Abstract

AbstractLand surface air temperature is an essential climate variable for understanding rapid global environmental changes. Sparse network coverage prior to the 1950s is a significant source of uncertainty in global climate change evaluations. Recognizing the importance of spatial coverage, more stations are continually being added to global climate networks. A challenge is how to best use the information introduced by the new station observations to enhance our understanding and assessment of global climate states and changes, particularly for times prior to the mid‐20th century. In this study, Data INterpolating Empirical Orthogonal Functions (DINEOF) were used to reconstruct mean monthly air temperatures from the Global Historical Climatology Network‐monthly (GHCNm version 4) over the land surface from 1880 through 2017. The final reconstructed air temperature dataset covers about 95% of the global land surface area, improving the spatial coverage by ~80% during 1880–1900 and by 10%–20% since the 1950s. Validation tests show that the mean absolute error of the reconstructed data is less than 0.82°C. Comparison with the Coupled Model Intercomparison Project Phase 5 (CMIP5) climate model output shows that the reconstructed dataset substantially reduces the bias in global datasets caused by sparse station coverage, particularly before the 1950s.

Highlights

  • Land air temperatures (1.5–2 m above the land surface) are one of the most fundamental climate variables for documenting and understanding rapid environmental changes (Stocker et al, 2013)

  • Spatial and temporal coverage issues are not independent, an issue that the Data INterpolating Empirical Orthogonal Functions (DINEOF) method (Beckers and Rixen, 2003) is designed to cope with. It determines the number of statistically significant empirical orthogonal functions (EOF) by a cross‐validation procedure for incomplete datasets, as well as quantification of the noise level and interpolation errors

  • Raw GHCNm V4 data were filtered and converted to the format required by DINEOF software: (a) Stations with missing latitude, longitude, or elevation information were deleted. (b) To avoid short records, stations with

Read more

Summary

| INTRODUCTION

Land air temperatures (1.5–2 m above the land surface) are one of the most fundamental climate variables for documenting and understanding rapid environmental changes (Stocker et al, 2013). There is increasing evidence that global air temperature datasets are biased due to the sparseness of observations prior to the 1950s, in the Arctic where enhanced climate changes are expected (Cowtan and Way, 2014; Huang et al, 2017; Wang et al, 2017b). Cowtan and Way (2014) used kriging to spatially interpolate the HadCRUT4 global temperature dataset, Huang et al (2017) used Data INterpolating Empirical Orthogonal Functions (DINEOF) to improve coverage in the Arctic, while Wang et al (2017a) used the Biased Sentinel Hospitals Areal Disease Estimation (BSHADE) method to improve global coverage over land.

WANG and CLOW
DATA HODS
Jan Jul
Raw Reconstruction
Findings
DATA MAT
Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call