Predicting Ground Cover with Deep Learning Models—An Application of Spatio-Temporal Prediction Methods to Satellite-Derived Ground Cover Maps in the Great Barrier Reef Catchments
<p>Spatial distribution of training (green), validation (blue), and testing (red) sample sites. The extent of grazing area on the map corresponds to the grazing native vegetation class from the Queensland Land Use Mapping [<a href="#B57-remotesensing-16-03193" class="html-bibr">57</a>].</p> "> Figure 2
<p>An overview of the workflow.</p> "> Figure 3
<p>Overview of the model architecture.</p> "> Figure 4
<p>Time series of ground cover prediction (<b>a</b>) and scoring metrics (<b>b</b>,<b>c</b>) for PredRNN. In plot (<b>a</b>), the solid and dash lines are the predicted and target ground cover at prediction step 1, respectively. The grey bars are the cloud cover rate, and the black vertical line indicates the date with smallest ground cover. The solid line in the plot (<b>b</b>,<b>c</b>) shows the spatial average of MAE and SSIM at prediction step 1, respectively, and the bars are the corresponding decay rate from prediction step 1 to step 8 (season 9~16 in target sequence in <a href="#remotesensing-16-03193-f003" class="html-fig">Figure 3</a>).</p> "> Figure 5
<p>Comparison of predicted and target ground cover maps for different prediction steps of the data sequence highlighted in <a href="#remotesensing-16-03193-f004" class="html-fig">Figure 4</a>. The first row shows the context images used to generate predictions. The second and third rows show the target images and their predictions. The fourth row shows the absolute difference between the targets and predictions, while the last row shows the histogram comparison of observations (targets) and predictions.</p> "> Figure 6
<p>Statistics of scoring metrics of PredRNN model for training, validation, and testing sites. Plot (<b>a</b>), (<b>b</b>), (<b>c</b>), and (<b>d</b>) are the results for MAE, SSIM, MAE decay, and SSIM decay, respectively. The MAE and SSIM results are the spatial–temporal mean for the first prediction step. Numbers above/below the bars are the median values of the metrics among all 100 sites.</p> "> Figure 7
<p>Comparison of different models and training scales for the arid testing site (site 78) in <a href="#remotesensing-16-03193-f001" class="html-fig">Figure 1</a>. Plots in the first column (<b>a1</b>–<b>e1</b>) shows the temporal variation of accuracy. Ground cover, MAE, and SSIM are based on the prediction step 1. ConvLSTM_site and PredRNN_site (dash lines) are model-trained sites, specifically. ConvLSTM_GBRCA and PredRNN_GBRCA (solid lines) are models trained on the GBRCA scale. The brown vertical line indicates the date with the least ground cover. The boxplot in the second column (<b>a2</b>–<b>e2</b>) summarizes the statistics of the time series in the first column. The blue triangle is the mean. Boxes extend from 25th to 75th percentiles, and whiskers from 5th to 95th percentiles. Circles represent outliers.</p> "> Figure 8
<p>Comparison of different models and training strategies for target images in a single data sequence indicated in <a href="#remotesensing-16-03193-f007" class="html-fig">Figure 7</a>. First row shows results from observation. Second to fifth rows are prediction results from different trained models. The last row shows comparison of histograms.</p> "> Figure 9
<p>Spatial distribution of MAE and SSIM. The two metrics are aggregated as the spatio-temporal mean for the first prediction step. The smoke-white and light-blue backgrounds represent land and sea, respectively. Polygons with different colors are the grazing areas in different catchments. The varying sizes of scatter points indicate the magnitude of scoring metrics.</p> "> Figure 10
<p>Correlation of scoring metrics to different environmental factors. GC_mean and GC_STD are the spatial–temporal mean and standard deviation of ground cover for a site. The solid line in each pair plot is the linear fit for the scatter points. <span class="html-italic">ρ</span> is the Pearson correlation coefficient. The background color gradient indicates the magnitude of <span class="html-italic">ρ</span>.</p> "> Figure 11
