Significant Remote Sensing Vegetation Indices: A Review of Developments and Applications

Vegetation Indices (VIs) obtained from remote sensing based canopies are quite simple and effective algorithms for quantitative and qualitative evaluations of vegetation cover, vigor, and growth dynamics, among other applications. These indices have been widely implemented within RS applications using different airborne and satellite platforms with recent advances using Unmanned Aerial Vehicles (UAV). Up to date, there is no unified mathematical expression that defines all VIs due to the complexity of different light spectra combinations, instrumentation, platforms, and resolutions used. Therefore, customized algorithms have been developed and tested against a variety of applications according to specific mathematical expressions that combine visible light radiation, mainly green spectra region, from vegetation, and nonvisible spectra to obtain proxy quantifications of the vegetation surface. In the real-world applications, optimization VIs are usually tailored to the specific application requirements coupled with appropriate validation tools and methodologies in the ground. The present study introduces the spectral characteristics of vegetation and summarizes the development of VIs and the advantages and disadvantages from different indices developed. This paper reviews morethan100VIs,discussingtheirspecificapplicabilityandrepresentativenessaccordingtothevegetationofinterest,environment,andimplementationprecision.Predictably,research,anddevelopmentofVIs,whicharebasedonhyperspectralandUAVplatforms,wouldhaveawideapplicabilityindifferentareas.


Introduction
Remote sensed information of growth, vigor, and their dynamics from terrestrial vegetation can provide extremely useful insights for applications in environmental monitoring, biodiversity conservation, agriculture, forestry, urban green infrastructures, and other related elds. Speci cally, these types of information applied to agriculture provide not only an objective basis (depending on resolution) for the macroand micromanagement of agricultural production but also in many occasions the necessary information for yield estimation of crops [ ]. ese latter applications have been developed to be a well-known discipline category, precision agriculture, which could be tracked back to three decades ago [ ]. However, the applicability of remote sensing and its di erent VIs extracted from these techniques usually relies heavily on the instruments and platforms to determine which solution is best to get a particular issue.
. . Remote Sensing Platform Considerations. In terms of platforms, the advantages of satellite based remote sensing include high spatial resolution, which makes possible the extraction of long time data series of consistent and comparable data, which can be cost e ective [ ]. Furthermore, some satellite platforms have free access to visible and multispectral data, such as Landsat -. However, there are two main problems with these platforms for precision agriculture applications, which are related to the per pixel resolution ( m 2 per pixel for Landsat and m 2 for MODIS) and the orbit period ( d for Landsat and d for SPOT). More recently, pixel resolution has been increased by newer satellites, such as WorldView-and -(DigitalGlobe, Longmont, CO, USA). WorldView-was the rst commercial high resolution satellite to provide eight spectral sensors in the visible to near infrared range. Along with the four typical multispectral bands: blue ( -nm), green ( -nm), red (nm), and near infrared (NIR) ( -nm), each sensor Journal of Sensors isnarrowlyfocusedonaparticularrangeoftheelectromagnetic spectrum that is sensitive to a particular feature from the ground or a property of the atmosphere. However, images fromthisplatformcanbecostprohibitiveforlongtimedata series studies. e second problem with satellite based remote sensing is the revisitation time, which is days in average, which makes the agricultural applications di cult, speci cally those related to water and nutrient management. Moreover, passive sensors cannot penetrate clouds; therefore, there is no usable data capture for overcast days.
To solve these two main problems, airborne and more recently UAV platforms can be used. e former can also be cost prohibitive due to the requirement of expensive aircra s andpilots. elatterhasbecomealmostofubiquitoususein the last ve years with a ordable aircra s and camera payloads ranging from visible, near and thermal infrared, and D LIDAR, which has been referred to as Unmanned Aerial System (UAS). Among UAS platforms, there are mainly xed wing and multirotor options available. ere is a compromise using these UAS platforms in relation to payload weight versus ying time. In general, longer ying time achieved by xed wing systems demands lighter weight payloads. For example, small high de nition visible cameras weighting less t h a n g r a m sa sp a y l o a do fa x e dw i n gU A Sa l l o wi tt o y for around two hours using currently available batterypower [ ]. On the contrary, battery-powered multirotor UAS with higher payload capacity have reduced y time that at the moment is around -to -minute duration. Using these UAS, higher spatial and temporal data resolution can be achieved, which makes possible precision agriculture applications to the submeter resolution per pixel.
is allows research and practical applications applied to growth and vigor dynamic assessment, plant water status sensing for irrigation scheduling applications, and evapotranspiration modelling, among others [ -].
