Constructing a High Temporal Resolution Global Lakes Dataset via Swin-Unet with Applications to Area Prediction

TL;DR

This work tackles the need for high-temporal-resolution global lake dynamics by introducing GLAKES-Additional, a biennial dataset for $152{,}567$ lakes from $1990$ to $2021$, produced with Swin-Unet to improve delineation on high-resolution satellite imagery. The pipeline fuses JRC Monthly Water History, Global Surface Water Occurrence, GRWL, GSHHS, and CRU TS climate inputs to generate CLAKES-Additional and enables LSTM-based forecasts of lake-area changes, achieving an RMSE of $0.317\ \mathrm{km^2}$ on the test set. Key contributions include an end-to-end workflow for high-resolution lake extraction, contour post-processing and indexing with GLAKES, and climate-driven area prediction that reveals regional patterns in lake area changes. The dataset and methods support improved monitoring and forecasting of freshwater resources under climate change and anthropogenic stress, aiding policy and management decisions.

Abstract

Lakes provide a wide range of valuable ecosystem services, such as water supply, biodiversity habitats, and carbon sequestration. However, lakes are increasingly threatened by climate change and human activities. Therefore, continuous global monitoring of lake dynamics is crucial, but remains challenging on a large scale. The recently developed Global Lakes Area Database (GLAKES) has mapped over 3.4 million lakes worldwide, but it only provides data at decadal intervals, which may be insufficient to capture rapid or short-term changes.This paper introduces an expanded lake database, GLAKES-Additional, which offers biennial delineations and area measurements for 152,567 lakes globally from 1990 to 2021. We employed the Swin-Unet model, replacing traditional convolution operations, to effectively address the challenges posed by the receptive field requirements of high spatial resolution satellite imagery. The increased biennial time resolution helps to quantitatively attribute lake area changes to climatic and hydrological drivers, such as precipitation and temperature changes.For predicting lake area changes, we used a Long Short-Term Memory (LSTM) neural network and an extended time series dataset for preliminary modeling. Under climate and land use scenarios, our model achieved an RMSE of 0.317 km^2 in predicting future lake area changes.

Figures (9)

• Figure 1: Global 0.5°x0.5° gridded map depicting the lake count per grid cell, derived from the GLAKES-Additional dataset obtained in this study.
• Figure 2: The workflow for constructing a high-resolution global lakes dataset using Swin-UNet, and predicting lake area changes utilizing a stacked LSTM network driven by meteorological data.
• Figure 3: (a) Detailed structure of the Swin Transformer Block. (b) Architecture of the Long Short-Term Memory (LSTM) network.
• Figure 4: Loss curve during the (a) training and (b) testing process of Swin-Unet.
• Figure 5: Evaluation indexes of the semantic segmentation results. (a) Mean Intersection over Union (b) Pixel Accuracy.