Evaluating wetland hydrological performance under three different conservation programs in Nebraska, United States, during 2018–2021

Assessing hydrological dynamics of wetlands is essential for understanding ecological services. This study utilized open‐access Sentinel‐2 satellite data to enhance conservation management by enabling near‐real‐time monitoring and assessment of hydrological dynamics in conserved lands across Nebraska, United States. Using machine learning and Google Earth Engine, this research classifies surface water cover rate for different conserved land sites in Nebraska in 2018–2021. The results of the study confirmed successful inundation performance in conserved wetland sites under Wildlife Management Areas (WMA), Wetlands Reserve Program (WRP), and Waterfowl Production Areas (WPA). The WMA sites had the highest inundated area rate of 16.41%, indicating active hydrological inundation of the core conserved land areas. The WRP and WPA sites reached a mean annual surface water cover rate of 8.07% and 7.51%, respectively, demonstrating occasional flooding or periodic inundation of core wetland areas but limited inundation coverages of the surrounding areas. The findings confirmed that wetland conservation practices are functioning very well on the sites with higher inundation rates, but hydrological restoration at the watershed scale could boost conservation performance for the entire conserved land areas. The findings of this research provide robust evidence for obtaining surface water inundation data, which is crucial for sustainable conservation assessment and achieving long‐term goals in wetland monitoring, protection, and management.