. . Remote Sensing and Vegetation Indices. Remote sensing of vegetation is mainly performed by obtaining the electromagnetic wave re ectance information from canopies using passive sensors. It is well known that the re ectance of light spectra from plants changes with plant type, water content within tissues, and other intrinsic factors [ ]. e re ectance from vegetation to the electromagnetic spectrum (spectral re ectance or emission characteristics of vegetation) is determined by chemical and morphological characteristics of the surface of organs or leaves [ ]. e main applications for remote sensing of vegetation are based on the following light spectra: (i) the ultraviolet region (UV), which goes from to nm; (ii) the visible spectra, which are composed of the blue ( -nm), green ( − nm), and red ( -nm) wavelength regions; and (iii) the near and mid infrared band ( -nm) [ , ]. e emissivity rate of the surface of leaves (equivalent to the absorptivity in the thermal waveband) of a fully grown green plant without any biotic or abiotic stress is generally in the range of . -. andismoreo enbetween . and . [ ].Onthecontrary, for dry plants, the emissivity rate generally has a larger range going from . to . [ ]. Vegetation emissivity in the near and mid infrared regions has been widely studied within plant canopies. Indices extracted from this light spectra range can be attributed to a range of characteristics beyond growth and vigor quanti cation of plants related to water content, pigments, sugar and carbohydrate content, protein content, and aromatics, among others [ , ]. Di erent applications are dependent on the re ectivity peaks or overtones for speci c compounds within the visible and near/mid infrared regions of light spectra [ , ]. Plant re ectivity in the thermal infrared spectral range ( -m) follows the blackbody radiation law [ ], which allows interpreting plant emission as directly r e l a t e dt op l a n tt e m p e r a t u r e .H e n c e ,i n d i c e so b t a i n e df r o m t h i ss p e c t r ar a n g ec a nb eu s e da sap r o xyt oa s s e s ss t o m a t a dynamics that regulates transpiration rate of plants. erefore,thelaterindicescanbeusedasindicatorofplantwater status [ -] and abiotic/biotic stress levels [ , ]. e latter considerations demonstrate that the quantitative interpretation of remote sensing information from vegetation is a complex task. Many studies have limited this interpretation by extracting vegetation information using individual light spectra bands or a group of single bands for data analysis. us, researchers o en combine the data from near infrared ( . -. m) and red ( . -. m) bands in di erent ways according to their speci c objectives [ ]. ese types of combinations present many disadvantages (e.g., lack of sensitivity) by using single or limited group of bands to detect, for example, vegetation biomass. ese limitations are particularly evident when trying to apply these types of VI on heterogeneous canopies, such as horticultural tree plantations. A mixed combination of soils, weeds, cover crops in the interrow, and the plants of interest makes the discrimination regions of interest and extraction of simple VI very di cult, speci cally, when the vegetation of interest has di erent VIs due to spatial variability, or VIs corresponding to other vegetation (weeds and cover crop), which can be similar to those of interest. e later will complicate imaging denoising and ltering processes. Several image analysis techniques and algorithms have been developed to go around these issues, which will be described later. Even though there are many considerations as described before, the construction of simple VI algorithm could many times render simple and e ective tools to measure vegetation status on the surface of the Earth [ ].

Vegetation Indices and Validation Process
With the use of high resolution spectral instrumentation, the number of bands obtained by remote sensing is increasing, and the bandwidth is getting narrower [ ]. One of the most used and implemented indices calculated from multispectral information as normalized ratio between the red and near infrared bands is the Normalized Di erence Vegetation Index (NDVI) [ ]. A direct use of NDVI is to characterize canopy growth or vigor; hence, many studies have compared it with the Leaf Area Index (LAI) [ ], where LAI is de ned as the area of single sided leaves per area of soil [ ].
