UC Irvine

Clouds have a strong impact on surface radiation fluxes and may have triggered multiple massive melt events in the Arctic. However, harsh and distinctive physical conditions there make it difficult to obtain the regular and reliable in situ observations of clouds and radiation necessary to study the cloud radiative effects (CRE). In this dissertation, we use radiation observed by 30+ automatic weather stations (AWS) all over Greenland, facilitated by a radiative transfer model, to establish the ground-truth of CRE temporal variability and spatial distribution in melt season (May to August). We then use our novel dataset of CRE estimated from in situ measurements to evaluate the CRE estimated by five well-known large-scale datasets from satellite retrievals, reanalyses, and climate models.

AWS provide valuable observations of radiation and basic meteorology. However, their results may contain considerable biases caused primarily by station tilt. We invent a method that relies only on solar geometry (no additional instrumentation) to retrospectively correct tilt-induced errors in insolation, which affect more than 60% of data and can reach up to 200 W m−2. The overall improvement is 11 W m−2 on average, equivalent to 0.24 m of snow melt in liquid during melt season. Albedo estimated using the adjusted insolation presents a consistent semi-smiling diurnal cycle, and agrees better with temperature changes on monthly and inter-annual time scales. Overheating and riming on sensor domes due to a lack of proper shading and ventilation can also contribute to tens W m−2 of biases in longwave measurements. We apply data quality control using physical limits and inter- variable principles to reduce their influences.

We then estimate CRE by subtracting simulated clear-sky radiation from corrected AWS all-sky observations, and examine the relative importance of major factors (such as cloud properties, surface albedo, and solar zenith angle) that determine the temporal and spatial distributions of CRE. Clouds currently warm Greenland during most of the melt season. However, the seasonal trends are contrasting in the ablation (elevation < 1800 m) and accumulation (elevation ≥ 1800 m) zones. Net CRE in the ablation zone, controlled mainly by shortwave CRE, decreases from May to July and increases afterwards. Net warming in the accumulation zone, controlled mainly by longwave CRE, increases from May to August. Average through melt season, clouds warm most of Greenland except in the lower southern ablation zone. CRE generally decreases with elevation, forming a “warm center” spatial distribution. In the ablation zone, the large variability of albedo dominates the seasonal trend and spatial distribution of CRE, shown by strong correlations for both (r > 0.90 and p << 0.01). In the accumulation zone where albedo is constantly high, CRE seasonal trend and spatial distribution are more likely associated with cloud properties, such as cloud fraction and liquid water path. On an hourly timescale, CRE exhibits a bimodal distribution with one peak near 0 W m−2 (i.e., clear state) and the other near 40 W m−2 (i.e., cloudy state), indicating that Greenland is either nearly clear or heavily cloudy with fast transitions between the two. At the cloudy state, CRE strongly correlates with the combination of solar zenith angle and albedo (r=0.85, p<0.01) probably because clouds are already thick enough for CRE to become saturated. The actual links among CRE, cloud properties, and environmental conditions need to be further examined using large-scale observations and determined by model simulations.

Therefore, we evaluate five well-known gridded datasets by assessing their CRE spatial distributions against AWS estimates and examining their cloud-radiation physics as well as simulations of the major determinants of CRE. CRE areal averages from the five datasets are similar (all around 10 W m−2). MERRA-2, ERA-Interim, and CERES CRE estimates agree with in-situ estimates and reproduce the “warm center” distribution. However, the NCAR Arctic System Reanalysis (ASR) and the CESM Large ENSemble community project (LENS) show strong warming in the south and northwest, forming a “warm L-shape” CRE distribution. Discrepancies are mainly caused by longwave CRE in the accumulation zone. MERRA-2, ERA-Interim, and CERES successfully reproduce cloud fraction and its dominant positive influence on longwave CRE in this region. On the other hand, longwave CRE from ASR and LENS correlates strongly with ice water path instead of with cloud fraction or liquid water path. In the ablation zone, MERRA-2 best captures the observed inter-station changes, due to its correct radiation physics and good simulations of surface albedo.

This dissertation provides the first CRE estimate over the entirety of Greenland using multi-year high-quality in-situ observations. It identifies the unique features of CRE temporal and spatial distributions, and uses them to evaluate the verisimilitude of large-scale observations and simulations. Our new methods and findings improve understanding of and ability to predict cloud-related contributions to the increasing widespread melting events in Greenland and, by extension, other polar regions.