The European Space Agency (ESA) launched the Sentinel-2A (S2A) and Sentinel-2B (S2B) twin sensors in 2015and 2017, respectively (European Space Agency, 2019).Sentinel satellites offer high-resolution and multispectral imagery, enabling detailed monitoring of land, water bodies, and vegetation on a global scale.Sentinel-2 capture data useful for water resource management and frequent revisit times of Sentinel satellites allow for near real-time monitoring, making them valuable for tracking changes in water resources over time (Phiri et al., 2020).These sensors provide 10 bands covering the visible, near-infrared (NIR), red-edge, and shortwave infrared (SWIR) spectra, with spatial resolutions of 10 and 20 m.Additionally, there are three additional bands at a spatial resolution of 60 m specifically used for atmospheric correction.
These freely accessible sensors have enabled regional monitoring worldwide, facilitating various applications (European Space Agency, 2019).
This high-quality multispectral imagery from Sentinel-2 provides the necessary resolution for wetland monitoring and assessment (Ludwig et al., 2019).Due to the varied spatial and temporal hydrological characteristics of wetlands, creating effective remote sensing algorithms and indices to detect water under wetland vegetation and wet soils has been difficult (Ozesmi & Bauer, 2002;Pena-Regueiro et al., 2020).
The extent and spectral signature of wetlands can be highly variable due to the daily and seasonal changes in the water within them.Shallow water areas that are mixed with vegetation or soil can make it difficult to detect spectral reflectance.Mono-temporal classification is therefore inadequate to capture the temporal dynamics of wetlands (Ludwig et al., 2019).NIR and SWIR spectroscopy can provide useful information about water and wet surfaces, leading to the development of water indices that can be applied to satellite images to map wetland hydrological dynamics.Popular indices for this purpose include the normalized difference water index (NDWI), modified normalized difference water index (MNDWI), normalized difference moisture index (NDMI), and normalized difference vegetation index (NDVI) (Bwangoy et al., 2010;Davranche et al., 2013;Islam et al., 2008;Kulawardhana et al., 2007;Ma et al., 2017;Tang et al., 2016).
In the last two decades, machine learning and artificial intelligence algorithms have received attention in wetland monitoring and management by imaging classification (Abdel-Hamid et al., 2018;Franklin et al., 2018;Maxwell et al., 2018).The researcher created an unsupervised method to estimate the degree of flooding using radiometrically adjusted Sentinel-2 data.They utilized water in wetlands (WIW) to monitor changes in the spatial and temporal patterns of flooding in wetlands with diverse heights and densities over time from images captured by Sentinel/Landsat (Kordelas et al., 2018;Lefebvre et al., 2019).In their study, Pena-Regueiro et al. (2020) utilized seven indices to examine Sentinel-2 images and identify small water bodies in wetlands that exhibited varying patterns of temporal and spatial flooding.Upon comparison of the indices, it was determined that the NDWI index yielded the most accurate results in extracting water surfaces.Huang et al. (2018) created an automatic classification tree method to classify surface water extent in a Prairie Pothole Region site using Sentinel-1 synthetic aperture radar (SAR) data.Their approach resulted in an overall accuracy (OA) ranging from 79% to 93%.Additionally, the study explored support vector machine (SVM) models as a surface water classification method, which is a widely used machine learning algorithm in the fields of remote sensing, land-cover classification, and mapping.SVM is often applied to address both classification and regression issues in remote sensing and has been commonly used for land-use and land-cover (LULC) classification tasks.Shao and Lunetta (2012) gathered MODIS timeseries data and utilized SVM to perform LULC classification in North Carolina.The findings demonstrated that SVM exhibited exceptional generalization capability, even when using a small number of training samples.In recent years, machine learning algorithms have demonstrated a considerable benefit in LULC classification, making them a key component in the remote sensing research, specifically for LULC classification (Whyte et al., 2018;Xu et al., 2017).SVM classifiers show the highest OA and require the least amount of training samples compared to RF classifiers.This difference in training sample size influences the performance of both SVM and RF classifiers (Thanh Noi & Kappas, 2017).Ma et al. (2017) examined the data and compared the SVM and RF classifiers, concluding that SVM was more effective with small training set sizes.Sheykhmousa et al. (2020) conducted a review of 251 journal papers to compare the results of RF and SVM models.Their analysis revealed that SVM had a generally higher OA than RF when used at resolutions higher than 10 m, while the reverse was true for resolutions lower than 100 m.In this study, Sentinel-2 had a spatial resolution of 10 m, which was consistent with most prior research.In Zhang et al. (2022), six main machine learning algorithms were compared, and it was confirmed that SVM was the optimal algorithm when using GEE to classify wetland land cover in Nebraska.It was found that SVM with a linear kernel presented the best performance in surface water classification.Consequently, this study has applied linear kernel SVM as a surface water classifier in Nebraska wetlands.Two common methods of remote sensing classification are pixel-based and objective-based (Dronova, 2015;Whyte et al., 2018;Xu et al., 2017).This study employed a pixel-based classification model, which has been successfully utilized for water body and wetland classification.The study provides a long-term and costeffective wetland mapping tool that can be used in hydrological performance monitoring for wetland conservation programs (Jia et al., 2018;Lang & McCarty, 2009;Whyte et al., 2018).
GEE provides a cost-free and open-source platform that allows users to access, process, and analyze Earth observation data, including Sentinel-2 imagery.GEE is a cloud-based platform with a vast repository of satellite imagery and geospatial data.It enables efficient analysis and processing of massive datasets, with tools for detecting changes in water bodies, assessing water quality, and monitoring land use changes affecting water resources (Tamiminia et al., 2020).In recent studies, GEE has been utilized to map and assess the conditions of wetlands (Mahdianpari et al., 2020;Tamiminia et al., 2020).GEE offers a wide selection of remote sensing information and data processing techniques, featuring various embedded machine learning algorithms (Farda, 2017;Gorelick et al., 2017).GEE was utilized to provide different machine learning classifiers for pixel-based classification and to import Sentinel-2 satellite images and machine learning models.By using GEE, it was possible to map all conservation sites across Nebraska and examine surface water changes.Sentinel data and GEE offer significant potential in water resource monitoring and management.Sentinel images allow for monitoring water quality parameters like turbidity, chlorophyll-a concentration, and water temperature.Additionally, Sentinel-1's Synthetic Aperture Radar enables all-weather, day-night monitoring, aiding flood mapping, and early warning systems.Through analysis of Sentinel-2 imagery and historical data, GEE facilitates drought monitoring and water availability assessment.Satellite data assist in estimating crop water requirements and evaluating water use efficiency for irrigation management.Furthermore, Sentinel images and GEE aid in assessing land use changes within watersheds, contributing to effective watershed management and long-term groundwater monitoring (Amani et al., 2020;Druce et al., 2021).Moreover, it was an efficient tool for mapping all conserved land sites across Nebraska and detecting changes in surface water.
Remote sensing techniques for tracking hydrological dynamics in conserved land, especially wetlands commonly found in landlocked regions like Nebraska, have been relatively scarce and limited in number.This research aims to develop a cost-effective, sustainable, and efficient method for monitoring surface water inundation in conserved land sites throughout Nebraska.Additionally, machine learning models, GEE platform, and GIS tools will be utilized to detect, map, and analyze surface water changes in these sites.This research answers two specific questions: (1) How well did the Nebraska's wetlands in major conservation programs, including WMA, WPA, and WRP, perform with hydrological inundation coverage during 2018-2021?and (2) What are the policy implications for wetland managers to improve hydrological performance of these conserved lands?