Vegetation information from remote sensed images is mainly interpreted by di erences and changes of the green leaves from plants and canopy spectral characteristics. e Journal of Sensors most common validation process is through direct or indirect correlations between VIs obtained and the vegetation characteristics of interest measured in situ, such as vegetation cover, LAI, biomass, growth, and vigor assessment. More established methods are used to assess VIs using direct and georeferenced methods by monitoring sentinel plants to be com-paredwithVIsobtainedfromthesameplantsforcalibration purposes.
e later process is known as allometric measurements and requires destructive methods to scan speci c area of total leaves per plant or tree in the case of LAI [ ]. Indirect validation methods are based on proximal instrumentation using the same or similar spectral instrumentation to assess georeferenced sentinel plants at the same angle as the aerial platforms. e latter method is useful to compare VIs obtained from satellite that are sensitive to atmospheric e ects and serve as a mean to obtain correction factors. More recent indirect methods based on cover photography to estimate canopy cover, LAI, porosity, and clumping index have used automated analysis methods.
For this purpose, upward looking cover photogrammetry at zero zenith angle is taken with visible cameras to obtain canopy architecture parameters calculated using computer vision algorithms. An automated image acquisition and calculation method was proposed by Fuentes et al. applied to Eucalyptus trees [ ] and it has been successfully applied f o ro t h e rc r o p ss u c ha sg r a p e v i n e sc o m p a r e dt oa l l o m e t r i c measurements and to validate NDVI calculated from satellite information (WorldView-) [ ], apple trees with increased accuracy by using a variable light extinction coe cient ( ) [ ], and cherry trees improving the method by extracting n o n l e a fm a t e r i a ls u c ha sb r a n c h e sf o rt a l lt r e e s [ , ] .I n late , a computer application (App) for smartphones and tablet PCs called VitiCanopy was released for free use to assess canopy architecture parameters using the cover photography automated algorithms, which can be applied to any other tree crop by changing to a speci c value [ , ].Other Apps using RGB photogrammetry to assess LAI have been later developed such as PoketLAI [ -].
. . Basic Vegetation Indices. Jordan [ ] proposed in one of the rst VIs named Ratio Vegetation Index (RVI), which is based on the principle that leaves absorb relatively more red than infrared light; RVI can be expressed mathematically as ( ) e DVI is very sensitive to changes in soil background; it can be applied to monitoring the vegetation ecological environment. us,DVIisalsocalledEnvironmentalV egetation Index (EVI).
e Perpendicular Vegetation Index (PVI) [ ] is a simulation of the Green Vegetation Index (GVI) in , NIR D data. In the NIR − coordinate system, the spectral response from soil is presented as a slash (soil brighten line). e latter e ect can be explained as the soil presents a high spectral response in the NIR and bands. e distance between the pointofre ectivity( , NIR) and the soil line has been de ned as the Perpendicular VI, which can be expressed as follows: where soil is the soil re ectance; veg is the vegetation re ectivity; PVI characterizes the vegetation biomass in red the soil background; the greater the distance, the greater the biomass. PVIcanalsobequantitativelyexpressedas where DN NIR and DN are the radiation re ected luminance values from the NIR and ,r e s p e c t i v e l y ; is the intercept of the soil baseline and the vertical axis of NIR re ectivity; and is the angle between the horizontal axis of re ectivity a n ds o i lb a s e l i n e .P V I l t e r so u ti nt h i sw a yt h ee e c t so f soil background in an e cient manner; PVI also has less sensitivity to atmospheric e ects and it is mainly used for the inversion of surface vegetation parameter (grass yield, chlorophyll content), the calculation of LAI, and vegetation identi cation and classi cation [ , ]. However, PVI is sensitive to soil brightness and re ectivity, especially in the case of low vegetation coverage and needs to be adjusted for this e ect [ ].
As mentioned before, the Normalized Di erence Vegetation Index (NDVI) is the most widely used as VI; it was proposed by Rouse Jr. et al. [ ], which can be expressed as Since the index is calculated through a normalization procedure,therangeofNDVIvaluesisbetween and ,havinga sensitive response to green vegetation even for low vegetation covered areas. is index is o en used in research related to regional and global vegetation assessments and was shown to be related not only to canopy structure and LAI but also to canopy photosynthesis [ , ]. However, NDVI is sensitive to the e ects of soil brightness, soil color, atmosphere, cloud and cloud shadow, and leaf canopy shadow and requires remote sensing calibration.

