Wetland delineation / state
Satellite(s)Landsat ETM+, ASTER, SPOT, Sentinel-1, Sentinel-2. | Monitoring elementLand surface reflectance, SAR backscatter. |
---|---|
Description techniqueAllan (2016) prepared a literature review on the remote sensing of Waikato wetlands. To date most techniques for wetland delineation have utilised supervised classification to extract relevant pixels from multispectral imagery based on spectral signature. Landsat is the most common choice due to the relatively fine resolution and spectral range of its imagery; and extensive catalogue (Guo et al, 2017). Sentinel-2 imagery, while lacking the deep archive of Landsat catalogue, has also been successfully applied to wetland delineation, and offers a higher revisit frequency and higher spatial resolution (Slagter et al, 2020). Commonly applied algorithms include random forests, support vector machines and neural nets (Chatziantoniou et al, 2017, Mahdavi et al, 2018). The utilisation of multitemporal imagery has been shown to significantly improve the accuracy of wetland delineation by taking advantage of the dynamic nature of these environments, namely vegetation phenology and water level (Mahdavi et al, 2018, Amani et al, 2019). OBIA (Object Based Image Analysis) has also been shown to be effective for wetland delineation and classification by introducing textural features along with the spectral information (Chatziantoniou et al, 2017). A growing number of studies have demonstrated the utility of SAR data and data hybridisation to wetland classification (e.g. Slagter et al, 2020, Chatziantoniou et al, 2017). Using SAR data by itself can be difficult, as is the use of single polarised data (Mahdavi et al, 2018). However, recent development within the SAR processing space and the deployment of sensors such as Sentinel-1 has allowed both wetland delineation and change detection to be achieved with SAR data alone, with results comparing favorably to optical sensors such as Landsat (Muro et al, 2016). Repeat classification of the type outlined above has been used to monitor wetland extent and degradation, with both short term (monthly) change (e.g. Muro et al, 2016) to long term (decadal, e.g. Han et al, 2015), using optical, SAR and a combination of both. Wetland degradation has been monitored using RS via changing cover class (e.g. Han et al, 2015) and time series vegetation index trends (e.g. Alonso et al, 2016). Wulder et al (2018) undertook a 33 year Landsat time series analysis of Canadian wetlands, plotting loss trends for both tree covered and open wetland. By comparing each annual classification, the authors were able to date the loss of wetland area, as well as determining the current persistence (in years) of areas classified as wetland. Significant regional variations in wetland extent were found, with both significant growth and loss recognized. Authors do caution that field data should also be used to corroborate classification findings (Mahdavi et al, 2018), a conclusion that Allan (2016) also reached - suggesting that RS technologies are a great compliment to field monitoring programmes, but are not sufficient to replace in situ monitoring at this stage. | Accuracy / ResolutionVariable spatial and temporal resolution according to sensors. |
Case studyAllan et al. (2020) utilised combined Sentinel-1 and Sentinel-2 data (extracted from Google Earth Engine) and computed spectral and backscatter metrics within existing Land Cover Data Base (LCDB) wetland polygons. This data was then used to train several machine learning models (Support Vector Machine and Neural Net) to perform a wetland area classification of segments extracted from the test region. With post-classification filtering, SVM reached 81% accuracy. Slagter et al (2020) describe an approach in which hybrid Sentinel-1 and Sentinel-2 data are used for wetland delineation and classification in South Africa. Using a Random Forest classification, combining the multitemporal SAR and optical sources significantly improved accuracy, with 69.1% reported when using 10 classes and 76.4% with 7 classes (confused classes collapsed). Amani et al (2018) used Google Earth Engine (GEE) to classify 30,000 Landsat 8 scenes captured between 2016 and 2018 to build a nationwide map of wetland areas with five classes separating bog, fen, marsh, swamp and shallow water. Applying a Random Forest classification, users and producers accuracies averaged 63 and 66% respectively. The authors noted that utilising multi temporal data captured during different seasons significantly improved classification accuracy. | |