<p>Ground cover for grazing areas in the GBRCA. (<b>a</b>,<b>d</b>) are the ground cover maps for June 2022 to August 2022 (the last season in the context sequence) from observation and prediction, respectively; (<b>b</b>,<b>e</b>) are the results for September 2022 to December 2022 (the first season in the target sequence); and (<b>c</b>,<b>f</b>) are the maps for the change in ground cover in the above two seasons. The inset plots in (<b>c</b>,<b>f</b>) compare the histograms of pixels in the two seasons.</p> "> Figure 12
<p>Accuracy of GBRCA ground cover prediction from September 2019 to September 2023. The solid and dashed lines in the first plot are the spatial average of observation and PredRNN prediction. The filled and unfilled bars represent the increment (negative value for decrement) of observation and prediction, respectively. In the second plot, the solid line with error bars represents the MAE and the standard deviation of the absolute error, while the dashed line represents the SSIM.</p> "> Figure 13
<p>Comparison of PredRNN model trained on sites in GBRCA and Burdekin. The first column (<b>a</b>) is the observation for the season from September to December 2022; the second column (<b>b</b>,<b>c</b>) shows predictions for the same season by different trained models. The last column (<b>d</b>,<b>e</b>) shows the errors (i.e., prediction–observation) of the different predictions. Inset plots in the last column compare the histograms for observation and model prediction.</p> ">
Abstract
:1. Introduction
1.1. Motivation
1.2. Related Works
1.3. Research Objectives
2. Study Area and Datasets
2.1. Study Area
2.2. Datasets
2.2.1. Ground Cover Data
2.2.2. Climate Data
2.2.3. Hydrological Data
2.2.4. Data Preparation
3. Methods
3.1. Overview of Methodology
3.2. Spatio-Temporal Models
3.2.1. ConvLSTM Model
3.2.2. PredRNN Model
3.3. Training Setup
3.3.1. Data Split for Different Training Scales
- GBRCA training: As Figure 1 shows, all data sequences in the training sites were used for training, validation sites for validation, and testing sites for testing. Therefore, there were 6400 data sequences used for training (50 sites × 128 data sequences). In this case, a single trained model was obtained for the entire GBRCA, and the model was applied to all sites in this region;
- Site-specific training: Instead of splitting the training, validation, and testing datasets according to sites, for each individual site, the 128 data sequences along the time series were split into 50% training, 25% validation, and the final 25% testing datasets. The model was specifically trained for the site.
3.3.2. Model Structure
3.3.3. Model Configuration
3.4. Model Evaluation
3.5. Performance Variation Analysis
3.6. GBRCA Ground Cover Prediction
4. Results
4.1. Model Evaluation
4.1.1. Single-Site Evaluation
4.1.2. Scoring Metrics for All Sites
4.2. Model Performance Comparison
4.2.1. Comparison of Different Models
4.2.2. Comparison of Different Training Scales
4.3. Model Performance Variation Analysis
4.4. GBRCA Next-Season Ground Cover Prediction
5. Discussion
5.1. Scalability
5.1.1. Impact of the Spatial Density of Training Sites
5.1.2. Time and Resources
5.2. Implications
5.3. Limitations and Future Directions
6. Conclusions
Author Contributions
Funding
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Food and Agriculture Organization of the United Nations. FAOSTAT—Land Use. 2021. Available online: http://www.fao.org/faostat/en/#data/RL (accessed on 27 August 2024).
- ABARES. Land Use of Australia 2010–11 to 2015–16, 250 m; ABARES: Canberra, Australia, 2022. [CrossRef]
- State of Queensland. Reef 2050 Water Quality Improvement Plan 2017–2022. 2018. Available online: https://www.reefplan.qld.gov.au/__data/assets/pdf_file/0017/46115/reef-2050-water-quality-improvement-plan-2017-22.pdf (accessed on 27 August 2024).