Journal of Sensors
. . Vegetation Indices considering Atmospheric E ects. Given the limitations of NDVI under atmospheric e ects, Kaufman and Tanré[ ]proposedtheAtmosphericallyResistantV egetation Index (ARVI). is index is based on the knowledge that the atmosphere a ects signi cantly compared to the NIR. us, Kaufman and Tanrémodi edtheradiationvalue of by the di erence between the blue ( )and . erefore, ARVI can e ectively reduce the dependence of this VI to atmospheric e ects, which can be expressed as is the di erence between and , is the re ectivity related to the molecular scattering and gaseous absorption for ozone corrections, and represents the air conditioning parameters.
e ARVI is commonly used to eliminate the e ects of atmospheric aerosols. e aerosols and ozone e ects in the atmosphere still need to be eliminated by the S atmospheric t r a n s po rtm od e l[ ] .H o w ev e r ,t oi m p l e m e n tt h e Sa t m ospheric transmission model, actual atmospheric parameters must be considered, which are di cult to obtain. If the ARVI index is not calculated using the S model, this index is not expected to perform much better than NDVI considering atmospheric e ects or large dust particles in the atmosphere.
us, Zhang et al. [ ] proposed a new atmospheric e ect resistant vegetation index, namely, IAVI, that can eliminate atmospheric interference without the use of the S model.
w h e r et h er a n g eo f values is between . and . ; a signi cant value of is close to for ARVI. A er testing, the error caused in IAVI by the atmosphere e ect is between . % and . %, which is less than those found when using NDVI inthesameconditions( -%).
. . Adjusted-Soil Vegetation Index. e distinction of vegetation from the soil background was originally proposed by Richardson and Wiegand [ ] by analyzing the soil line, which can be considered as a linear relationship on the D plane of the soil spectral re ectance values between the NIR and . erefore, it can be considered as a comprehensive description of a large number of soil spectral information f r o man u m b e ro fe n v i r o n m e n t s[ ] .M a n yV I st h a tt a k e into account the e ect of soil background have been based on this principle. In addition to PVI (( )-( )), the Soil Line Atmospheric Resistance Index (SLRA) was developed based ontheimprovementofthesoillineprinciple. eSLRAwas then combined with the Transformed Soil-Adjusted Vegetation Index (TSAVI) to develop the Type Soil Atmospheric Impedance Vegetation Index (TSARVI) [ ], which will be discussed later.
As shown before, NDVI ( ) is very sensitive to background factors, such as the brightness and shade of the vegetation canopies and soil background brightness. Researches have shown that when the background brightness is increased, NDVI also increases systematically. Given the e ect of soil background, radiation increases signi cantly when the vegetation cover is sparse; conversely NIR radiation is reduced to make the relationship between vegetation and soil more sensible. Many VIs have been developed to adjust to the soil e ect.
Since NDVI and PVI have some de ciencies in describing the spectral behavior of vegetation and soil background, Huete [ ] established the Soil-Adjusted Vegetation Index (SAVI), which can be expressed as follows: ( ) e above model of a soil vegetation system was established to improve the sensitivity of NDVI to soil backgrounds, where is the soil conditioning index, which improves the sensitivity of NDVI to soil background. e range of is from to . In practical applications, the values of are determined according to the speci c environmental conditions. When the degree of vegetation coverage is high, is close to , showing that the soil background has no e ect on the extraction of vegetation information. is kind of ideal conditions is rarely found in natural environments and can be applicable only in thecaseofalargecanopydensityandcoverage[ ]. evalue of is around . under most common environmental conditions. When iscloseto ,thevalueofSAVIisequaltoNDVI. However, factor should vary inversely with the amount of vegetation present to obtain the optimal adjustment for the soil e ect. us, a modi ed SAVI (MSAVI) replaces factor inth eSA VIeq ua tio n( )wi thava ria b le function. In this way, MSAVI [ ] reduces the in uence of bare soil on SAVI, which can be expressed as follows: ( ) e SAVI is much less sensitive than the RVI ( ) to changes in the background caused by soil color or surface soil moisture content. ree new versions of SAVI (SAVI , SA VI ,andSA VI )weredevelopedbasedonthetheoretical considerations of the e ects of wet and dry soils [ ]. SAVI , SAVI , and SAVI reduce the e ect of background soil brightness, by taking into account the e ect of the variation of the solar incidence angle and changes in the soil physical structure.