| Study area
The study area of this research is the 914 wetland sites in Nebraska under the WRP, WMA, and WPA programs.This study measured the quantitative monitoring and assessment of inundation and hydrological dynamics conditions within the boundary of Nebraska.Conserved land sites that overlapped Nebraska's state boundary were excluded to ensure the accuracy of the assessment.Finally, this study conducted a detailed mapping and analysis of all conserved land sites, with the locations of these sites shown in Figure 1.

| Data sources
In this project, the GEE was utilized to obtain and preprocess remote sensing data from the selected research area.The acquisition of Sentinel-2 multispectral images played a crucial role in training and classifying the data.The Sentinel-2 system consists of two satellites and provides a temporal resolution of 5 days, allowing for the regular acquisition of images for the study area on a weekly basis.The datasets were imported into the GEE platform and processed using its built-in command.The Sentinel-2B satellite was launched in March 2017, which allowed for the full functionality of the Sentinel-2 system in 2018.For this research, data from 2018 to 2021 were utilized, with a focus on the months between March and November due to the climate of Nebraska.This range of months was chosen to avoid snow and ice coverage in the study area.
To enhance the training accuracy in this study, index bands were calculated and incorporated.The following index bands were included for training: NDWI, calculated as NDWI = (B3 − B8)/(B3 + B8); NDVI, NDVI = (B8 − B4)/(B8 + B4); MDNWI, MNDWI = (B3 − B11)/(B3 + B11); and NDMI, NDMI = (B8 − B11)/(B8 + B11).These index bands were computed using code in GEE, and the Sentinel-2 spectral data were directly imported into the GEE code editor.The shapefiles for the conserved sites were acquired from the Nebraska Office of the USDA-NRCS.After preprocessing with QGIS is a free and open-source cross-platform desktop geographic information system (GIS) application, these shapefiles were uploaded to Google Drive, where they could be directly imported and processed within the GEE framework.

| Data analysis
The schematic representation of workflow for this study is depicted in Figure 2. The data mentioned earlier, which include shapefile data with the boundaries of the research area, are available on GEE and can be acquired using an image collection code.Classification results from GEE as TIFF files are exported in QGIS and convert them into shapefile.These shapefiles are utilized for the calculation of change in water cover from 2018 to 2021.The water cover data for all conserved sites are used to calculate surface water cover rate inundation frequency.For each site, at least 25 images are used for this purpose.In the last water cover rate data, the inundation frequency mapping was performed through QGIS and ArcGIS software.

