Simulating Heat Stress of Coal Gangue Spontaneous Combustion on Vegetation Using Alfalfa Leaf Water Content Spectral Features as Indicators

Background: Vegetation heat-stress assessment in the reclamation areas of coal gangue dumps is of 14 great significance in controlling spontaneous combustion. Methods: The study simulated the heat-stress 15 environment of a coal gangue dump reclamation area through a temperature gradient experiment. We 16 collected leaf spectrum and water content data on alfalfa plants commonly planted in such areas. We then 17 obtained the optimal spectral features of appropriate leaf water content indicators through time series 18 analysis, correlation analysis, and least absolute shrinkage operator (lasso) regression analysis. A spectral 19 feature-based long short-term memory (SF-LSTM) model is proposed to estimate alfalfa's heat stress 20 level. Results: Comparing three leaf water content indicators, we found that the live fuel moisture content 21 (LFMC) varies significantly with time and has high regularity. Correlation analysis of the raw spectrum, first-derivative spectrum, vegetation indexes and leaf water content data shows that LFMC and spectral 23 data were the most strongly correlated. Combined with lasso regression analysis, the optimal spectral 24 features were the first-derivative spectral value at 1661 nm (abbreviated as FDS (1661)), RVI (1525, 25 1771), DVI (1412, 740) and NDVI (1447,1803). When the classification strategies were divided into 26 three categories and the time sequence length of the spectral features was set to five consecutive 27 monitoring dates, the SF-LSTM model had the highest accuracy in estimating the heat stress level in 28 alfalfa. The accuracy of the training set was > 95% and the accuracy of the verification set was about 29 90%. Conclusion: The results provide an important theoretical basis and technical support for vegetation 30 heat-stress assessment in coal gangue dump reclamation areas.

first-derivative spectrum, vegetation indexes and leaf water content data shows that LFMC and spectral 23 data were the most strongly correlated. Combined with lasso regression analysis, the optimal spectral 24 features were the first-derivative spectral value at 1661 nm (abbreviated as FDS (1661)

53
Heat accumulation inside gangue dumps increases the surface soil temperature, which can reduce 54 root numbers, roots' absorption of water and nutrients, and plant fresh weights [16]. High soil 55 temperature is far more influential than high air-temperature on plant growth [

98
Seeds were sown on September 10, 2020, at a sowing density of 10 holes per pot and two seeds per 99 hole. Ten seedlings per pot were grown to the three-leaf stage and harvested on November 15, 2020. The 100 inner diameter of the bottom of the barrel was 20 cm, the inner diameter of the mouth was 28 cm, the 101 height of the pots was 31.5 cm, and the empty barrel weighed 0.54 kg. Each barrel was loaded with 10 102 kg air-dried light loam and 5.28 g compound fertilizer with an N-P-K ratio of 15%-15%-15%. One kg of 103 soil was used to cover the seeds after sowing. The first alfalfa crop took about 60 days to grow from the 104 sowing to the flowering stage. The gradient experiment of heat stress was started on October 16, 2020.

105
One control group and five experimental groups were set. For the experimental groups, five heat sources 106 of different temperature T (T1 = 60 ℃, T2 = 90 ℃, T3 = 120 ℃, T4 = 150 ℃, and T5 = 180 ℃) were 107 placed at a depth of 30 cm in the soil layer, which is the typical thickness of overlying soil used in 108 reclamation projects [43] ( Fig. 1(b)). Each group was replicated five times, as shown in Fig. 1

115
A portable ground object spectrometer (Spectra Vista Corporation SVC HR-1024I) was used to measure 116 the spectral reflectance of alfalfa leaves. The spectral measurement range was 340-2500 nm and the 117 spectral sampling intervals were 1.5 nm (sampling range 350-1000 nm), 3.8 nm (sampling range 1000-118 1885 nm), and 2.5 nm (sampling range 1885-2500 nm). The resample interval was 1 nm. The 119 measurements were synchronized with the heating. The first measurement was made on October 16, 120 2020, and then every 4 days. The spectral reflectance of leaves was measured between 10:00 and 14:00 121 on sunny and windless days. The spectral data were collected eight times until November 15, 2020, when 122 it was overcast and rainy. A standard whiteboard was used for calibration of measurements using a hand-123 held leaf spectrum detector with a light source. This was clamped to the middle part of a leaf sample to 124 measure its spectrum. Each process measured three pots and each pot was measured six times, with the 7 average taken as the processed alfalfa leaf spectrum reflectance. During the measurement process, 126 standard whiteboard calibration was performed every 30 minutes.

127
Leaf water content 128 Leaf water content data were collected synchronously with spectral data. Three alfalfa samples were 129 selected for each treatment and packed in self-sealing plastic bags to avoid water loss from the plants as 130 much as possible. Samples were quickly brought back to the laboratory to weigh their fresh weight ( f m ) 131 with a precision balance and manually measure their leaf area. Each treated fresh leaf was put into a 132 beaker filled with distilled water and soaked for 24 hours. After reaching a constant weight, the saturated 133 fresh weight was measured ( t m ). Then a blade put into the paper bag, which was placed in an oven at Where t m is the measured saturation weight of the leaves.

159
In the formula, i  is the wavelength, () i  and '( ) i  are the reflectance and first-derivative 161 spectrum of the wavelength i  , respectively, and   ∆ is the interval between the wavelength 162

181
Where N is the sample number, j y is the predicted true value, i x is the observed value, 0  is 194 the bias,  is the weight of the observed variable, and  is a non-negative regularization parameter. 195 where  is the logistic sigmoid function, W is the weight matrix,  is a dot product, and b is 235 a bias term.  Table 2 show that these classical vegetation indices are not quite adequate for application in this paper, and a vegetation index with better correlation needs to be constructed.

20
The regular term constructed in the Lasso regression model makes it possible to compress the 332 dimension of the input sample. First, we need to determine the optimal regular coefficient Lambda (λ) 333 and adopt 10-fold cross-validation for the dataset (Figure 7(a)). As shown in Fig. 7(a), the minimum λ 334 of the RMSE was obtained after multiple iterations and was used as the regular term coefficient of the 335 model. Then, the compressed spectral characteristic parameters were determined and the accuracy of 336 the regression model was tested. The results are shown in Table 3 and Figure 7(b). It can be seen from 337

374
When divided into six categories, the model cannot converge many times and its stability is very poor.

375
The heat-stress level was divided into three categories. The model training set had the highest accuracy

501
The processed data required to reproduce these findings cannot be shared at this time as the data also 502 forms part of an ongoing study.        Coe cients of correlation between EWT, RWC, LFMC and the raw leaf and rst-derivative spectral data Figure 6 Coe cients of correlation between EWT, RWC, and LFMC with RVI (λ1, λ2), NDVI (λ1, λ2), DVI (λ1, λ2), and ratio/normalized difference/difference vegetation indexes constructed from raw spectral data