# A Practical Split-Window Algorithm for Estimating Land Surface Temperature from Landsat 8 Data

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

**Figure 1.**Spectral response functions of the thermal bands of the different sensors on board the Landsat platforms.

## 2. Methodology

#### 2.1. Split-Window Algorithm Principle

_{i}(T

_{i}) measured on top of the atmosphere (TOA) in a thermal infrared channel of the sensor aboard the satellite is provided with a significant approximation as follows [16,18]:

_{i}is the effective transmittance of the atmosphere in channel i, and ε

_{i}is the channel effective surface emissivity. B

_{i}is the Planck function, and B

_{i}(T

_{s}) is the measured radiance if the surface is a black body with a surface temperature T

_{s}(K), ${R}_{atm\_i}^{\uparrow}$ and ${R}_{atm\_i}^{\downarrow}$ are the upward and downward atmospheric thermal radiance. The first term on the right side of Equation (1) represents the surface emission attenuated by the atmosphere. The second term represents the downward atmospheric thermal radiance reflected by the surface and which reaches the sensor. The third term represents the upward atmospheric emission towards the sensor. From Equation (1), we can deduce that LST retrieval requires knowledge of surface emissivity and atmospheric information.

_{i}, T

_{j}) of the adjacent thermal infrared channels, which can be expressed as

_{i}and T

_{j}are the TOA brightness temperatures measured in channels i (~11.0 μm) and j (~12.0 µm), respectively; ε is the average emissivity of the two channels (i.e., ε = 0.5 [ε

_{i}+ ε

_{j}]), whilst Δε is the channel emissivity difference (i.e., Δε = ε

_{i}− ε

_{j}); b

_{k}(k = 0,1,...7) are the algorithm coefficients derived in the following simulated dataset.

#### 2.2. Algorithm Development for Landsat 8

_{k}in Equation (2) are obtained through numerical simulation with different atmospheric and surface conditions.

^{2}to 8 g/cm

^{2}[20,21]. In one of the levels, the profiles with a relative humidity greater than 90% are discarded as under a cloudy condition, which results in a total of 946 atmospheric situations under clear skies with water vapor ranging from 0.06 g/cm

^{2}to 6.3 g/cm

^{2}. According to these profiles, the MODTRAN 5.2 atmospheric transmittance/radiance code is used to calculate the channel atmospheric parameters (${\tau}_{i}$, ${R}_{atm\_i}^{\uparrow}$ and ${R}_{atm\_i}^{\downarrow}$ ) in Equation (1) of each atmospheric profile with a spectral integration of the response filters of the two TIRS channels.

_{0}of each atmospheric profile, that is, the LST range from T

_{0}− 10 K to T

_{0}+ 20 K in a 5 K step. In addition, a total of 53 emissivity spectra in 3 µm to 14 µm are selected from the American Advanced Spaceborne Thermal Emission Reflection (ASTER) emissivity database [22], including 5 water types, 8 man-made target types, 4 vegetation types, 5 rock types, 30 soil types and 1 mineral type. Notably, by contrast to other studies [23,24], man-made targets are considered in this study because the LST products from TIRS are used in urban environmental studies, given the advantages of these products in terms of the finer resolution compared with the Moderate-resolution Imaging Spectroradiometer (MODIS) and Advanced Very High Resolution Radiometer (AVHRR) products, and also the free charge compared with the ASTER product with a resolution of 90 m.

_{i}and T

_{j}are obtained from the inverse of Planck’s law in the two channels. Given that the field of view (FOV) of the TIRS is about 15 degrees and almost observes the land surface at a nadir direction, meanwhile, the angular effect of atmospheric data, land surface emissivity, and LST as reported in our previous study is also not remarkable [25,26] in FOV = 15 degrees, it is reasonable to ignore the angular variation in the development of SW algorithm. Thus, a total of 350,966 different groups of T

_{i}and T

_{j}, LST, and ε and ∆ε are obtained (946 atmospheres × 53 emissivity × 7 surface temperatures). Thus, the coefficients b

_{0}–b

_{7}in Equation (2) can be determined through a statistical regression method.

## 3. Algorithm Results

#### 3.1. Algorithm Coefficients

_{0}–b

_{7}in Equation (2) independently on the CWV to improve the LST retrieval accuracy. Thus, in our SW algorithm, the CWV was divided into 5 sub-ranges and an overlap of 0.5 g/cm

^{2}was considered between 2 adjacent sub-ranges, which resulted in [0.0, 2.5], [2.0, 3.5], [3.0, 4.5], [4.0, 5.5] and [5.0, 6.3] g/cm

^{2}. The CWV was retrieved from a modified split-window covariance and variance ratio method, as stated in Section 3.3. However, given the somewhat unsuccessful CWV retrieval, a group of coefficients for the entire CWV range needed to be calculated to ensure the spatial continuity of the LST product. Table 1 displays the coefficients in different CWV sub-ranges and the root-mean-square error (RMSE) of the temperature that are estimated based on the simulation data. The table shows a significant variation in the coefficients with the CWV sub-ranges, particularly in the heavy CWV loading. The RMSE was smaller than the variation from 0.34 K to 0.93 K, which increased as the CWV increased.