- Trimble, S.W.; Mendel, A.C. The cow as a geomorphic agent—A critical review. Geomorphology 1995, 13, 233–253. [Google Scholar] [CrossRef]
- Bartley, R.; Hawdon, A.; Post, D.A.; Roth, C.H. A sediment budget for a grazed semi-arid catchment in the Burdekin basin, Australia. Geomorphology 2007, 87, 302–321. [Google Scholar] [CrossRef]
- Risk, M.J. Assessing the effects of sediments and nutrients on coral reefs. Curr. Opin. Environ. Sustain. 2014, 7, 108–117. [Google Scholar] [CrossRef]
- Baird, M.E.; Mongin, M.; Skerratt, J.; Margvelashvili, N.; Tickell, S.; Steven, A.D.L.; Robillot, C.; Ellis, R.; Waters, D.; Kaniewska, P.; et al. Impact of catchment-derived nutrients and sediments on marine water quality on the Great Barrier Reef: An application of the eReefs marine modelling system. Mar. Pollut. Bull. 2021, 167, 112297. [Google Scholar] [CrossRef]
- McCulloch, M.; Fallon, S.; Wyndham, T.; Hendy, E.; Lough, J.; Barnes, D. Coral record of increased sediment flux to the inner Great Barrier Reef since European settlement. Nature 2003, 421, 727–730. [Google Scholar] [CrossRef] [PubMed]
- Fabricius, K.E. Effects of terrestrial runoff on the ecology of corals and coral reefs: Review and synthesis. Mar. Pollut. Bull. 2005, 50, 125–146. [Google Scholar] [CrossRef]
- Waterhouse, J.; Brodie, J.; Lewis, S.; Mitchell, A. Quantifying the sources of pollutants in the Great Barrier Reef catchments and the relative risk to reef ecosystems. Mar. Pollut. Bull. 2012, 65, 394–406. [Google Scholar] [CrossRef]
- McCloskey, G.L.; Baheerathan, R.; Dougall, C.; Ellis, R.; Bennett, F.R.; Waters, D.; Darr, S.; Fentie, B.; Hateley, L.R.; Askildsen, M. Modelled estimates of fine sediment and particulate nutrients delivered from the Great Barrier Reef catchments. Mar. Pollut. Bull. 2021, 165, 112163. [Google Scholar] [CrossRef]
- O’Mahoney, J.; Simes, R.; Redhill, D.; Heaton, K.; Atkinson, C.; Hayward, E.; Nguyen, M. At What Price? The Economic, Social and Icon Value of the Great Barrier Reef; Deloitte Access Economics: Sydney, Australia, 2017. [Google Scholar]
- Kroon, F.J.; Thorburn, P.; Schaffelke, B.; Whitten, S. Towards protecting the Great Barrier Reef from land-based pollution. Glob. Chang. Biol. 2016, 22, 1985–2002. [Google Scholar] [CrossRef]
- Coggan, A.; Thorburn, P.; Fielke, S.; Hay, R.; Smart, J.C. Motivators and barriers to adoption of Improved Land Management Practices. A focus on practice change for water quality improvement in Great Barrier Reef catchments. Mar. Pollut. Bull. 2021, 170, 112628. [Google Scholar] [CrossRef]
- Australian and Queensland Government. Grazing Water Quality Risk Framework 2017–2022; Australian and Queensland Government: Brisbane, Australia, 2020.
