USGS Science⌕ Search
Home/Archive/USGS

USGS · 70197124

Do downscaled general circulation models reliably simulate historical climatic conditions?

Andrew R. Bock·Lauren E. Hay·Gregory J. McCabe·Steven L. Markstrom·R. Dwight Atkinson
View original source ↗Export metadata ↓

Abstract

The accuracy of statistically downscaled (SD) general circulation model (GCM) simulations of monthly surface climate for historical conditions (1950–2005) was assessed for the conterminous United States (CONUS). The SD monthly precipitation (PPT) and temperature (TAVE) from 95 GCMs from phases 3 and 5 of the Coupled Model Intercomparison Project (CMIP3 and CMIP5) were used as inputs to a monthly water balance model (MWBM). Distributions of MWBM input (PPT and TAVE) and output [runoff (RUN)] variables derived from gridded station data (GSD) and historical SD climate were compared using the Kolmogorov–Smirnov (KS) test For all three variables considered, the KS test results showed that variables simulated using CMIP5 generally are more reliable than those derived from CMIP3, likely due to improvements in PPT simulations. At most locations across the CONUS, the largest differences between GSD and SD PPT and RUN occurred in the lowest part of the distributions (i.e., low-flow RUN and low-magnitude PPT). Results indicate that for the majority of the CONUS, there are downscaled GCMs that can reliably simulate historical climatic conditions. But, in some geographic locations, none of the SD GCMs replicated historical conditions for two of the three variables (PPT and RUN) based on the KS test, with a significance level of 0.05. In these locations, improved GCM simulations of PPT are needed to more reliably estimate components of the hydrologic cycle. Simple metrics and statistical tests, such as those described here, can provide an initial set of criteria to help simplify GCM selection.

Explore related subjects

Earth InteractionsArticle

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Andrew R. Bock, Lauren E. Hay, Gregory J. McCabe, Steven L. Markstrom, R. Dwight Atkinson. 2018-05-17. Do downscaled general circulation models reliably simulate historical climatic conditions?. https://doi.org/10.1175/ei-d-17-0018.1

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2024

Investigating the atmospheric conditions associated with impactful shallow landslides in California (USA)

Shallow landslides are often triggered during rainfall events, which can increase subsurface soil water pressure and destabilize hillslopes. The likelihood of regional shallow landslide initiation is often assessed through a comparison of rainfall intensity and duration to pre-established thresholds. While informative for landslide warning, this exclusive focus on rainfall exceeding thresholds does not consider the meteorological conditions producing the rainfall. Here, we ask the question, are there common meteorological characteristics that lead to landslide-triggering precipitation? We develop a catalog of 18 post-1995 widespread, impactful shallow landslide events occurring within 13 storms across California, USA, where initiation time could be constrained to a ≤6-h window. We examine storm characteristics during the landslide initiation window using atmospheric reanalysis products, radar observations, and quantitative precipitation estimates. We find that, while there are some common atmospheric characteristics across landslide events, they can occur under a range of atmospheric conditions. For example, all Northern California landslide events assessed are associated with moderate to strong atmospheric rivers (ARs), while Southern California landslides feature non-AR to strong AR conditions. The storm events evaluated herein share many characteristics of hydrologically important storms in California that did not necessarily result in landslides; thus, atmospheric characteristics alone may not be sufficient to determine whether landslides will occur. However, documenting the characteristics of landslide-triggering storms defines the conditions under which landslides tend to occur, provides analog events that can be useful in forecast applications, helps define future research directions relating to atmospheric conditions and landslides, and supports interdisciplinary research efforts.

California↗
USGS2022

Temporal greenness trends in stable natural land cover and relationships with climatic variability across the conterminous United States

Assessment of temporal trends in vegetation greenness and related influences aids understanding of recent change in terrestrial ecosystems and feedbacks from weather, climate, and environment. We analyzed 1-km normalized difference vegetation index (NDVI) timeseries data (1989–2016) derived from the Advanced Very High Resolution Radiometer (AVHRR) and developed growing season time-integrated NDVI (GS-TIN) for estimating seasonal vegetation activity across stable natural land cover in the conterminous United States (CONUS). After removing areas from analysis that had experienced land cover conversion or modification, we conducted a monotonic trend analysis on the GS-TIN timeseries and found that significant positive temporal trends occurred over 35% of the area, while significant negative trends were observed over only 3.5%. Positive trends were prevalent in the forested lands of the eastern third of CONUS and far northwest, as well as in grasslands in the north central plains. We observed negative and nonsignificant trends mainly in the shrublands and grasslands across the northwest, southwest, and west central plains. To understand the relationship of climate variability with these temporal trends, we conducted partial and multiple correlation analyses on GS-TIN, growing season temperature, and water-year precipitation timeseries. The GS-TIN trends in northern forests were positively correlated with temperature. The GS-TIN trends in the central and western shrublands and grasslands were negatively correlated with temperature and positively correlated with precipitation. Our results revealed spatial patterns in vegetation greenness trends for different stable natural vegetation types across CONUS, enhancing understanding gained from prior studies based on coarser 8-km AVHRR data.

Earth Interactions↗
USGS2022

Extensive droughts in the conterminous United States during multiple centuries

Extensive and severe droughts have substantial effects on water supplies, agriculture, and aquatic ecosystems. To better understand these droughts, we used tree-ring-based reconstructions of the Palmer drought severity index (PDSI) for the period 1475–2017 to examine droughts that covered at least 33% of the conterminous United States (CONUS). We identified 37 spatially extensive drought events for the CONUS and examined their spatial and temporal patterns. The duration of the extensive drought events ranged from 3 to 12 yr and on average affected 43% of the CONUS. The recent (2000–08) drought in the southwestern CONUS, often referred to as the turn-of-the-century drought, is likely one of the longest droughts in the CONUS during the past 500 years. A principal components analysis of the PDSI data from 1475 through 2017 resulted in three principal components (PCs) that explain about 48% of the variability of PDSI and are helpful to understand the temporal and spatial variability of the 37 extensive droughts in the CONUS. Analyses of the relations between the three PCs and well-known climate indices, such as indices of El Niño–Southern Oscillation, indicate statistically significant correlations; however, the correlations do not appear to be large enough (all with an absolute value less than 0.45) to be useful for the development of drought prediction models.

Conterminous United States↗