Based on the implementation of the MSAVI, Richardson and Wiegand ( ) proposed a Modi ed Secondary Soil-Adjusted Vegetation Index (MSAVI ) [ ], which can be expressed as ( )

Journal of Sensors
MSAVI does not rely on the soil line principle and has a simpler algorithm; it is mainly used in the analysis of plant growth, deserti cation research, grassland yield estimation, LAI assessment, analysis of soil organic matter, drought monitoring, and the analysis of soil erosion [ ].
Baretetal.studiedthesensitivityof veVIs(NDVI,SAVI, Transformed Soil-Adjusted Vegetation Index (TSAVI), Modi ed Soil-Adjusted Vegetation Index (MSAVI), and Global Environment Monitoring Index (GEMI)) to the soil background. ey simulated the performance of the VIs for di erent soil textures, moisture levels, and roughness by using the Scattering from Arbitrarily Inclined Leaves (SAIL) model. ey determined an optimum value of SAIL = . to reduce the e ects of soil background and then proposed an Optimized Soil-Adjusted Vegetation Index (OSAVI) [ ] that can be expressed as follows: where SAIL is . and OSAVI does not depend on the soil line and can eliminate the in uence of the soil background e ectively. However, the applications of OSAVI are not extensive; it is mainly used for the calculation of aboveground biomass, leaf nitrogen content, and chlorophyll content, among others [ ].
. . Tasseled Cap Transformation of Greenness Vegetation Index (GVI, YVI, and SBI). Kauth and omas studied the spectral pattern of the vegetation growth process and called it the "spike cap" pattern, including the soil background re ectivity and brightness line. e Tasseled Cap Transformation is a conversion of the original bands of an image into a new set of bands with de ned interpretations that are useful for vegetation mapping. A Tasseled Cap Transformation is performed by taking "linear combinations" of the original image bands, which is similar in concept to the multivariate data analysis technique called principal components analysis (PCA) [ ].
e Tasseled Cap can convert Landsat MSS, Landsat TM, and Landsat ETM data. For Landsat MSS data, furthermore, the Tasseled Cap performs orthogonal transformation on the original data, which converts it into a D space. is conversion includes the Soil Brightness Index (SBI) ( ), degree of Green Vegetation Index (GVI) ( ), and the degree of Yellow Vegetation Index (YVI) ( ). It also includes Nonsuch Index (NSI) mainly for noise reduction. e NSI is closely related to atmospheric e ects. For the Landsat TM data, the Tasseled Cap results consist of three factors: brightness, greenness, and a third component related to soil. Among them,thebrightnessandthegreennessareequivalenttoSBI and GVI in the MSS Tasseled Cap. e third component is r e l a t e dt os o i lc h a r a c t e r i s t i c sa n dh u m i d i t y .F o rL a n d s a t ETM data, the Tasseled Cap Transformation generates six bands, namely, brightness, greenness, humidity, the fourth component (noise), a h component, and a sixth component.
( ) e GVI, YVI, and SBI ignore the interaction and e ects of theatmosphere,soil,andvegetation.SBIandGVIcanbeused to evaluate the behavior of vegetation and bare soil [ ]. e GVI has a strong correlation with di erent vegetation covers.
us, GVI increases the processing of atmospheric e ects.