| Accuracy assessment of the site's classification
To evaluate accuracy, a testing sample representing 30% of the input data was utilized.The OA, recall, precision, and F1 score were calcu- condition that is present.Specifically for water, a TP means that pixels classified as water are water in the real world.The F1 score, which is calculated using precision and recall, ranges from 0 to 100 and serves as an accuracy metric for understanding how machine learning methods can classify multispectral land cover image data in natural conservation land settings in Nebraska.

| RE SULTS
We conducted water classifications to assess the performance of different machine learning classifiers available on the GEE platform (Zhang et al., 2022).The mean OA for all classifications was 99.79%, ranging from 95.45% to 100%.The mean recall was 99.57%, ranging from 87.53% to 100%.Mean precision was 99.21%, ranging from 90.60% to 100%.The mean F1 score was 99.36%, ranging from 93.35% to 100%.These results indicate that the SVM with a linear kernel classifier is accurate and reliable for surface water classification in Nebraska during the growing season.The results of the OA assessment based on Zhang et al. (2022) research are presented in Table 2. Additionally, we verified the classification results using ground truth data.When comparing the Sentinel-2 classification results with the Rainwater Basin's annual habitat survey data in 2020, we found that 72 conserved land sites with visible surface water from the Sentinel-2 classification were 100% verified in the annual habitat survey data.The total area of surface water in conserved sites within the Rainwater Basin was 2.80 km 2 as counted by the annual habitat surveys, while our classification result covered 2.46 km 2 , accounting for 87.68% of the total surface water area.
Figure 3 and Table 3 present the results, highlighting a significant representation of surface water conditions across all WRP sites throughout the 4-year period from 2018 to 2021.In Figure 3, each dot represents the observed surface water cover rate.The median surface water cover rate for all sites over the 4-year period was 3.33%.Specifically, in 2019, there was a notably wetter condition with a median surface water cover rate of 4.70%, varying from 0% to 99.9%.The median surface water cover rates for the remaining 3 years were as follows: 2018: 2.92%, 2020: 3.25%, and 2021: 2.90%.These values were lower than the median surface water cover rate for all sites during the 4-year period.

TA B L E 1
The geospatial data sources from the Sentinel-2 satellites.

TA B L E 2
Accuracy assessment of testing sample using algorithm in Google Earth Engine (GEE) platform (Zhang et al., 2022).The data presented in Figure 4 and Table 4 demonstrate the surface water conditions of all WMAs sites in the study period from 2018 to 2021.The median for all sites of all 4 years is 9.0%, with 2021 showing a relatively higher wetter condition with a median surface water cover rate of 12.02%, ranging from 0% to 95.72%.The medians for the other 3 years (2018: 8.66%, 2019: 7.73%, and 2020: 7.63%) are less than the median of all sites for all 4 years.

Algorithms
Figure 5 and Table 5  The analysis of surface water classification results revealed that the average surface water cover rate for all WRP and WPA sites was 8.07% and 7.51%, respectively.Out of these sites, eight (1.21%) showed no surface water presence throughout the 4-year study period, while 178 sites (26.90%) consistently had surface water.Additionally, 475 sites (71.75%) experienced partial ponding at some point during the 4 years.
The research also highlighted that 64.64% of the total conserved area in WRP and WPA remained unaffected by inundation, indicating potential for improving the hydrological functions of these lands.
Figure 6 illustrates the surface water cover rate by area in each category of conserved land sites, while Water cover rate in all WPA sites during 2018-2021.This figure represents a comprehensive analysis of the water cover dynamics within these conserved areas over a 4-year period.
control.WPAs have the lowest surface water cover rate among all three types of conserved sites, with significantly less water than the other two (WRP and WMA).This is evident in the data for each of the 4 years, suggesting it is the least conserved land type.The year 2021 was considered as a wet year for all WMAs conserved sites.The yearly mean water cover rate for all WRPs conserved sites are 6.99%, 12.99%, 6.44%, and 5.87% for the years 2018, 2019, 2020, and 2021, respectively.WRPs sites have an average mean cover rate of 8.07% for the 2018-2021 study period time.For WRPs sites, 2019 is considered as wet year with higher mean water cover rate 12.99%.The WPAs sites have the lowest mean water cover rate 4.40%, 13.03%, 5.00%, and 7.60% for the years 2018, 2019, 2020, and 2021, respectively.
The 4-year average water cover rate mean for all WPAs site was 7.51%.The other analysis shows that 2018 was a relatively dry year for all three conserved sites.