**Table 1.**The coefficients b

_{k}(k = 0,1,…7) in different atmospheric column water vapor (CWV) sub-ranges and the root-mean-square error (RMSE) of the temperature estimated based on the simulation data.

CWV (g/cm^{2}) | b_{0} | b_{1} | b_{2} | b_{3} | b_{4} | b_{5} | b_{6} | b_{7} | RMSE |
---|---|---|---|---|---|---|---|---|---|

[0.0, 2.5] | −2.78009 | 1.01408 | 0.15833 | −0.34991 | 4.04487 | 3.55414 | −8.88394 | 0.09152 | 0.34 K |

[2.0, 3.5] | 11.00824 | 0.95995 | 0.17243 | −0.28852 | 7.11492 | 0.42684 | −6.62025 | −0.06381 | 0.60 K |

[3.0, 4.5] | 9.62610 | 0.96202 | 0.13834 | −0.17262 | 7.87883 | 5.17910 | −13.26611 | −0.07603 | 0.71 K |

[4.0, 5.5] | 0.61258 | 0.99124 | 0.10051 | −0.09664 | 7.85758 | 6.86626 | −15.00742 | −0.01185 | 0.86 K |

[5.0, 6.3] | −0.34808 | 0.98123 | 0.05599 | −0.03518 | 11.96444 | 9.06710 | −14.74085 | −0.20471 | 0.93 K |

[0.0, 6.3] | −0.41165 | 1.00522 | 0.14543 | −0.27297 | 4.06655 | −6.92512 | −18.27461 | 0.24468 | 0.87 K |

_{s}in the simulated dataset and the T

_{s}estimated through the SW algorithm using the coefficients (see Table 1) of the 5 CWV sub-ranges. In those figures, the temperature difference evidently fell in the ranges of [−1.0, 1.0] K, which covered about 97.92%, 94.42%, 91.67%, 86.77%, 75.99% and 92.11% of the total cases in Figure 2a–f, respectively. The maximum temperature error was about −3.09 K (see Figure 2e), which corresponds to the heavy CWV content. In the case of the entire water vapor range, the RMSE was about 0.87 K.

**Figure 2.**Histograms of temperature difference between the actual T

_{s}and the T

_{s}estimated using the split-window algorithm; (

**a**) for CWV ∈ [0.0, 2.5]; (

**b**) for CWV ∈ [2.0, 3.5]; (

**c**) for CWV ∈ [3.0, 4.5]; (

**d**) for CWV ∈ [4.0, 5.5]; (

**e**) for CWV ∈ [5.0, 6.3]; (

**f**) for CWV ∈ [0.0, 6.3] g/cm

^{2}. CWV: Column Water Vapor, RMSE: Root-Mean-Square Error, ΔLST: actual T

_{s}and the T

_{s}estimated using the split-window algorithm.

#### 3.2. Determination of LSEs

**Table 2.**Land cover schemes in FROM-GLC (Finer Resolution Observation and Monitoring of Global Land Cover) at levels 1 and 2 and their components.

Land Cover (Emissivity Class) | Models | Component Description | ||
---|---|---|---|---|

Level-1 Type | Level-2 Type | Vegetation Type | Ground Type | |

Cropland | Rice fields | Vol: bF = 0.0–0.2 | Gr. Veg | Mollisols, Liquid Water |

Greenhouse farming | Gr. Veg | Mollisols, Liquid Water | ||

Other croplands | Gr. Veg | Mollisols | ||

Forest | Broadleaf forests | Vol: bF = 0.7–5.0 | Bdlf | Alfisols, Spodosols |

Needleleaf forests | Ndle | Alfisols, Spodosols | ||

Mixed forests | Bdlf , Ndle | Alfisols, Spodosols | ||

Orchards | Bdlf | Alfisols, Spodosols | ||

Grasslands | Pastures | Vol: bF = 1.2–5.0 | Gr. Veg | Aridisols, Gr. Veg, Tree and Bush |

Other grasslands | Gr. Veg | Aridisols, Gr. Veg, Tree and Bush | ||

Shrublands | - | Vol: bF = 1.2–5.0 | Bdlf, Ndle | Aridisols, Gr. Veg, Tree and Bush |

Wetlands | Marshland | Spec: σ = 0.2 | - | Alfisols, Gr. Veg and Water |

Mudflats | - | Alfisols, Gr. Veg and Water | ||

Waterbodies | Lake | Spec: σ = 0.2 | - | Liquid Water |

Reservoir/Pond | - | Liquid Water | ||

River | - | Liquid Water | ||

Ocean | - | Liquid Water | ||

Tundra | Shrub and Brush Tundra | Vol: bF = 0.35–0.7 | Bdlf, Ndle | Aridisols, Gr. Veg, Tree and Bush |

