pixel ratio, characterized in that: The following steps are involved: Step 1: Data preprocessing - image cropping: Obtain four images of the two surface temperature bands of the lake during the day and at night from the two datasets and preprocess the images; Step 2: QC band quality screening: The QC band evaluates the quality of the image and divides the pixels into qualified pixels and unqualified pixels; Step 3: Data fusion: assign null values ​​to unqualified pixels, and take the maximum value of the same pixel in the four images of the day as the fusion value; Step 4: Pixel category determination: The LST of the day is converted into LWST by band calculation, and the pixels are divided into water pixels and ice pixels according to the pixel temperature threshold; Step 5: Time series generation - extraction of phenological information using the threshold method: Calculate the number of ice pixels and water pixels in the lake vector range every day to generate a time series of the daily lake coverage ratio; Step 6: Accuracy verification: Use the lake ice phenology data provided by measured data or public remote sensing data as the true value and the phenology value extracted by this algorithm as the test value to calculate the determination coefficient R 2 , RMSE, MAE and p-value were used to evaluate the extraction accuracy.

2. The method for automatically extracting lake ice phenological information based on ice-water pixel ratio according to claim 1, characterized in that: The image in step 1 is specifically an image generated by calling the two surface temperature bands of MOD11A1 and MYD11A1 in the daytime and nighttime of the MODIS LST product in the GEE platform, with a spatial resolution of 1 km and a temporal resolution of daily. The lake vector boundary is derived from the Hydro LAKES global lake dataset.

3. The method for automatically extracting lake ice phenological information based on ice-water pixel ratio according to claim 1, characterized in that: The QC band in step 2 is a binary number on the image, each bit represents a different image quality attribute, and is used to evaluate the quality of remote sensing images.

4. The method for automatically extracting lake ice phenological information based on ice-water pixel ratio according to claim 1, characterized in that: The data fusion in step 3 is to fuse the temperature values ​​of the image pixels that are qualified after the QC band screening and take the maximum temperature value of the same pixel as the LST of the day to ensure the accuracy of the temperature value of the day.

5. The method for automatically extracting lake ice phenological information based on ice-water pixel ratio according to claim 1, characterized in that: The calculation in step 4 is to convert the pixel temperature value unit Fahrenheit (K) into Celsius (℃).

6. The method for automatically extracting lake ice phenological information based on ice-water pixel ratio according to claim 1, characterized in that: The band calculation formula in step 4 is: LWST=LSTx0.02-273.15, Where: LWST is the lake surface temperature, in °C, LST is the land surface temperature value from MODIS, in K; The pixel temperature threshold formula is:

7. The method for automatically extracting lake ice phenological information based on ice-water pixel ratio according to claim 1, characterized in that: The percentage of lake ice coverage in step 5 is defined as: Where N ice and N total Corresponding to the number of ice pixels and the total number of lake pixels, the four parameters of lake ice phenology extracted according to the threshold are: the beginning of lake freezing FUS, complete freezing FUE, the beginning of ice melting BUS and complete melting BUE. The difference between FUS and BUE is the lake ice duration ICD, and the difference between FUE and BUS is the lake ice complete freezing time CFD. The formula for judging the threshold value of phenological parameters is:

8. The method for automatically extracting lake ice phenological information based on ice-water pixel ratio according to claim 1, characterized in that: The absolute coefficient R in step 6 2 The calculation formula is: Where: y i is the true value, is the verification value, is the true value average, n is the number of samples; the calculation formula of RMSE is: Where: y i is the verification value, is the true value, n is the number of samples; The calculation formula of MAE is: Where: y i is the verification value, is the true value, and n is the number of samples.

Images