Estimation of Forest Canopy Fuel Moisture Content in Dali Prefecture by Combining Vegetation Indices and Canopy Radiative Transfer Models from MODIS Data

Forest fires pose significant and formidable challenges in today’s world. Countries across temperate, subtropical, and tropical regions are grappling with the menace of forest fires due to the escalating impacts of global warming and heightened weather extremes [1,2]. The accurate and efficient early warning and monitoring of forest fires pose a critical issue that governments must presently confront [3].

Forest fire events are influenced by multiple interacting factors. According to Pyne’s wildfire triangle model [4], the occurrence of wildfires is primarily determined by climate, topography, and fuels. Currently, early-warning methods for forest fires can be classified into three main categories: fire weather forecasts, forest fire occurrence forecasts, and forest fire behavior forecasts. The advancement of satellite remote sensing and internet big data technology has facilitated the integration of weather conditions, combustible materials, and fire sources, leading to a mainstream approach known as forest fire forecasting, which aims to predict the risk of forest fires [5].

Due to the complexity of the forest ecosystem, various factors in the wildfire triangle model interact with each other within the natural environment [6]. This phenomenon means that the influence of other factors on the relevant parameters of many forest areas cannot be ignored in forest fire forecasting [7,8]. Undoubtedly, this complexity and difficulty increase the challenges associated with forest fire forecasting.

Forest canopy fuel moisture content (FMC) acts as an important indicator reflecting the dryness or wetness levels of wild vegetation while also significantly contributing to evaluating susceptibility to forest fires. Previous research has demonstrated that when the FMC falls below 100%, there is a significant increase in forest fire occurrence probability [9]. The sensitivity of FMC to weather conditions and internal vegetation dynamics makes it one of an exemplary factor for forecasting forest fires among numerous others [10]. Consequently, the precise retrieval of FMC through remote sensing assumes paramount importance as a fundamental step in evaluating forest fire risk.

Traditional methods for measuring FMC suffer from several limitations, such as their high demand for manpower and resources. Consequently, they are generally inefficient and unable to comprehensively cover vast forest areas. However, satellite remote sensing technology offers a solution by providing real-time observations across extensive regions [11]. This technological advancement overcomes the inefficiencies associated with traditional physical parameter measurement methods and facilitates the prolonged monitoring of FMC in large forest regions [12,13].

The most direct approach to quantitatively inverting FMC by using remote sensing is to establish a correlation between vegetation canopy spectra and FMC. Some researchers have explored this relationship through spectroscopic and spectral analyses, aiming to construct empirical models that link spectral indices to FMC. These models rely on spectral measurements of vegetation and corresponding field data on FMC [14,15]. However, the construction of such empirical models requires a substantial number of field observations and data samples. Additionally, these empirical models are area-specific in nature. As a result, many researchers are inclined towards utilizing physical models for the quantitative retrieval of FMC. Numerous studies have indicated that FMC can be approximated by two parameters within the PROSPECT model [16] of leaf reflectance: equivalent water thickness (EWT) and dry matter content (DMC). Therefore, radiative transfer models can be employed to simulate vegetation spectral reflectance and estimate FMC [17]. Commonly used models include the two-dimensional PROSAIL model [18], the Liberty model [19], the GEOSAIL model [20], and the three-dimensional DART model [21,22,23]. The merit and demerit of the FMC estimation methods described above are presented in Table 1.

The PROSAIL model, which integrates the leaf optical model PROSPECT with the canopy radiative transfer model SAIL, is extensively employed for biophysical and chemical parameter retrieval, as well as spectral simulation. It has demonstrated high accuracy in estimating FMC for uniformly distributed vegetation. Numerous researchers have utilized the PROSAIL model to estimate FMC in grasslands and crops [24]. Previous studies have indicated significant correlations between various indices, such as NDVI, EVI, NDWI, and NDMI, and FMC, and regression models incorporating these indices along with measured FMC data achieve remarkable accuracy [25]. However, when dealing with forests characterized by tall and irregular canopy structures along with complex understory environments, the selection of suitable models for FMC estimation becomes more diverse [26,27].

Further research and analysis have shown good agreement between vegetation parameters derived from physical models or look-up table-based methods and measured data [28]. Multiple-model analyses and accuracy evaluations of inverse canopy FMC estimation further revealed that the PROGEOSAIL model outperformed the Liberty model for sparse broadleaf and mixed conifer forests [29,30].

In response to the complex forest structure, improvements are being made to radiative transfer models. For the prevalent “tall trees + low vegetation” double-layered forest structures, some scholars have proposed combining the PROSAIL model with the PROGEOSAIL model to construct a look-up table approach, approximating FMC through a loss function [31]. While estimation of FMC based on physical models enhances applicability compared with empirical models, the presence of loss functions in the look-up table method creates an error between estimated and real values of FMC.

The objective of this research is to investigate the correlation between vegetation index/water index and FMC by examining their numerical relationship to develop a model for estimating FMC. Section 2 provides an overview of the study area and data utilized in this work, and the method of combining vegetation indices and canopy radiative transfer models to estimate FMC is presented. The results are shown in Section 3, followed by a discussion in Section 4. Finally, the conclusions are drawn in Section 5.