- Pickup, G. New land degradation survey techniques for arid Australia—Problems and prospects. Rangel. J. 1989, 11, 74–82. [Google Scholar] [CrossRef]
- O’Neill, A.L. Satellite-derived vegetation indices applied to semi-arid shrublands in Australia. Aust. Geogr. 1996, 27, 185–199. [Google Scholar] [CrossRef]
- Wallace, J.; Behn, G.; Furby, S. Vegetation condition assessment and monitoring from sequences of satellite imagery. Ecol. Manag. Restor. 2006, 7, S31–S36. [Google Scholar] [CrossRef]
- Jafari, R.; Lewis, M.M.; Ostendorf, B. Evaluation of vegetation indices for assessing vegetation cover in southern arid lands in South Australia. Rangel. J. 2007, 29, 39–49. [Google Scholar] [CrossRef]
- Trevithick, R.; Scarth, P.; Tindall, D.; Denham, R.; Flood, N. Cover under Trees: RP64G Synthesis Report; Department of Science Information Technology, Innovation and the Arts: Brisbane, Australia, 2014. [Google Scholar]
- Scarth, P.; Röder, A.; Schmidt, M.; Denham, R. Tracking grazing pressure and climate interaction-the role of Landsat fractional cover in time series analysis. In Proceedings of the 15th Australasian Remote Sensing and Photogrammetry Conference, Alice Springs, Australia, 13–17 September 2010. [Google Scholar]
- Barnetson, J.; Phinn, S.; Scarth, P.; Denham, R. Assessing Landsat Fractional Ground-Cover Time Series across Australia’s Arid Rangelands: Separating Grazing Impacts from Climate Variability. ISPRS—Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2017, XLII-3/W2, 15–26. [Google Scholar] [CrossRef]
- Guerschman, J.P.; Hill, M.J.; Renzullo, L.J.; Barrett, D.J.; Marks, A.S.; Botha, E.J. Estimating fractional cover of photosynthetic vegetation, non-photosynthetic vegetation and bare soil in the Australian tropical savanna region upscaling the EO-1 Hyperion and MODIS sensors. Remote Sens. Environ. 2009, 113, 928–945. [Google Scholar] [CrossRef]
- Beutel, T.S.; Shepherd, R.; Karfs, R.A.; Abbott, B.N.; Eyre, T.; Hall, T.J.; Barbi, E. Is ground cover a useful indicator of grazing land condition? Rangel. J. 2021, 43, 55–64. [Google Scholar] [CrossRef]
- Queensland Government. Reef Protection Regulations. 2022. Available online: https://www.qld.gov.au/environment/agriculture/sustainable-farming/reef/reef-regulations/about (accessed on 27 August 2024).
- Navin, M.S.; Agilandeeswari, L. Multispectral and hyperspectral images based land use/land cover change prediction analysis: An extensive review. Multimed. Tools Appl. 2020, 79, 29751–29774. [Google Scholar] [CrossRef]
- Barrett, A.B.; Duivenvoorden, S.; Salakpi, E.E.; Muthoka, J.M.; Mwangi, J.; Oliver, S.; Rowhani, P. Forecasting vegetation condition for drought early warning systems in pastoral communities in Kenya. Remote Sens. Environ. 2020, 248, 111886. [Google Scholar] [CrossRef]
- Queensland Government. FORAGE REPORT: Pasture Growth Alert. 2021. Available online: https://www.longpaddock.qld.gov.au/forage/report-information/pasture-growth-alert/ (accessed on 27 August 2024).