Herbaceous Tundra | Gr. Veg | Aridisols, Gr. Veg, Tree and Bush | ||

Impervious | Impervious-high albedo | Vol: bF = 0.2–0.5 | Gr. Veg | Paving concrete |

Impervious-low albedo | Gr. Veg | Paving asphalt | ||

Barren Land | Dry salt flats | Vol: bF = 0.0–0.2 | Sn.Veg | Salty soil |

Sandy areas | Sn.Veg | Sand soil | ||

Bare exposed rock | Sn.Veg | Coarse sandstone | ||

Bare herbaceous croplands | Sn.Veg | Aridisols, Gr. Veg, Tree and Bush | ||

Dry lake/river bottom | Sn.Veg | Aridisols, Gr. Veg, Tree and Bush | ||

Other barren lands | Sn.Veg | Aridisols, Gr. Veg, Tree and Bush | ||

Snow and ice | Snow | Spec: σ = 0.2 | - | Snow |

Ice | - | Ice |

**Table 3.**Average emissivity for two TIRS (Thermal Infrared Sensor) channels at different land covers of FROM-GLC (Finer Resolution Observation and Monitoring of Global Land Cover)

Emissivity Class | Mean | |
---|---|---|

TIRS-10 | TIRS-11 | |

Cropland | 0.971 | 0.968 |

Forest | 0.995 | 0.996 |

Grasslands | 0.970 | 0.971 |

Shrublands | 0.969 | 0.970 |

Wetlands | 0.992 | 0.998 |

Waterbodies | 0.992 | 0.998 |

Tundra | 0.980 | 0.984 |

Impervious | 0.973 | 0.981 |

Barren Land | 0.969 | 0.978 |

Snow and ice | 0.992 | 0.998 |

#### 3.3. Determination of Atmospheric CWV

_{0}, c

_{1}and c

_{2}are the coefficients obtained from the simulated data; τ is the band effective atmospheric transmittance; N is the number of adjacent pixels (always excluding water and cloud pixels) in a spatial window size n (i.e., N = n × n); T

_{i,k}and T

_{j,k}are the respective brightness temperatures (K) of bands i and j at the TOA level for the kth pixel; and $\overline{{T}_{i}}$ and $\overline{{T}_{j}}$ are the mean or median brightness temperatures of the N pixels for the two bands. Using the aforementioned 946 cloud-free TIGR atmospheric profiles, we first used the new high accurate atmospheric radiative transfer model MODTRAN 5.2 to simulate the band effective atmospheric transmittance, and then we obtained the coefficients through regression, which resulted in c

_{0}= −9.674, c

_{1}= 0.653 and c

_{2}= 9.087. The model analysis indicated that this method will obtain a CWV RMSE of about 0.5 g/cm

^{2}. The details about the CWV retrieval can be found in [40].

## 4. Sensitivity Analysis

#### 4.1. Sensitivity Analysis to Instrument Noises

_{i}and T

_{j}in Equation (2). For CWV ∈ [0, 2.5] considering most parts of the 946 atmospheric profiles, we selected the coefficients in this CWV range (see Table 1) as examples to check the variation of the RMSEs affected by the given NEΔT. Thus, the LST error was 0.355 K for NEΔT = 0.1 K, 0.397 K for NEΔT = 0.2 K and 0.531 K for NEΔT = 0.4 K. Compared with the LST error (i.e., 0.34 K) for the no instrument noise, NEΔT = 0.1 K contributed only 4.5% of the error in the retrieved LST, whereas NEΔT = 0.2 K contributed about 16.8% and NEΔT = 0.4 K contributed up to 56.3%.

#### 4.2. Sensitivity Analysis to LSEs

**Table 4.**The values of α and β in Equations (4a) and (4b). α and β are coefficients of $\frac{1-\epsilon}{\epsilon}$ and $\frac{\u2206\epsilon}{{\epsilon}^{2}}$, respectively, mentioned in Equations (4a) and (4b).

CWV Sub-Ranges (g/cm^{2}) | [0.0, 2.5] | [2.0, 3.5] | [3.0, 4.5] | [4.0, 5.5] | [5.0, 6.3] | |||||
---|---|---|---|---|---|---|---|---|---|---|

Variables | α | β | α | β | α | β | α | β | α | β |

Range of Value (K) | [30.70, 61.38] | [−138.62, −66.76] | [45.77, 56.46] | [−112.25, −68.18] | [30.57, 60.88] | [−99.19, −28.91] | [20.94, 54.83] | [−81.41, −10.97] | [12.73, 46.30] | [−58.40, −4.40] |

Mean (K) | 42.29 | −93.57 | 51.20 | −90.02 | 46.03 | −64.04 | 38.20 | −47.09 | 30.53 | −33.11 |

Standard deviation (K) | 4.00 | 9.20 | 1.82 | 6.60 | 4.95 | 11.41 | 6.08 | 12.65 | 7.10 | 11.34 |

**Table 5.**LST (Land Surface Temperature) errors caused by 1% uncertainties in LSEs (Land Surface Emissivities) for different ranges of CWV (Column Water Vapor).

