Implications of a Varying Observational Network for Accurately Estimating Recent Climate Trends

Abstract

Gridded data sets that are widely used to characterize recent historical trends in regional and global climate are derived from a temporally varying and spatially inhomogeneous observational network. Lin and Huybers (2018, https://doi.org/10.1029/2018GL079709) demonstrate that such network variations underlying two widely used precipitation data sets have biased trends in mean and extreme rainfall over India. I highlight similar concerns raised by studies over other regions and discuss the implications for climate change research. Evaluating uncertainties arising from such nonclimatic factors requires access to underlying station data that is currently unavailable for several vulnerable regions but is critical for accurately characterizing recent climate change and evaluating climate models.

Observed Trends in the South Asian Monsoon Low‐Pressure Systems and Rainfall Extremes Since the Late 1970s

“…One caveat pertinent to the dipole trend patterns is the artificial extreme rainfall trend introduced by the time-varying rain-gauge networks, as pointed out by Lin and Huybers (2019). We thus repeat the trend calculation using the relatively coarser IMD 1° × 1° gridded rainfall product, which has comparatively moderate temporal changes in station numbers compared to the IMD 0.25° × 0.25° gridded rainfall shown in Figures 1 and 2 (Figure S7; Singh, 2019). We also compare the time series of rainfall extremes over the two boxed regions using the longest temporal coverage of two satellite-based gridded rainfall data sets (late 1990s-2018; TRMM and GPCP1DD) (Figure S8), all show consistent trends as seen in Figure 2.…”

“…They also showed that no detectable trend can be identified in either monsoon depressions or monsoon lows according to reanalysis‐based LPS tracks. Lin and Huybers (2019), on the other hand, pointed out that one of the greatest uncertainties in estimating the Indian extreme rainfall trend comes from the temporal variations in the rain‐gauge network, which artificially enhanced rainfall extremes over central India after the 1970s (see also Singh et al., 2019).…”

Reply to Comment by E. T. Swenson, D. Das, and J. Shukla on “Unraveling the Mystery of Indian Summer Monsoon Prediction: Improved Estimate of Predictability Limit”

“…More importantly, however, there is a large variation of the number of stations reporting from 1 day to another. The sampling problem associated with such a variable network and analysis into a small grid size (0.25° × 0.25°) could give rise to significant sampling induced spurious daily variability influencing both variance of the daily rainfall as well as the seasonal mean (Lin & Huybers, 2019; Singh, 2019). On the other hand, SA19 uses the Rajeevan et al (2006) data set based on a fixed 1,803 stations and analyzed into 1° × 1° grids.…”

Long‐term changes in spatially coherent extreme precipitation systems over Central India

“…Lin and Huybers (2018) and Singh (2019) indicated that the number of reporting stations decreased significantly in the IMD gridded rainfall datasets with 1.0° resolution from the early 2000s. The variation in the number of stations can also affect the size of extreme precipitation systems and their characteristics.…”