( ) M i s r aa n dW h e e l e r( )p e r f o r m e dP C Ao fL a n d s a t images and computed the multiple factors of these indexes.
is analysis was the basis of the development of the Misra Soil Brightness Index (MSBI), Misra Green Degree Vegetation Index (MGVI), and Misra Yellow Degree Vegetation Index (MYVI) [ ], which can be expressed as follows: ( ) S i n c eN D V Ih a sb e e nf o u n dt ob ea e c t e do n l yb ys o i l brightness, it presents a negative correlation between NDVI and soil brightness. A positive correlation is found when only atmospheric e ects a ect NDVI. Under natural conditions, the soil and atmosphere in uence NDVI in a complex manner, which interacts with the vegetation cover in uence. erefore, atmosphere and vegetation have a collective e ect on NDVI based on the soil characteristics and exposure. Liu andHuetecomprehensivelyanalyzedmultiplesoiltypesand atmospheric enhanced VIs. ey developed the Atmosphere Antivegetation Index (ARVI) and Soil-Adjusted Vegetation Index (SAVI) for a comprehensive analysis of vegetation in these conditions. ey found that, as a result of the interaction between the soil and the atmosphere, reducing one of them may increase the other. ey introduced a feedback mechanism by building a parameter to simultaneously correct soil Journal of Sensors and atmospheric e ects. is parameter is the Enhanced Vegetation Index (EVI) [ ] that can be expressed as follows: which includes the values of NIR, R,a n dB,w h i c ha r e corrected by the atmosphere; L represents soil adjustment parameters, and its value is equal to ; and parameters correspond to constant values equivalent to and . , respectively.
. . Vegetation Indices Based on UAS Remote Sensing in the Visible Spectra Region. UAS remote sensing is a low altitude remote sensing technology ( -m), which is less a ected by atmospheric factors during the data acquisition process. It has the advantages of a ordability, simple operation, fast imaging speed, and high spatial and temporal resolutions, which is unparalleled compared with traditional [ ] remote sensing technologies based on satellites. At present, the UAS remote sensing technology plays a crucial role in the eld of aerial remote sensing with increased interest in applying these platforms on di erent studies of vegetation assessment [ ].
e practical applications of UAS are mainly related to images acquisition in the visible bands (RGB) due to easy access of ubiquitous high resolution cameras at low price and weight. However, due to rapid advances in technology, multispectral and infrared thermal cameras are becoming increasingly cheaper and miniaturized.
As previously shown through di erent VIs, most of them a r eb a s e do nt h em i x t u r eo fv i s i b l el i g h tb a n d sa n dN I Rt o generatealgorithmsandthosebasedonlyonthevisiblelight spectra are not common. However, weightless high de nition cameras are appearing in the market that includes the NIR band, which will enhance the practical applicability of UAS in the near future. ese types of re ectance are commonly measured using visible, multispectral, and hyperspectral cameras [ ]. According to Gago et al. ( ), NDVI is one of the most employed indices for UAS applications and is de ned speci cally as where 800 is the re ectance at nm and 680 at nm. Due to the high NIR re ectance of chlorophyll, this index is used to detect plants greenness [ ]. Some studies described the use of UAS with multispectral cameras and high resolution multispectral satellites to estimate LAI (Leaf Area Index) through NDVI [ , ]. e optimized index transformed chlorophyll absorption in re ectance Transformed Chlorophyll Absorption in Re ectance Index/Optimized Soil-Adjusted Vegetation Index (TCARI/OSAVI) was proposed as more sensitive VI to chlorophyll content. In this way, avoiding other factors that could a ect the re ectance values such as canopy re ectance and soil re ectance among others [ ]. Another index evaluated by Zarco-Tejada et al. ( ) was the PRInorm, which is an improvement of the Photochemical Re ectance Index (PRI).
is index considers xanthophyll changes related to water stress but also generates a normalization considering chlorophyll content and canopy leaf area reduction which is mainly a ected by water stress [ ]. However, by obtaining a quick and e ective method to extract vegetation information based on UAS visible images, it will enhance and popularize the scope of application of UAS immediately [ ]. In this sense, Wang et al. ( ) comprehensively considered the spectral characteristics of healthy green vegetation and the spectral characteristics of typical features of UAS imagery [ ]. ey use green ( ) band instead of the NIR band to calculate NDVI, ( red + blue ) compared to for NDVI, and the band multiplied by for ( red + blue ). us,aV isible-Band Di erence Vegetation Index (VDVI) is created based on the three bands of visible light, which can be expressed as follows:  ) found that LAI, chlorophyll, and the chlorophyll-LAI interaction accounted for , , and % of the MCARI variation. Even though the MCARI formula is not related to the NIR bands, good predictions were still found.