LST Error (K) | CWV (g/cm^{2}) | ||||
---|---|---|---|---|---|

[0.0, 2.5] | [2.0, 3.5] | [3.0, 4.5] | [4.0, 5.5] | [5.0, 6.3] | |

Range of Value | [0.73, 1.52] | [0.82,1.25] | [0.42,1.16] | [0.24,0.98] | [0.13,0.75] |

Mean | 1.02 | 1.04 | 0.79 | 0.61 | 0.45 |

Standard deviation | 0.10 | 0.06 | 0.12 | 0.13 | 0.13 |

^{2}. The error generally decreased as the CWV increased, and the minimum was obtained for the highest atmospheric condition consequently. In addition, the range of the LST error in the wet atmospheric condition was narrower than that in the dry atmospheric condition. This finding indicates that in wet atmospheric conditions, the dominant factor that affects the LST error is the uncertainty from the atmospheric information, rather than from that of the surface conditions. In this case, the LST retrieval was insensitive to the error included in the LSEs.

#### 4.3. Sensitivity Analysis to the Atmospheric CWV

^{2}theoretically. Thus, misclassifying CWV from the correct sub-range to the incorrect sub-range was possible, which can result in wrong coefficients for the SW algorithm. Table 6 lists the LST errors caused by using the wrong coefficients in the adjacent CWV sub-ranges. Given that the CWV error varies from −0.5 g/cm

^{2}to 0.5 g/cm

^{2}, we considered only the cases of misclassifying CWV in the adjacent CWV sub-ranges. Thus, for a true CWV value of 2.6 g/cm

^{2}, the retrieved water vapor value may be ranged in [0.0, 2.5], [2.0, 3.5] or [3.0, 4.5] g/cm

^{2}.

**Table 6.**LST (Land Surface Temperature) errors caused by using wrong coefficients in adjacent CWV (Column Water Vapor) sub-ranges.

CWV(g/cm^{2}) | [0.0, 2.5] | [2.0, 3.5] | [3.0, 4.5] | [4.0, 5.5] | [5.0, 6.3] |
---|---|---|---|---|---|

[0.0, 2.5] | 0.34 K | 1.45 K | - | - | - |

[2.0, 3.5] | 1.40 K | 0.60 K | 1.40 K | - | - |

[3.0, 4.5] | - | 1.34 K | 0.71 K | 1.21 K | - |

[4.0, 5.5] | - | - | 1.29 K | 0.86 K | 2.45 K |

[5.0, 6.3] | - | - | - | 1.85 K | 0.93 K |

_{inc}for convenience in the following discussion. The δLST

_{inc}was less than 1.5 K with a CWV of less than 4.0 g/cm

^{2}, but δLST

_{inc}increased dramatically with a CWV of more than 4.0 g/cm

^{2}. However, for a CWV in the sub-ranges of [2.5, 3.0], [3.5, 4.0], [4.5, 5.0] or [5.5, 6.0] g/cm

^{2}, the CWV retrieval value using the MSWCVR method may fall in the CWV adjacent sub-ranges, which consequently decreases the LST accuracy. For instance, a CWV is 4.8 g/cm

^{2}and belongs to a sub-range of [4.5, 5.0] g/cm

^{2}. If this CWV is misclassified into [3.5, 4.0], the δLST

_{inc}will be 1.29 K, whereas if this CWV is misclassified into [5.5, 6.0], the δLST

_{inc}will reach 2.45 K. To reduce the influence of the CWV error on the LST, for a CWV within the overlap of two adjacent CWV sub-ranges, we first use the coefficients from the two adjacent CWV sub-ranges to calculate the two initial temperatures and then use the average of the initial temperatures as the pixel LST. For example, the LST pixel with a CWV of 2.1 g/cm

^{2}is estimated by using the coefficients of [0.0, 2.5] and [2.0, 3.5]. This process initially reduces the δLST

_{inc}and improves the spatial continuity of the LST product.

#### 4.4. Comparison amongst Different Split-Window Algorithms

_{k}(k = 0, 1…6) and e

_{k}(k = 1, 2, 3, 4) are the algorithm coefficients; w is the CWV; ε and ∆a are the average emissivity and emissivity difference of two adjacent thermal channels, respectively, which are similar to Equation (2); and f

_{k}(k = 0 and 1) is related to the influence of the atmospheric transmittance and emissivity, i.e., f

_{k}= f(ε

_{i},ε

_{j},τ

_{i},τ

_{j}). Note that the algorithm (Equation (6a)) proposed by Jiménez-Muñoz et al. added CWV directly to estimate LST. Rozenstein et al. used CWV to estimate the atmospheric transmittance (τ

