Paper Reading Note | Developing an Effective Model for Predicting Spatially and Temporally Continuous Stream Temperatures from Remotely Sensed Land Surface Temperatures

这篇文章用到了很多统计学的方法！

JDR是全局global模型

Abstract

Models were built at three spatial scales: site-specific(特定站点), subwatershed(分水岭), and basin-wide(流域). Model quality was assessed using jackknife and cross-validation.

eight-day composite LST

1.Introduction

2.Study Area

The Columbia River basin

3.Materials and Methods

3.1 Geospatial Data

10m DEM,

3.2. Land Surface Temperature Data and Handling Data Gaps

Daily Land Surface Temperature (LST) for 2000–2009 (MOD11A1 v005, 1 km2 spatial resolution rasters), and 8-day LST (MOD11A2 v005) for 2012 (also 1 km2), from the MODIS sensor were downloaded from NASA’s Earth Observing System Data and Information System

Gapfilling

3.3 In-Stream Thermal Logger Data

3.4. Model Development

4 th polynomial models ——> overfitting
cross-correlation function (CCF) (Time Series Analysis (TSA) package in R)

3.5. Sensitivity Analysis

• ackknife
• bootstrapped

4 Result

4.1. Estimating Missing Land Surface Temperature Data

We therefore concluded a simple temporal linear interpolation was most effective for filling in cloudy-day gaps in the LST for use in our models.

4.2 Predictor Variable Selection and Temporal Lag Analyses

We reviewed the residuals from linear regressions of potential predictor variables and stream temperatures to determine the most informative and least correlated physiographic variables to include in the LST model.

\[Temp = LST + JulianDay + Elevation \tag{1} \]

5 Discussion

We have demonstrated that remotely-sensed LST can be used in linear regression models to generate robust, spatially and temporally continuous estimate of stream temperature. The strongest relationship was between DMWT and LST, though the inclusion of Julian day did improve predictive power. Elevation was not included in the site-specific models, and it was not always significant in the HUC and global models.

---

[1] MCNYSET K, VOLK C, JORDAN C. Developing an Effective Model for Predicting Spatially and Temporally Continuous Stream Temperatures from Remotely Sensed Land Surface Temperatures [J]. Water, 2015, 7(12): 6827-46.