TA B L E 5
The percentage of inundated areas on WPAs conserved footprints during 2018-2021.

F I G U R E 6
Comparison of water cover rate in three types of conservation lands in each year.

TA B L E 6
The percentage of inundated areas on all conserved footprints during 2018-2021.2018 and 2019.In the context of the WPAs, the conserved land exhibited the highest water coverage (29.65%) in June 2019.Conversely, the lowest water coverage (0%) was recorded for a duration of 8 months within the preceding 4-year period.Figure 8 shows that this site was highly inundated during the wet season.The results for this site showed that substantial differences occurred over the years.The WPAs sites have the highest mean surface water cover rate was 29.65% in 2019, which was much higher than that of the other 3 years, while the lowest was 0.18% in 2018.
Figure 9 shows the surface water cover rate for conserved lands (WRPs, WPAs, and WMAs) during the fall of 2021.Results indicate that WMAs had the highest rate of surface water cover compared to the other conserved sites during the month of October.The surface water cover rate in the month of October during 2018, 2019, 2020, and 2021 was 13.86%, 14.87%, 13.18%, and 15.57%, respectively, at WMAs sites.In October 2021, the WRPs and WPAs conserved lands had a relatively low percentage of 3.33% and 2.96%, respectively, compared to the WMA conserved sites.

| DISCUSS ION
This research presents a methodology that combines Sentinel-2 imagery, GEE, and machine learning models to detect surface water inundation status at conserved land sites in Nebraska.It explored an economical and efficient way to observe big land zones with high-resolution satellite imagery and demonstrated that the approach is dependable and can be utilized for long-term, continuous studies.Furthermore, conserving existing wetlands can be a cost-effective approach compared to other methods aimed at mitigating damages.Wetland conservation offers comparable levels of damage mitigation while potentially requiring lower financial investment (Barbier, 2015;Boutwell & Westra, 2016).
Significant lessons were gained related to the utilization of machine learning and GEE for conserved site assessments, monitoring, and wetland conservation research.
This research presents a cost-effective, long-term wetland mapping tool that can be utilized for monitoring hydrological performance and conservation efforts (Lang & McCarty, 2009).The findings of Pekel et al. (2016) and Donchyts et al. (2016)  approach presented in this study could be beneficial for hydrological dynamic assessment and monitoring of conserved land.Previous studies on LULC classification, particularly concerning wetlands, were supported by the classification results of our research (Ma et al., 2017;Shao & Lunetta, 2012;Sheykhmousa et al., 2020;Zhang et al., 2022).This research can assist in recognizing and ranking wetland conservation land that may require additional restoration.
Surface water conditions in conserved sites were found to vary, influenced by multiple contextual factors.Topography (e.g., upland or lower lands, slopes), human activities (e.g., irrigation, agricultural pits), conservation practices (e.g., pumping, sediment removal), and watershed context (e.g., soil types, vegetation cover, drainage pattern) all play a significant role in determining the hydrological performance.As many of the conservation land sites studied contain wetlands, changes in hydrological dynamics are particularly significant as they can alter water depth and hydrophyte communities, as well as wetland edges (Tang et al., 2014;Whyte et al., 2018).The surface water cover maps created as part of this study provide valuable insights into the changes in surface water of wetland-related conservation sites.
In this research, the limitations of Sentinel-2 data should be acknowledged, and improvements should be implemented in future studies.These limitations can be summarized in terms of climate and resolution: Satellite data availability is greatly affected by climatic conditions.This research examined hydrological dynamics of conservation land sites spanning from 2018 to 2021, dividing them into three distinct categories.In May, the typically wet spring season in Nebraska, the most difficult conditions were encountered in trying to obtain cloud-free images for each site.This study was set up with the intention of getting weekly satellite imagery for a more precise evaluation of all conservation sites in the state.The rainy season in Nebraska can sometimes lead to missing data in May, a potential issue that may be unavoidable.To address this limitation, long-term data collection or manual data collection from aerial or field surveys could be employed.Sentinel-1 SAR can be used for land observation despite cloud cover in our future studies.The recently released Dynamic World dataset is a useful resource for conducting global LULC studies, which may help alleviate this issue (Brown et al., 2022;Huang et al., 2018;Mahdavi et al., 2018).The spatial resolution of 10 m used in most of the Sentinel-2 bands was insufficient to accurately detect small streams with widths smaller than 10 m, as well as FNs due to algae cover in small inundation areas.Furthermore, this resolution was not sufficient for vegetation and soil classification.To improve the outcome of the study at a more precise level, data collection with higher spatial resolution should be used, such as drone-based imagery.In addition, the Sentinel data can only be used to evaluate the hydrological performance, the influential factors (e.g., such as land management practices, surrounding land use, or hydrological connectivity) which have caused the variations of wetland hydrologic performance still need more experimental data to examine their relationships.Moreover, more research is still needed to understand the implications of wetland hydrological changes in surface water cover on wetland biodiversity, ecosystem functioning, and habitat suitability.