_{i}, τ

_{j}) and optimize retrieval accuracy explicitly. Therefore, if the atmospheric CWV is unknown or cannot be obtained successfully, neither of the two algorithms in Equations (6a) and (6b) will work. By contrast, although our algorithm also needs CWV to determine the coefficients, this algorithm still works for unknown CWVs because the coefficients are obtained regardless of the CWV, as shown in Table 1. We first obtained the coefficients of Equations (6a) and (6b) using the same simulated dataset in our algorithm development, and then we analyzed the difference between the SW algorithms. Table 7 presents the LST errors caused by the different algorithms for various CWV sub-ranges. From this table, we know that the errors of all the algorithms were close to one another for a CWV of less than 3.5 g/cm

^{2}, which is much less than 1.0 K. However, under wet atmospheric conditions, the LST error increased quickly as the CWV increased, especially for the algorithm of Rozenstein et al., Jiménez-Muñoz’s algorithm had similar results with our algorithm. Moreover, we added an uncertainty of ±0.5 g/cm

^{2}to the CWV, and we found that the LST error was about 0.8 K for the CWV of −of t tha

^{2}and 1.1 K for the CWV of +0.5 g/cm

^{2}for the algorithm of Jiménez-Muñoz et al.; the results were somewhat better than those of our algorithm, as shown in Table 7, probably because of the direct usage of CWV in the algorithm of Jiménez-Muñoz et al. (Equaiton (6a)) to reduce the influence of the CWV on the LST retrieval accuracy. As stated above, all three algorithms relied on the CWV input and were impractical without this parameter. To deal with this potential problem, our algorithm also obtained the coefficients in Equation (2) for the entire CWV range, and this group of coefficients only resulted in an LST error of about 0.87 K for all the CWV, about 0.46 K for the CWV sub-range of [0.0, 2.5] g/cm

^{2}and about 1.11 K for a CWV of less than 3.5 g/cm

^{2}. This CWV range contains most cases of atmospheric moisture in polar, mid-latitude and tropical profiles. Therefore, compared with the other two algorithms, our algorithm is more practical and can even obtain LST with high accuracy for cases even without known CWV information.

**Table 7.**LST (Land Surface Temperature) error for different split-window (SW) algorithms at different sub-ranges of CWV (Column Water Vapor). The second, third and fourth columns correspond to the RMSEs (Root-Mean-Square Errors) of Jiménez-Muñoz et al., Rozenstein et al., and our algorithm, while the last column is RMSE from our algorithm using the coefficients derived from all water vapor range [0, 6.3] g/cm

^{2}when no CWV information can be obtained.

CWV (g/cm^{2}) | Jiménez-Muñoz | Rozenstein | SW in This Paper | |
---|---|---|---|---|

[0.0, 2.5] | 0.46 K | 0.32 K | 0.34 K | 0.46 K |

[2.0, 3.5] | 0.51 K | 0.56 K | 0.60 K | 1.11 K |

[3.0, 4.5] | 0.71 K | 0.79 K | 0.71 K | 2.00 K |

[4.0, 5.5] | 0.87 K | 1.32 K | 0.86 K | 2.33 K |

[5.0, 6.3] | 0.93 K | 1.26 K | 0.93 K | 3.13 K |

[0.0, 6.3] | 0.72 K | 1.25 K | 0.87 K | 0.87 K |

## 5. Application of the Split-Window Algorithm

_{i}and T

_{j}) of the two adjacent bands of the TIRS, FROM-GLC land cover products and emissivity lookup table, which are a fraction of the FVC that can be estimated from the red and near-infrared reflectance of the Operational Land Imager (OLI). These parameters can be easily obtained, which makes our algorithm generally operational in practice. Figure 3 shows the main flowchart of the LST retrieval LST from the TIRS data, where the clouds in the images are eliminated by using the band quality files along with the OLI and TIRS data, which are consequently removed before the LST retrieval. The output can contain LST products and emissivity images in the two channels and in the CWV product.

**Figure 3.**The main flowchart of retrieving LST (Land Surface Temperature) from Landsat 8 image. OLI: Operational Land Imager; TIRS: Thermal Infrared Sensor; NDVI: Normalized Different Vegetation Index; FVC: Fraction of Vegetation Cover; MSWCVM: Modified Split-Window Covariance and Variance Ratio; CWV: Column Water Vapor.

**Figure 4.**The LST (Land Surface Temperature) retrieved from TIRS (Thermal Infrared Sensor) data at urban area of Beijing city and Gobi in northwest China. (

**a**) and (

**d**) are the false color images; (

**b**) and (

**e**) are the retrieved LST; while (

**c**) and (

**f**) present LST histograms.

## 6. Discussions

^{2}, 100 profiles for CWV ∈ [2.0, 3.5] g/cm

^{2}, 76 profiles for CWV ∈ [3.0, 4.5] g/cm

^{2}, 36 profiles for CWV ∈ [4.0, 5.5] g/cm

^{2}and 15 profiles for CWV ∈ [5.0, 6.3] g/cm

^{2}. The dry atmosphere takes great part in the used atmospheric profiles, but our algorithm developed several groups of coefficients depending on the CWV sub-ranges, so the current algorithm will not cause significant uncertainty to the final LST retrieval with known CWV, but it will lead some error to the LST result for the cases without input CWV. To reduce this effect, we are trying to find more representative profiles to optimize the current algorithm in the coming future work.

