Joel Michaelsen

Cross-Validation in Statistical Climate Forecast Models

Abstract Cross-validation is a statistical procedure that produces an estimate of forecast skill which is less biased than the usual hindcast skill estimates. The cross-validation method systematically deletes one or more cases in a dataset, derives a forecast model from the remaining cases, and tests it on the deleted case or cases. The procedure is nonparametric and can be applied to any automated model building technique. It can also provide important diagnostic information about influential cases in the dataset and the stability of the model. Two experiments were conducted using cross-validation to estimate forecast skill in different predictive models of North Pacific sea surface temperatures (SSTs). The results indicate that bias, or artificial predictability (defined here as the difference between the usual hindcast skill and the forecast skill estimated by cross-validation), increases with each decision—either screening of potential predictors or fixing the value of a coefficient—drawn from the da...

Snow accumulation and distribution in an Alpine Watershed

Distribution of snow water equivalence (SWE) was measured in the Emerald Lake watershed located in Sequoia National Park, California, by taking hundreds of depth measurements and density profiles at six locations during the 1986, 1987 and 1988 water years. A stratified sampling scheme was evaluated by identifying and mapping zones of similar snow properties on the basis of topographic parameters that account for variations in both accumulation and ablation. Elevation, slope, and radiation values calculated from a digital elevation model were used to determine the zones. Of the variables studied, net radiation was of primary importance. Field measurements of SWE were combined with the physical attributes of the watershed and clustered to identify similar classes of SWE. The entire basin was then partitioned into zones for each survey date. Statistical analysis showed that partitioning the watershed on the basis of topographic and radiation variables does produce superior results over a simple random sample.

Warming of the Indian Ocean threatens eastern and southern African food security but could be mitigated by agricultural development

Since 1980, the number of undernourished people in eastern and southern Africa has more than doubled. Rural development stalled and rural poverty expanded during the 1990s. Population growth remains very high, and declining per-capita agricultural capacity retards progress toward Millennium Development goals. Analyses of in situ station data and satellite observations of precipitation have identified another problematic trend: main growing-season rainfall receipts have diminished by ≈15% in food-insecure countries clustered along the western rim of the Indian Ocean. Occurring during the main growing seasons in poor countries dependent on rain-fed agriculture, these declines are societally dangerous. Will they persist or intensify? Tracing moisture deficits upstream to an anthropogenically warming Indian Ocean leads us to conclude that further rainfall declines are likely. We present analyses suggesting that warming in the central Indian Ocean disrupts onshore moisture transports, reducing continental rainfall. Thus, late 20th-century anthropogenic Indian Ocean warming has probably already produced societally dangerous climate change by creating drought and social disruption in some of the worlds most fragile food economies. We quantify the potential impacts of the observed precipitation and agricultural capacity trends by modeling “millions of undernourished people” as a function of rainfall, population, cultivated area, seed, and fertilizer use. Persistence of current tendencies may result in a 50% increase in undernourished people by 2030. On the other hand, modest increases in per-capita agricultural productivity could more than offset the observed precipitation declines. Investing in agricultural development can help mitigate climate change while decreasing rural poverty and vulnerability.

Forest responses to increasing aridity and warmth in the southwestern United States

In recent decades, intense droughts, insect outbreaks, and wildfires have led to decreasing tree growth and increasing mortality in many temperate forests. We compared annual tree-ring width data from 1,097 populations in the coterminous United States to climate data and evaluated site-specific tree responses to climate variations throughout the 20th century. For each population, we developed a climate-driven growth equation by using climate records to predict annual ring widths. Forests within the southwestern United States appear particularly sensitive to drought and warmth. We input 21st century climate projections to the equations to predict growth responses. Our results suggest that if temperature and aridity rise as they are projected to, southwestern trees will experience substantially reduced growth during this century. As tree growth declines, mortality rates may increase at many sites. Increases in wildfires and bark-beetle outbreaks in the most recent decade are likely related to extreme drought and high temperatures during this period. Using satellite imagery and aerial survey data, we conservatively calculate that ≈2.7% of southwestern forest and woodland area experienced substantial mortality due to wildfires from 1984 to 2006, and ≈7.6% experienced mortality associated with bark beetles from 1997 to 2008. We estimate that up to ≈18% of southwestern forest area (excluding woodlands) experienced mortality due to bark beetles or wildfire during this period. Expected climatic changes will alter future forest productivity, disturbance regimes, and species ranges throughout the Southwest. Emerging knowledge of these impending transitions informs efforts to adaptively manage southwestern forests.