- Zhang, B.; Carter, J. FORAGE—An online system for generating and delivering property-scale decision support information for grazing land and environmental management. Comput. Electron. Agric. 2018, 150, 302–311. [Google Scholar] [CrossRef]
- Xie, Z.; Phinn, S.R.; Game, E.T.; Pannell, D.J.; Hobbs, R.J.; Briggs, P.R.; McDonald-Madden, E. Using Landsat observations (1988–2017) and Google Earth Engine to detect vegetation cover changes in rangelands—A first step towards identifying degraded lands for conservation. Remote Sens. Environ. 2019, 232, 111317. [Google Scholar] [CrossRef]
- von Keyserlingk, J.; de Hoop, M.; Mayor, A.G.; Dekker, S.C.; Rietkerk, M.; Foerster, S. Resilience of vegetation to drought: Studying the effect of grazing in a Mediterranean rangeland using satellite time series. Remote Sens. Environ. 2021, 255, 112270. [Google Scholar] [CrossRef]
- Xie, Z.; Zhao, Y.; Jiang, R.; Zhang, M.; Hammer, G.; Chapman, S.; Brider, J.; Potgieter, A.B. Seasonal dynamics of fallow and cropping lands in the broadacre cropping region of Australia. Remote Sens. Environ. 2024, 305, 114070. [Google Scholar] [CrossRef]
- Song, Y.; Kalacska, M.; Gašparović, M.; Yao, J.; Najibi, N. Advances in geocomputation and geospatial artificial intelligence (GeoAI) for mapping. Int. J. Appl. Earth Obs. Geoinf. 2023, 120, 103300. [Google Scholar] [CrossRef]
- Sun, S.; Mu, L.; Feng, R.; Wang, L.; He, J. GAN-Based LUCC Prediction via the Combination of Prior City Planning Information and Land-Use Probability. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2021, 14, 10189–10198. [Google Scholar] [CrossRef]
- Yubo, Z.; Zhuoran, Y.; Jiuchun, Y.; Yuanyuan, Y.; Dongyan, W.; Yucong, Z.; Fengqin, Y.; Lingxue, Y.; Liping, C.; Shuwen, Z. A Novel Model Integrating Deep Learning for Land Use/Cover Change Reconstruction: A Case Study of Zhenlai County, Northeast China. Remote Sens. 2020, 12, 3314. [Google Scholar] [CrossRef]
- Wang, J.; Yin, X.; Liu, S.; Wang, D. Spatiotemporal change and prediction of land use in Manasi region based on deep learning. Environ. Sci. Pollut. Res. 2023, 30, 82780–82794. [Google Scholar] [CrossRef] [PubMed]
- Rumelhart, D.E.; Hinton, G.E.; Williams, R.J. Learning internal representations by error propagation. In Parallel Distributed Processing: Explorations in the Microstructure of Cognition; MIT Press: Cambridge, MA, USA, 1985; Volume 1. [Google Scholar]
- Hochreiter, S.; Schmidhuber, J. Long Short-Term Memory. Neural Comput. 1997, 9, 1735–1780. [Google Scholar] [CrossRef]
- Yuan, Q.; Shen, H.; Li, T.; Li, Z.; Li, S.; Jiang, Y.; Xu, H.; Tan, W.; Yang, Q.; Wang, J.; et al. Deep learning in environmental remote sensing: Achievements and challenges. Remote Sens. Environ. 2020, 241, 111716. [Google Scholar] [CrossRef]
- Freeman, B.S.; Taylor, G.; Gharabaghi, B.; Thé, J. Forecasting air quality time series using deep learning. J. Air Waste Manag. Assoc. 2018, 68, 866–886. [Google Scholar] [CrossRef]
- Yang, Y.; Dong, J.; Sun, X.; Lima, E.; Mu, Q.; Wang, X. A CFCC-LSTM Model for Sea Surface Temperature Prediction. IEEE Geosci. Remote Sens. Lett. 2018, 15, 207–211. [Google Scholar] [CrossRef]
- Reddy, D.S.; Prasad, P.R.C. Prediction of vegetation dynamics using NDVI time series data and LSTM. Model. Earth Syst. Environ. 2018, 4, 409–419. [Google Scholar] [CrossRef]