## 7. Conclusions

## Acknowledgments

## Author Contributions

## Conflicts of Interest

## References and Notes

- Price, J.C. The potential of remotely sensed thermal infrared data to infer surface soil-moisture and evaporation. Water Resour. Res.
**1980**, 16, 787–795. [Google Scholar] [CrossRef] - Duan, S.-B.; Li, Z.-L.; Tang, B.-H.; Wu, H.; Tang, R.L.; Bi, Y.; Zhou, G. Estimation of diurnal cycle of land surface temperature at high temporal and spatial resolution from clear-sky MODIS data. Remote Sens.
**2014**, 6, 3247–3262. [Google Scholar] [CrossRef] - Duan, S.-B.; Li, Z.-L.; Tang, B.-H.; Wu, H.; Tang, R.L. Generation of a time-consistent land surface temperature product from MODIS data. Remote Sens. Environ.
**2014**, 140, 339–349. [Google Scholar] [CrossRef] - Becker, F.; Li, Z.-L. Towards a local split window method over land surfaces. Int. J. Remote Sens.
**1990**, 11, 369–393. [Google Scholar] [CrossRef] - Becker, F.; Li, Z.-L. Surface temperature and emissivity at various scales: Definition, measurement and related problems. Remote Sens. Rev.
**1995**, 12, 225–253. [Google Scholar] [CrossRef] - Wan, Z.; Dozier, J. A generalized split-window algorithm for retrieving land-surface temperature from space. IEEE Trans. Geosci. Remote Sens.
**1996**, 34, 892–905. [Google Scholar] [CrossRef] - Gillespie, A.; Rokugawa, S.; Matsunaga, T.; Cothern, J.S.; Hook, S.; Kahle, A.B. A temperature and emissivity separation algorithm for advanced spaceborne thermal emission and reflection radiometer (ASTER) images. IEEE Trans. Geosci. Remote Sens.
**1998**, 36, 1113–1126. [Google Scholar] [CrossRef] - Becker, F.; Li, Z.-L. Temperature-independent spectral indexes in thermal infrared bands. Remote Sens. Environ.
**1990**, 32, 17–33. [Google Scholar] [CrossRef] - Watson, K. Two-temperature method for measuring emissivity. Remote Sens. Environ.
**1992**, 42, 117–121. [Google Scholar] [CrossRef] - Peres, L.F.; DaCamara, C.C. Land surface temperature and emissivity estimation based on the two-temperature method: Sensitivity analysis using simulated MSG/SEVIRI data. Remote Sens. Environ.
**2004**, 91, 377–389. [Google Scholar] [CrossRef] - Wan, Z.; Li, Z.-L. A physics-based algorithm for retrieving land-surface emissivity and temperature from EOS/MODIS data. IEEE Trans. Geosci. Remote Sens.
**1997**, 35, 980–996. [Google Scholar] [CrossRef] - Ingram, P.M.; Muse, A.H. Sensitivity of iterative spectrally smooth temperature/emissivity separation to algorithmic assumptions and measurement noise. IEEE Trans. Geosci. Remote Sens.
**2001**, 39, 2158–2167. [Google Scholar] [CrossRef] - Kealy, P.S.; Hook, S.J. Separating temperature and emissivity in thermal infrared multispectral scanner data - implications for recovering land-surface temperatures. IEEE Trans. Geosci. Remote Sens.
**1993**, 31, 1155–1164. [Google Scholar] [CrossRef] - Roy, D.P.; Wulder, M.A.; Loveland, T.R.; Woodcock, C.E.; Allen, R.G.; Anderson, M.C.; Helder, D.; Irons, J.R.; Johnson, D.M.; Kennedy, R.; et al. Landsat-8: Science and product vision for terrestrial global change research. Remote Sens. Environ.
**2014**, 145, 154–172. [Google Scholar] [CrossRef] - Li, Z.-L.; Tang, B.-H.; Wu, H.; Ren, H.; Yan, G.; Wan, Z.; Trigo, I.F.; Sobrino, J.A. Satellite-derived land surface temperature: Current status and perspectives. Remote Sens. Environ.
**2013**, 131, 14–37. [Google Scholar] [CrossRef] - Li, Z.-L.; Wu, H.; Wang, N.; Qiu, S.; Sobrino, J.A.; Wan, Z.; Tang, B.-H.; Yan, G. Land surface emissivity retrieval from satellite data. Int. J. Remote Sens.
**2013**, 34, 3084–3127. [Google Scholar] [CrossRef] - Gao, C.; Li, Z.-L.; Qiu, S.; Tang, B.; Wu, H.; Jiang, X. An improved algorithm for retrieving land surface emissivity and temperature from MSG-2/SEVIRI data. IEEE Trans. Geosci. Remote Sens.
**2014**, 52, 3175–3191. [Google Scholar] [CrossRef] - Li, Z.