Also fits domainFreshwater. | |
Benefits
| Limitations
|
Applicability for NorthlandYes. Northland has a large number of wetlands of sufficient extent to apply RS-derived techniques for delineation and monitoring. Like other applications utilising spectral reflectance, the number of observations available will be limited by cloud cover, which might hinder year-round monitoring. Year on year monitoring should be feasible across the region, e.g., Allan et al. (2020) achieved promising results from their classification approach to wetland delineation in the Waikato region but suggested that care must be taken for small wetland fragments which are present in Northland. Techniques applying optical data will be limited in coverage and temporal granularity by the persistent cloud cover in the region, particularly during the winter months. Mature cloud-masking techniques are directly available for open access multispectral data (e.g. Landsat and Sentinel-2). When using commercial data, care must be taken to ensure that there is sufficiently cloud free imagery available, as cloud masking is not as mature and ordering a large volume of imagery to ensure complete cloud free coverage between multiple observations can become cost prohibitive. | |
Publications references Allan M 2016. Remote sensing of Waikato wetlands; a literature review. Environmental Research Institute, University of Waikato, Hamilton. 30 p. Allan M, Westerhoff R, Borman D, Briggs C, McLeod J, Dutton P, Lim N, Hamilton D, Hicks B, Brabyn L, Muraoka K. 2020. Remote sensing of Waikato lakes water quality and remote sensing of wetland extent. Powerpoint presentation. https://gitlab.com/envirosattools/documentation/-/raw/master/M_Allan_Envirosat_10-12-2020.pdf Mahdavi, S., Salehi, B., Granger, J., Amani, M., Brisco, B. and Huang, W., 2018. Remote sensing for wetland classification: A comprehensive review. GIScience & Remote Sensing, 55(5), pp.623-658. Slagter, B., Tsendbazar, N.E., Vollrath, A. and Reiche, J., 2020. Mapping wetland characteristics using temporally dense Sentinel-1 and Sentinel-2 data: A case study in the St. Lucia wetlands, South Africa. International Journal of Applied Earth Observation and Geoinformation, 86, p.102009. Alonso, A., Muñoz-Carpena, R., Kennedy, R.E. and Murcia, C., 2016. Wetland landscape spatio-temporal degradation dynamics using the new Google Earth Engine cloud-based platform: Opportunities for non-specialists in remote sensing. Transactions of the ASABE, 59(5), pp.1331-1342. Amani, M., Mahdavi, S., Afshar, M., Brisco, B., Huang, W., Mohammad Javad Mirzadeh, S., White, L., Banks, S., Montgomery, J. and Hopkinson, C., 2019. Canadian wetland inventory using google earth engine: The first map and preliminary results. Remote Sensing, 11(7), p.842. Dronova, I., 2015. Object-based image analysis in wetland research: A review. Remote Sensing, 7(5), pp.6380-6413. Kaplan, G. and Avdan, U., 2017. Mapping and monitoring wetlands using Sentinel-2 satellite imagery. ISPRS Annals of Photogrammetry, Remote Sensing & Spatial Information Sciences, 4. Faber-Langendoen, D., Nichols, W., Rocchio, J., Walz, K., Lemly, J., Smyth, R. and Snow, K., 2016. Rating the condition of reference wetlands across states: NatureServe Ecological Integrity Assessment method. National Wetlands Newsletter, 38(3), pp.12-16. Muro, J., Canty, M., Conradsen, K., Hüttich, C., Nielsen, A.A., Skriver, H., Remy, F., Strauch, A., Thonfeld, F. and Menz, G., 2016. Short-term change detection in wetlands using Sentinel-1 time series. Remote Sensing, 8(10), p.795. Chatziantoniou, A., Psomiadis, E. and Petropoulos, G.P., 2017. Co-Orbital Sentinel 1 and 2 for LULC mapping with emphasis on wetlands in a mediterranean setting based on machine learning. Remote Sensing, 9(12), p.1259. Han, X., Chen, X. and Feng, L., 2015. Four decades of winter wetland changes in Poyang Lake based on Landsat observations between 1973 and 2013. Remote Sensing of Environment, 156, pp.426-437. https://www.sciencedirect.com/science/article/pii/S003442571400399X | |
Other comments or informationHyperspectral data captured from airborne-based images has the potential to gain both high spectral resolution and high spatial resolution information. However, airborne-based remote sensing techniques can be costly, which often limits them to applications over small areas or at a low temporal resolution. |