China currently faces a severe issue from soil heavy metal contamination in major grain-producing areas, which slightly alters the chlorophyll content of rice leaves. Although soil microbes are unable to break down heavy metals found in the soil, they may readily collect and transform into more hazardous methyl compounds. Some even make their way into human cells through the food chain and build up to potentially harmful levels there, endangering human health. Therefore, the creation of a universal spectrum index that can detect even minute variations in the amount of chlorophyll in leaves is crucial for tracking non-point source pollution in agricultural areas and estimating crop production. In this paper, two cities in the north and south were chosen to provide a matching test field with three distinct types of pollution: non-pollution, moderate pollution, and severe pollution., namely Changchun city and Suzhou city, and 260 groups of ASD spectral data and leaf chlorophyll content data were collected to construct leaf chlorophyll sensitive index. The universal applicability of the model can be proved by the difference of environment and crop varieties in the two regions, and the scientific supply of other conditions to avoid unnecessary stress and other environmental factors. First, three indexes, namely NDSI (normalized difference spectral index), RSI (ratio spectral index) and NVI (new vegetation index), were constructed, and each index was used as an independent variable to deduce the chlorophyll content of leaves. At the same time, 7 vegetation indexes (DattA, MaccioniA, Vogelmann2A, SR3A, SR6A, VogelmannA and RSI(550,800)) which are sensitive to chlorophyll content in leaves are selected as input variables to build a model that can accurately forecast the amount of chlorophyll in rice leaves with a 0.75 accuracy rate. The results showed that NVI(R720, R800, R698) was the best index to evaluate chlorophyll content in leaves. The application effect of this index in two regions with different varieties and environments is good, and the determination coefficient R2 is 0.83. This index (NVI) is obviously better than other indexes and random forest model. By analysing the spectral properties of rice leaves, it was found that the “red edge” had a slight “blue shift”, the height of green peak increased, the absorption depth decreased, and the absorption width decreased after the moderate and severe pollution of rice. It also explains why the best exponential bands are in the red-edged region. In this paper, a universal index sensitive to the weak change of chlorophyll content in leaves under arsenic stress was found. The index has good application effect in two regions with different varieties and different environments. It may be utilized to field non-point source pollution monitoring by estimating the little variations in chlorophyll concentration in rice leaves under stress from heavy metals in various locales.