| Policy recommendations
This study confirms the conservation values in wetland hydrological performance and supports the necessity of adaptive management for wetland conservation programs.Results revealed water boundaries and inundation conditions to be constantly fluctuating.Utilizing the high temporal resolution of Sentinel-2, inundation conditions of all protected land sites across the state can be calculated and mapped more frequently.Due to their isolation, wetlands in remote areas can be difficult to access when conducting traditional systematic research.To assess the effects of nearby agricultural properties or road construction on these conserved lands, a rapid method is proposed to determine the impact of such activities on their hydrological conditions.This can help to prevent potential degradation of conserved land.This study suggests that areas with frequent inundation should be considered for future conservation programs.Based on the findings, the following recommendations are provided for the management of preserved lands: to enhance flood mitigation strategies, promote effective water resource management, and maintain biodiversity.
The study revealed that a high percentage of wetland sites had experienced regular or partial inundation over the past 4 years.To protect these sites at a watershed scale, it is recommended that hydrological restoration or partial recovery be undertaken.Such restoration, either full or partial, is a key step for conserving wetlands and protecting them at the watershed level.The findings revealed that there were only a few areas of inundation at the conservation land sites.Furthermore, the ponding size, frequency, and duration can be improved through monitoring of hydrological dynamics at the watershed scale.Tang et al. (2018) suggested employing hydrological restoration treatments to enhance the performance of wetlands.Various treatments, such as sediment removal, drain closure, and irrigation reuse pit closure, can be employed to conserve and revive wetland functions.These treatments, when implemented at the conserved land sites studied in this research, can promote sustainable conservation and maintain the wetland functions across the state.In addition, vegetation plays a crucial role in conservation land, and one commonly employed tool for conservation management is vegetation buffers.These buffers serve as a protective barrier, shielding wetlands from agricultural pollutants and sediment by regulating water runoff through upland vegetation.By slowing down the flow of water, sediment is given the opportunity to settle out of the water column before reaching the wetland (Drahota & Reichart, 2015).Restoring vegetation is a long-term initiative that contributes to the enhancement of native wetland vegetation communities.
This study employed a pixel-based classification model for geospatial analysis of all conservation sites.The Sentinel-2 images from March to November for the study period(2018)(2019)(2020)(2021) are imported into GEE for surface water classification, and shapefile of the conserved land sites used as the boundary in GEE.The GEE geometric tool was utilized to label different land cover classes.In 2021, a field survey was conducted to validate the water and land classes in the training region.Random sampling of all the attributes was done using the random column function in GEE.NDVI, NDWI, NDMI, and MNDWI are computed through GEE's built-in normalized difference function.Subsequently, these index bands are combined with spectral bands as training bands.On the basis of previous studies, the Linear Kernel SVM is selected as machine learning classifiers for the conserved land classifications.In this process, 11 spectral bands and four index bands have been used as the training bands.
lated using the following formulas, to compare different classifiers with varying parameters: OA = ((TA + TN)/(TP + TN + FP + FN)), Recall = (TP/ (TP + FN)), Precision = (TP/(TP + FP)), and F1 score = {2 × [(Precision × Recall)/(Precision + Recall)]}.In this context, a true positive (TP) indicates that a detected condition is present, while a true negative (TN) suggests that an undetected condition is absent.A false positive (FP) represents a detected condition that is absent, and a false negative (FN) indicates an undetected F I G U R E 1 Location maps of Wetlands Reserve Program (WRP) sites, Waterfowl Production Areas (WPAs), and Wildlife Management Areas (WMAs) in Nebraska.