- Gamboa, J.C.B. Deep learning for time-series analysis. arXiv 2017, arXiv:1701.01887. [Google Scholar]
- Wu, T.; Feng, F.; Lin, Q.; Bai, H. A spatio-temporal prediction of NDVI based on precipitation: An application for grazing management in the arid and semi-arid grasslands. Int. J. Remote Sens. 2020, 41, 2359–2373. [Google Scholar] [CrossRef]
- Shi, X.; Chen, Z.; Wang, H.; Yeung, D.-Y.; Wong, W.-K.; Woo, W.-c. Convolutional LSTM network: A machine learning approach for precipitation nowcasting. Adv. Neural Inf. Process. Syst. 2015, 1, 802–810. [Google Scholar]
- Boulila, W.; Ghandorh, H.; Khan, M.A.; Ahmed, F.; Ahmad, J. A novel CNN-LSTM-based approach to predict urban expansion. Ecol. Inform. 2021, 64, 101325. [Google Scholar] [CrossRef]
- Diaconu, C.-A.; Saha, S.; Günnemann, S.; Zhu, X.X. Understanding the Role of Weather Data for Earth Surface Forecasting Using a ConvLSTM-Based Model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, LA, USA, 18–24 June 2022. [Google Scholar]
- Sefrin, O.; Riese, F.M.; Keller, S. Deep Learning for Land Cover Change Detection. Remote Sens. 2021, 13, 78. [Google Scholar] [CrossRef]
- Ma, Y.; Hu, Y.; Moncrieff, G.R.; Slingsby, J.A.; Wilson, A.M.; Maitner, B.; Zhenqi Zhou, R. Forecasting vegetation dynamics in an open ecosystem by integrating deep learning and environmental variables. Int. J. Appl. Earth Obs. Geoinf. 2022, 114, 103060. [Google Scholar] [CrossRef]
- Kladny, K.-R.; Milanta, M.; Mraz, O.; Hufkens, K.; Stocker, B.D. Enhanced prediction of vegetation responses to extreme drought using deep learning and Earth observation data. Ecol. Inform. 2024, 80, 102474. [Google Scholar] [CrossRef]
- Wang, Y.; Wu, H.; Zhang, J.; Gao, Z.; Wang, J.; Yu, P.; Long, M. Predrnn: A recurrent neural network for spatiotemporal predictive learning. IEEE Trans. Pattern Anal. Mach. Intell. 2023, 45, 2208–2225. [Google Scholar] [CrossRef]
- Yu, W.; Lu, Y.; Easterbrook, S.; Fidler, S. Efficient and information-preserving future frame prediction and beyond. In Proceedings of the International Conference on Learning Representations, Addis Ababa, Ethiopia, 26–30 April 2020. [Google Scholar]
- Liu, Z.-x.; Zhang, D.-g.; Luo, G.-z.; Lian, M.; Liu, B. A new method of emotional analysis based on CNN–BiLSTM hybrid neural network. Clust. Comput. 2020, 23, 2901–2913. [Google Scholar] [CrossRef]
- Wang, J.; Jaboyedoff, M.; Chen, G.; Luo, X.; Derron, M.-H.; Hu, Q.; Fei, L.; Prajapati, G.; Choanji, T.; Luo, S.; et al. Landslide susceptibility prediction and mapping using the LD-BiLSTM model in seismically active mountainous regions. Landslides 2024, 21, 17–34. [Google Scholar] [CrossRef]
- Kim, I.-S.; Latif, K.; Kim, J.; Sharafat, A.; Lee, D.-E.; Seo, J. Vision-Based Activity Classification of Excavators by Bidirectional LSTM. Appl. Sci. 2023, 13, 272. [Google Scholar] [CrossRef]
- Gao, Z.; Shi, X.; Wang, H.; Zhu, Y.; Wang, Y.; Li, M.; Yeung, D.-Y. Earthformer: Exploring space-time transformers for earth system forecasting. arXiv 2022, arXiv:2207.05833. [Google Scholar]
- Queensland Government. Land Use Mapping—2021—Great Barrier Reef NRM Regions. 2021. Available online: https://www.data.qld.gov.au/dataset/2021_great_barrier_reef_land_use (accessed on 27 August 2024).