-L.; Petitcolin, F.; Zhang, R.H. A physically based algorithm for land surface emissivity retrieval from combined mid-infrared and thermal infrared data. Sci China Ser. E.
**2000**, 43, 23–33. [Google Scholar] [CrossRef] - Wan, Z. New refinements and validation of the collection-6 MODIS land-surface temperature/emissivity product. Remote Sens. Environ.
**2014**, 140, 36–45. [Google Scholar] [CrossRef] - Chedin, A.; Scott, N.A.; Wahiche, C.; Moulinier, P. The improved initialization inversion method - a high-resolution physical method for temperature retrievals from satellites of the Tiros-N Series. J. Climate Appl. Meteor.
**1985**, 24, 128–143. [Google Scholar] [CrossRef] - Chevallier, F.; Chedin, A.; Cheruy, F.; Morcrette, J.J. TIGR-like atmospheric-profile databases for accurate radiative-flux computation. Quart. J. Roy. Meteorol. Soc.
**2000**, 126, 777–785. [Google Scholar] [CrossRef] - Hulley, G.C.; Hook, S.J. The north American ASTER land surface emissivity database (NAASLED) version 2.0. Remote Sens. Environ.
**2009**, 113, 1967–1975. [Google Scholar] [CrossRef] - Tang, B.-H.; Bi, Y.; Li, Z.-L.; Xia, J. Generalized split-window algorithm for estimate of land surface temperature from Chinese geostationary Fengyun meteorological satellite (FY-2C) data. Sensors
**2008**, 8, 933–951. [Google Scholar] [CrossRef] - Snyder, W.C.; Wan, Z. BRDF models to predict spectral reflectance and emissivity in the thermal infrared. IEEE Trans. Geosci. Remote Sens.
**1998**, 36, 214–225. [Google Scholar] [CrossRef] - Ren, H.; Yan, G.; Chen, L.; Li, Z.-L. Angular effect of MODIS emissivity products and its application to the split-window algorithm. ISPSR J. Photogramm.
**2011**, 66, 498–507. [Google Scholar] [CrossRef] - Ren, H.; Liu, R.; Yan, G.; Mu, X.; Li, Z.-L.; Nerry, F.; Liu, Q. Angular normalization of land surface temperature and emissivity using multiangular middle and thermal infrared data. IEEE Trans. Geosci. Remote Sens.
**2014**, 52, 4913–4931. [Google Scholar] [CrossRef] - Snyder, W.C.; Wan, Z.; Zhang, Y.; Feng, Y.Z. Classification-based emissivity for land surface temperature measurement from space. Int. J. Remote Sens.
**1998**, 19, 2753–2774. [Google Scholar] [CrossRef] - Sun, D.; Pinker, R. Implementation of GEOS—based land surface temperature diurnal cycle to avhrr. Int. J. Remote Sens.
**2005**, 26, 3975–3984. [Google Scholar] [CrossRef] - Trigo, I.F.; Peres, L.F.; DaCamara, C.C.; Freitas, S.C. Thermal land surface emissivity retrieved from SEVIRI/METEOSAT. IEEE Trans. Geosci. Remote Sens.
**2008**, 46, 307–315. [Google Scholar] [CrossRef] - Gong, P.; Wang, J.; Yu, L.; Zhao, Y.; Zhao, Y.; Liang, L.; Niu, Z.; Huang, X.; Fu, H.; Liu, S.; et al. Finer resolution observation and monitoring of global land cover: First mapping results with Landsat TM and ETM+ data. Int. J. Remote Sens.
**2013**, 34, 2607–2654. [Google Scholar] [CrossRef] - Belward, A.S.; Estes, J.E.; Kline, K.D. The IGBP-DIS global 1-km land-cover data set discover: A project overview. Photogramm. Eng. Remote Sens.
**1999**, 65, 1013–1020. [Google Scholar] - Loveland, T.R.; Reed, B.C.; Brown, J.F.; Ohlen, D.O.; Zhu, Z.; Yang, L.; Merchant, J.W. Development of a global land cover characteristics database and IGBP discover from 1 km AVHRR data. Int. J. Remote Sens.
**2000**, 21, 1303–1330. [Google Scholar] [CrossRef] - Wan, Z.; Ng, D.; Dozier, J. Spectral emissivity measurements of land-surface materials and related radiative transfer simulations. Adv. Space Res.
**1994**, 14, 91–94. [Google Scholar] [CrossRef] - Carlson, T.N.; Ripley, D.A. On the relation between NDVI, fractional vegetation cover, and leaf area index. Remote Sens. Environ.
**1997**, 62, 241–252. [Google Scholar] [CrossRef] - Sobrino, J.A.; Raissouni, N.; Li, Z.-L. A comparative study of land surface emissivity retrieval from NOAA data. Remote Sens. Environ.