F
I G U R E 4 Water cover rate in all WMA sites during 2018-2021, this figure represents the temporal fluctuations of water cover rate in the WMAs during the years 2018, 2019, 2020, and 2021.TA B L E 4 The percentage of inundated areas on WMAs conserved footprints during 2018-2021.

Figure 7
Figure 7 represents the surface water cover change of all three types of conservation land during growing season of each year.High values of surface water cover appear in summer and low values occur in fall.The curves of different conserved sites are compared and WMAs sites show more surface water cover change during the 4 years of study time.The curves show relatively wet summer and dry winter growing season.During the growing season of each year, high values appear for WMAs conserved area with mean area water cover rate 14.61%, 16.07%, 16.22%, and 18.72% for the years 2018, 2019, 2020, and 2021, respectively.The 4-year average mean value of water cover rate is 16.41%.

Figures 8
Figures 8 and 9 show the water cover rate in March 2021 and October 2021, respectively.The highest water cover (38.93%) occurred in May 2021, while the lowest water cover (0%) was observed over 2 months during the past 4 years for all WMAs conserved sites.For the WRPs conserved land, the highest water cover (25.08%) occurred in April 2019 with the lowest cover rate during the month of March in the years are based on the Landsat satellite dataset, which has a resolution of up to 30 m, insufficient to monitor the hydrological dynamics of protected land.Sentinel-2 imagery with a 10-m spatial resolution offers a greater monitoring range compared to Landsat satellite, especially for smaller sites.The mapping F I G U R E 9 Surface water cover rate in October 2021 (peak fall migration season) in all conserved land.

Table 6
cover rates of 8.29%, 13.51%, 8.19%, and 8.48% in 2018-2021, respectively.2019provedto be the wettest of the 4 years, with a surface water cover rate of 13.51%, exceeding the mean value of 9.62%.The conserved land sites under the WMAs had the highest inundated area rate of 16.41%, indicating active hydrological inundation in the floodplain areas.The conserved lands under the WRP reached the mean annual surface water cover rate by area at 8.07%, indicating the core wetland areas were inundated periodically or regularly.Other types of conserved land sites serving for upland conservation purposes had a lower level of inundation but provided critical conservation needs in soil erosion F I G U R E 3 Water cover rate in all WRP sites over a comprehensive 4-year study period spanning from 2018 to 2021.TA B L E 3 The percentage of inundated areas on WRPs conserved footprints during 2018-2021.Yearly surface water cover rate of all WRP sites, n = 600 2018 (%) 2019 (%) 2020 (%) 2021 (%) Total (%)