- Naga Rajesh, A.; Abinaya, S.; Purna Durga, G.; Lakshmi Kumar, T.V. Long-term relationships of MODIS NDVI with rainfall, land surface temperature, surface soil moisture and groundwater storage over monsoon core region of India. Arid Land Res. Manag. 2023, 37, 51–70. [Google Scholar] [CrossRef]
- Bartley, R.; Roth, C.H.; Ludwig, J.; McJannet, D.; Liedloff, A.; Corfield, J.; Hawdon, A.; Abbott, B. Runoff and erosion from Australia’s tropical semi-arid rangelands: Influence of ground cover for differing space and time scales. Hydrol. Process. 2006, 20, 3317–3333. [Google Scholar] [CrossRef]
- Turnbull, L.; Wainwright, J.; Brazier, R.E. A conceptual framework for understanding semi-arid land degradation: Ecohydrological interactions across multiple-space and time scales. Ecohydrology 2008, 1, 23–34. [Google Scholar] [CrossRef]
- Department of Environment and Science, Queensland Government. Seasonal Ground Cover—Landsat, JRSRP Algorithm, Australia Coverage (Version 1.0.0). Terrestrial Ecosystem Research Network. 2021. Available online: https://geonetwork.tern.org.au/geonetwork/srv/eng/catalog.search#/metadata/65878a57-f1b0-4e6b-8e7a-8a38ebe7960e (accessed on 27 August 2024).
- Joint Remote Sensing Research Program. Seasonal Fractional Cover—Landsat, JRSRP Algorithm, Australia Coverage (Version 1.0.0). Terrestrial Ecosystem Research Network. 2021. Available online: https://geonetwork.tern.org.au/geonetwork/srv/eng/catalog.search#/metadata/f0c32576-9ad7-4c9c-9aa9-22787867e28b (accessed on 27 August 2024).
- Jeffrey, S.J.; Carter, J.O.; Moodie, K.B.; Beswick, A.R. Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environ. Model. Softw. 2001, 16, 309–330. [Google Scholar] [CrossRef]
- Jones, D.; Wang, W.; Fawcett, R.J.B. High-quality spatial climate data-sets for Australia. Aust. Meteorol. Oceanogr. J. 2009, 58, 233–248. [Google Scholar] [CrossRef]
- Dai, H.; Huang, G.; Wang, J.; Zeng, H.; Zhou, F. Prediction of Air Pollutant Concentration Based on One-Dimensional Multi-Scale CNN-LSTM Considering Spatial-Temporal Characteristics: A Case Study of Xi’an, China. Atmosphere 2021, 12, 1626. [Google Scholar] [CrossRef]
- Bureau of Meteorology Australian Government. Australian Landscape Water Balance. 2022. Available online: https://awo.bom.gov.au/products/historical/soilMoisture-rootZone/4.5,-27.564,134.221/nat,-25.609,134.362/r/d/2022-10-31 (accessed on 27 August 2024).
- Frost, A.; Shokri, A. The Australian Landscape Water Balance Model (AWRA-L v7); Technical Description of the Australian Water Resources Assessment Landscape Model Version, 6; The Australian Bureau of Meteorology: Melbourne, Australia, 2021.