**2001**, 75, 256–266. [Google Scholar] [CrossRef] - Sobrino, J.A.; Raissouni, N.; Simarro, J.; Nerry, F.; Petitcolin, F. Atmospheric water vapor content over land surfaces derived from the AVHRR data: Application to the Iberian Peninsula. IEEE Trans. Geosci. Remote Sens.
**1999**, 37, 1425–1434. [Google Scholar] [CrossRef] - Li, Z.-L.; Jia, L.; Su, Z.; Wan, Z.; Zhang, R. A new approach for retrieving precipitable water from ATSR2 split-window channel data over land area. Int. J. Remote Sens.
**2003**, 24, 5095–5117. [Google Scholar] [CrossRef] - Schroedter-Homscheidt, M.; Drews, A.; Heise, S. Total water vapor column retrieval from MSG-SEVIRI split window measurements exploiting the daily cycle of land surface temperatures. Remote Sens. Environ.
**2008**, 112, 249–258. [Google Scholar] [CrossRef] - Wang, N.; Li, Z.-L.; Tang, B.-H.; Zeng, F.N.A.; Li, C.R. Retrieval of atmospheric and land surface parameters from satellite-based thermal infrared hyperspectral data using a neural network technique. Int. J. Remote Sens.
**2013**, 34, 3485–3502. [Google Scholar] [CrossRef] - Ren, H.; Du, C.; Qin, Q.; Liu, R.; Meng, J.; Li, J. Atmospheric water vapor retrieval from landsat 8 and its validation. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Quebec, QC, Canada, July 2014; pp. 3045–3048.
- Irons, J.R.; Dwyer, J.L.; Barsi, J.A. The next landsat satellite: The landsat data continuity mission. Remote Sens. Environ.
**2012**, 122, 11–21. [Google Scholar] [CrossRef] - Sobrino, J.A.; Li, Z.-L.; Stoll, M.P.; Becker, F. Improvements in the split-window technique for land-surface temperature determination. IEEE Trans. Geosci. Remote Sens.
**1994**, 32, 243–253. [Google Scholar] [CrossRef] - Xiong, X.X.; Barnes, W. An overview of MODIS radiometric calibration and characterization. Adv. Atmos Sci.
**2006**, 23, 69–79. [Google Scholar] [CrossRef] - Ren, H.; Du, C.; Liu, R.; Qin, Q.; Meng, J.; Li, Z.-L.; Yan, G. Evaluation of radiometric performance for the thermal infrared sensor onboard Landsat 8. Remote Sens.
**2014**, 6, 12776–12788. [Google Scholar] [CrossRef] - Jimenez-Munoz, J.C.; Sobrino, J.A.; Skokovic, D.; Mattar, C.; Cristobal, J. Land surface temperature retrieval methods from Landsat-8 thermal infrared sensor data. IEEE Geosci. Remote Sens. Lett.
**2014**, 11, 1840–1843. [Google Scholar] [CrossRef] - Sobrino, J.A.; Li, Z.-L.; Stoll, M.P.; Becker, F. Multi-channel and multi-angle algorithms for estimating sea and land surface temperature with ATSR data. Int. J. Remote Sens.
**1996**, 17, 2089–2114. [Google Scholar] [CrossRef] - Rozenstein, O.; Qin, Z.; Derimian, Y.; Karnieli, A. Derivation of land surface temperature for Landsat-8 TIRS using a split window algorithm. Sensors
**2014**, 14, 5768–5780. [Google Scholar] [CrossRef] [PubMed] - Montanaro, M.; Lunsford, A.; Tesfaye, Z.; Wenny, B.; Reuter, D. Radiometric calibration methodology of the Landsat 8 thermal infrared sensor. Remote Sens.
**2014**, 6, 8803–8821. [Google Scholar] [CrossRef] - Montanaro, M.; Gerace, A.; Lunsford, A.; Reuter, D. Stray light artifacts in imagery from the Landsat 8 thermal infrared sensor. Remote Sens.
**2014**, 6, 10435–10456. [Google Scholar] [CrossRef]

© 2015 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Du, C.; Ren, H.; Qin, Q.; Meng, J.; Zhao, S.
A Practical Split-Window Algorithm for Estimating Land Surface Temperature from Landsat 8 Data. *Remote Sens.* **2015**, *7*, 647-665.
https://doi.org/10.3390/rs70100647

**AMA Style**

Du C, Ren H, Qin Q, Meng J, Zhao S.
A Practical Split-Window Algorithm for Estimating Land Surface Temperature from Landsat 8 Data. *Remote Sensing*. 2015; 7(1):647-665.
https://doi.org/10.3390/rs70100647

**Chicago/Turabian Style**

Du, Chen, Huazhong Ren, Qiming Qin, Jinjie Meng, and Shaohua Zhao.
2015. "A Practical Split-Window Algorithm for Estimating Land Surface Temperature from Landsat 8 Data" *Remote Sensing* 7, no. 1: 647-665.
https://doi.org/10.3390/rs70100647