Establishing the Downscaling and Spatiotemporal Scale Conversion Models of NDVI Based on Fractal Methodology

Abstract

Scale effect is a crucial scientific problem in quantitative remote sensing (RS), and scholars attempt to solve it with scale conversion models, which can characterize the numerical relationship of RS land surface parameters at different resolutions (scales). As a significant land surface parameter, scale conversion of normalized difference vegetation index (NDVI) has been studied for a long time. Therefore, taking NDVI as an example, the development of scaling research is described and analyzed in the paper, and based on fractal theory, the development trends are discussed for land surface parameters in quantitative remote sensing. These are our conclusions: it will be the new trend to establish downscaling models based on fractal theory for land surface parameters in quantitative remote sensing; additionally, it still is the hotspot to establish spatiotemporal scale conversion models for land surface parameters in quantitative remote sensing in the future, and addressed on that, the multi-fractal scaling methodology is proposed, and its availability is analyzed in the paper, which presents significant potential.

A New Spatial Downscaling Method for Long-Term AVHRR NDVI by Multiscale Residual Convolutional Neural Network

“…This strategy facilitated the extraction of a broader range of image features, yielding significant results, albeit with certain limitations in the context of multiperiod images [29,30]. To better efficiently realize the feature transfer, Lim et al introduced the Multiscale Downscaling Architecture (MDSR) that allowed parameter sharing among multiple architectures, expediting convergence and reducing parameter count [31]. Furthermore, Wang et al proposed the Residual Network (ResNet), a method that integrates deep and shallow strategies in network design and utilizes an external network to efficiently transfer gradients, thereby enhancing learning effectiveness and overall model performance.…”

“…The current research focus leans towards enhancing various aspects of the neural network model, including architecture, activation function, and optimization methods [24,25,42,43]. Through comparative analyses, improvement of visual representation activation function [43][44][45] and optimization methods [46,47] can improve the computational efficiency and providing more accurate mappings than traditional techniques [31,48]. > REPLACE THIS LINE WITH YOUR MANUSCRIPT ID NUMBER (DOUBLE-CLICK HERE TO EDIT) < Despite the numerous studies on downscaling algorithms and the relentless efforts of many researchers, the aforemen-tioned issues have been partially resolved.…”

Can urban green space water planes be quantified? A new way of thinking about landscape planning and design

“…This paper investigates the planar morphology of water bodies in urban green spaces based on fractal theory, fractal theory takes complex irregular entities as the research object, can quantify and analyze objects with complex morphology in nature (Luan, 2020), and has increasingly become a powerful tool for studying objective entities. Since it was first proposed, fractal theory has been combined with various disciplines, such as geography, urban planning, aesthetics, and architecture, but its utilization in the field of landscape architecture is still in its infancy (Lagarias and Prastacos, 2020).…”