In agriculture, crop growth is directly linked to water supply and plant water status. When the soil water supply is insu cient, plants will be under water stress, which leads to reduced crop yield and even crop failures under extreme drought conditions. So it is very important to evaluate the crop water status in a timely and accurate manner, which has direct implications on crop growth, yield, and quality of produce [ ]. In recent years, the development of infrared thermal remote sensing technology made it possible to measure canopy temperature changes and dynamics from crop populations. ese changes are related to the transpiration rate of plants and stomatal conductance. where canopy is the temperature of fully sunlit canopy leaves ( ∘ C), nws isthetemperatureoffullysunlitcanopyleaves( ∘ C) when the crop is non-water-stressed (well-watered); dry is the temperature of fully sunlit canopy leaves ( ∘ C) when the crop is severely water stressed due to low soil water availability. nws and dry are the lower and upper baselines used to normalize CWSI for the e ects of environmental conditions (air temperature, relative humidity, solar radiation, and wind speed) on canopy . eC W S Ih a st w om o d e l s ,a ne m p i r i c a l model and a theoretical model; however, the theoretical model involves too many parameters, and these parameters are not easy to obtain. erefore, the theoretical model is only used for research purposes [ -]. e empirical model can be obtained only by using crop canopy temperature, air temperature, and air saturation di erence, so the empirical model has been further studied and used in many crop applications [ ].
Besides the use of infrared thermal radiation to detect plant water stress detection, the visible part of the spectrum has also been useful for early water stress detection. is involves using indices focused on bands at speci c wavelengths where photosynthetic pigments are a ected by water stress conditions such as chlorophyll.
e Photochemical Re ectance Index (PRI) has been used as a stress index of stress based on this principle, with initial developments to b ea p p l i e dt od i s e a s es y m p t o m sd e t e c t i o n ,w h i c hc a nb e expressed as PRI = 531 − 570 531 − 570 .
( ) It has been shown that the Light Use E ciency (LUE) is a key variable to estimate Net Primary Productivity (NPP) [ , ]. When obtaining reliable accuracy in LUE measurements, it is possible to study the distribution of energy and global climate change. e PRI is a normalized di erence VI of re ectivity at nm and nm and the re ectance of these two bands is a ected by the xanthophyll cycle and is closely related to LUE of leaves.
erefore, PRI provides a good estimation of leaf LUE.
. . Summary of Vegetation Indices. Asummaryofthemain VIs discussed in this paper can be found in Table with

Conclusions
Simple VIs combining visible and NIR bands have significantly improved the sensitivity of the detection of green vegetation. Di erent environments have their own variable and complex characteristics, which needs to be accounted when using di erent VIs. erefore, each VI has its speci c expression of green vegetation, its own suitability for speci c uses, and some limiting factors. erefore, for practical applications, the choice of a speci c VI needs to be made with caution by comprehensively considering and analyzing the advantages and limitations of existing VIs and then combine them to be applied in a speci c environment. In this way, the u sa g eo fV I sca nbeta il o r edt os pec i ca p p l i ca ti o n s ,i n s trumentation used, and platforms. With the development of hyperspectral and multispectral remote sensing technology, new VIs can be developed, which will broaden research areas. It is envisioned that these new developments will be readily applied and adopted by UAS platforms and will become one of the most important research areas in aerospace remote sensing in the near future.  [ ] J. Grace, C. Nichol, M. Disney, P. Lewis, T. Quaife, and P. Bowyer, "Can we measure terrestrial photosynthesis from space directly, using spectral re ectance and uorescence?" Global Change Biology,vol. ,no. ,pp. [ ]L .W a n g ,J .L i u ,L .Y a n g ,Z .C h e n ,X .W a n g ,a n dB .O u y a n g , "Applications of unmanned aerial vehicle images on agricultural remote sensing monitoring, " Nongye Gongcheng Xuebao/ Transactions of the Chinese Society of Agricultural Engineering, vol. , no. , pp. [ ]Z .Y a n g ,P .W i l l i s ,a n dR .M u e l l e r ," I m p a c to fb a n d -r a t i o enhanced awifs image on crop classi cation accuracy, " in