- Bengio, S.; Vinyals, O.; Jaitly, N.; Shazeer, N. Scheduled sampling for sequence prediction with recurrent neural networks. Adv. Neural Inf. Process. Syst. 2015, 1, 1171–1179. [Google Scholar]
- Huber, P.J. Robust Estimation of a Location Parameter. In Breakthroughs in Statistics: Methodology and Distribution; Kotz, S., Johnson, N.L., Eds.; Springer: New York, NY, USA, 1992; pp. 492–518. [Google Scholar]
- Zhou, W.; Bovik, A.C.; Sheikh, H.R.; Simoncelli, E.P. Image quality assessment: From error visibility to structural similarity. IEEE Trans. Image Process. 2004, 13, 600–612. [Google Scholar] [CrossRef]
- Loshchilov, I.; Hutter, F. Decoupled weight decay regularization. arXiv 2017, arXiv:1711.05101. [Google Scholar]
- Bastin, G.; Scarth, P.; Chewings, V.; Sparrow, A.; Denham, R.; Schmidt, M.; O’Reagain, P.; Shepherd, R.; Abbott, B. Separating grazing and rainfall effects at regional scale using remote sensing imagery: A dynamic reference-cover method. Remote Sens. Environ. 2012, 121, 443–457. [Google Scholar] [CrossRef]
- Xu, L.; Cai, R.; Yu, H.; Du, W.; Chen, Z.; Chen, N. Monthly NDVI Prediction Using Spatial Autocorrelation and Nonlocal Attention Networks. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17, 3425–3437. [Google Scholar] [CrossRef]
- Kartal, S.; Iban, M.C.; Sekertekin, A. Next-level vegetation health index forecasting: A ConvLSTM study using MODIS Time Series. Environ. Sci. Pollut. Res. 2024, 31, 18932–18948. [Google Scholar] [CrossRef]
Channel | Data | Category | Data Type | Whether in Output |
---|---|---|---|---|
1 | Ground cover | Target feature | Float | Yes |
2 | Ground cover mask | Mask | Binary | No |
3 | Rainfall | Auxiliary | Float | No |
4 | Temperature | Auxiliary | Float | No |
5 | Soil moisture | Auxiliary | Float | No |
6 | Runoff | Auxiliary | Float | No |
Model | MAE | SSIM | MAE_Decay | SSIM_Decay |
---|---|---|---|---|
PredRNN | 4.41 | 0.66 | 0.21 | −0.018 |
ConvLSTM | 4.72 | 0.64 | 0.19 | −0.015 |
Seasons | MAE | SSIM | Ground Cover (%) |
---|---|---|---|
Summer (DEC–FEB) | 5.43 | 0.75 | 82.27 |
Autum (MAR–JUN) | 4.40 | 0.61 | 88.64 |
Winter (JUL–AUG) | 3.91 | 0.63 | 88.65 |
Spring (SEP–NOV) | 4.84 | 0.71 | 82.06 |
Model Name | GPU Memory | Epoch to Achieve Optimum | Training Time to Achieve Optimum (Hours) | Average Time per Epoch (Hour/Epoch) |
---|---|---|---|---|
ConvLSTM | 16 GB | 210 | 137 | 0.65 |
PredRNN | 32 GB | 9 | 45 | 5 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Mao, Y.; Turner, R.D.R.; McMahon, J.M.; Correa, D.F.; Chamberlain, D.A.; Warne, M.S.J. Predicting Ground Cover with Deep Learning Models—An Application of Spatio-Temporal Prediction Methods to Satellite-Derived Ground Cover Maps in the Great Barrier Reef Catchments. Remote Sens. 2024, 16, 3193. https://doi.org/10.3390/rs16173193
Mao Y, Turner RDR, McMahon JM, Correa DF, Chamberlain DA, Warne MSJ. Predicting Ground Cover with Deep Learning Models—An Application of Spatio-Temporal Prediction Methods to Satellite-Derived Ground Cover Maps in the Great Barrier Reef Catchments. Remote Sensing. 2024; 16(17):3193. https://doi.org/10.3390/rs16173193
Chicago/Turabian StyleMao, Yongjing, Ryan D. R. Turner, Joseph M. McMahon, Diego F. Correa, Debbie A. Chamberlain, and Michael St. J. Warne. 2024. "Predicting Ground Cover with Deep Learning Models—An Application of Spatio-Temporal Prediction Methods to Satellite-Derived Ground Cover Maps in the Great Barrier Reef Catchments" Remote Sensing 16, no. 17: 3193. https://doi.org/10.3390/rs16173193