The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes

The Climate Hazards group Infrared Precipitation with Stations (CHIRPS) dataset builds on previous approaches to ‘smart’ interpolation techniques and high resolution, long period of record precipitation estimates based on infrared Cold Cloud Duration (CCD) observations. The algorithm i) is built around a 0.05° climatology that incorporates satellite information to represent sparsely gauged locations, ii) incorporates daily, pentadal, and monthly 1981-present 0.05° CCD-based precipitation estimates, iii) blends station data to produce a preliminary information product with a latency of about 2 days and a final product with an average latency of about 3 weeks, and iv) uses a novel blending procedure incorporating the spatial correlation structure of CCD-estimates to assign interpolation weights. We present the CHIRPS algorithm, global and regional validation results, and show how CHIRPS can be used to quantify the hydrologic impacts of decreasing precipitation and rising air temperatures in the Greater Horn of Africa. Using the Variable Infiltration Capacity model, we show that CHIRPS can support effective hydrologic forecasts and trend analyses in southeastern Ethiopia.

Regression Tree Analysis of satellite and terrain data to guide vegetation sampling and surveys

. Monitoring of regional vegetation and surface biophysical properties is tightly constrained by both the quantity and quality of ground data. Stratified sampling is often used to increase sampling efficiency, but its effectiveness hinges on appropriate classification of the land surface. A good classification must be sufficiently detailed to include the important sources of spatial variability, but at the same time it should be as parsimonious as possible to conserve scarce and expensive degrees of freedom in ground data. As part of the First ISLSCP (International Satellite Land Surface Climatology Program) Field Experiment (FIFE), we used Regression Tree Analysis to derive an ecological classification of a tall grass prairie landscape. The classification is derived from digital terrain, land use, and land cover data and is based on their association with spectral vegetation indices calculated from single-date and multi-temporal satellite imagery. The regression tree analysis produced a site stratification that is similar to the a priori scheme actually used in FIFE, but is simpler and considerably more effective in reducing sample variance in surface measurements of variables such as biomass, soil moisture and Bowen Ratio. More generally, regression tree analysis is a useful technique for identifying and estimating complex hierarchical relationships in multivariate data sets.

Interactions of Factors Affecting Seedling Recruitment of Blue Oak (Quercus Douglasii) in California

Acorn germination and seedling recruitment of blue oak (Quercus douglasii) were studied in relation to postdispersal predators, planting depth, oak canopy cover, slope angle and aspect, and herb layer. Removal of acorns and seedling recruitment from surface—sown and buried acorns were measured for 2 yr at two sites, a north—slope forest and a ridgetop savanna. Nested exclosures were used to measure acorn predation by birds, mice, gophers, deer, and cattle. The rate of seedling recruitment (Ps) was related to treatment variables using cluster analysis combined with stepwise logistic regression. Average acorn fall ranged from 3.5 to 58.7 acorns/m2, with germination rates varying from 28 to 85% in different years. 8172 sown acorns yielded 2922 (35.76%) seedlings, but Ps varied from 0.09 to 0.71 among the different treatments. Year, site, and rodents interacted strongly to affect seedling recruitment. Ps averaged 0.48 during a cool wet year compared to 0.23 in drier years. Low correlations between seedling recruitment rates for the same plot in different years were due to annual changes in the distribution of favorable microsites and patchy, unpredictable acorn predation. Recruitment rates for buried acorns were twice those of surface—sown acorns due to improved germination and reduced predation. Pocket gophers (Thomomys bottae) were the major predator of buried acorns. Surface acorns suffered high mortality from dying and overheating, as well as from predation by mice, gophers, and cattle. Seedling recruitment was positively associated with increasing canopy cover and more mesic microsites at the low elevation site, but was negatively associated with these factors at the cooler, high elevation site. Hierarchichal classification combined with stratum—specific logistic regression models was important in revealing strong interactions among seedling recruitment factors.

Estimating grassland biomass and leaf area index using ground and satellite data

Abstract We compared estimates of regional biomass and LAI for a tallgrass prairie site derived from ground data versus estimates derived from satellite data. Linear regression models were estimated to predict LAI and biomass from Landsat-TM data for imagery acquired on three dates spanning the growing season of 1987 using co-registered TM data and ground measurements of LAl and biomass collected at 27 grassland sites. Mapped terrain variables including burning treatment, land-use, and topographic position were included as indicator variables in the models to acccount for variance in biomass and LAI not captured in the TM data. Our results show important differences in the relationships between Kauth-Thomas greenness (from TM), LAI, biomass and the various terrain variables. In general, site-wide estimates of biomass and LAI derived from ground versus satellite-based data were comparable. However, substantial differences were observed in June. In a number of cases, the regression models exhibited signific...

Attribution of 2012 and 2003-2012 rainfall deficits in eastern Kenya and southern Somalia

Africa has experienced more frequent boreal spring dry events (Funk et al. 2008; Williams and Funk 2011; Lyon and DeWitt 2012; Funk 2012). In the spring of 2012, below-average March–May rains across parts of eastern Kenya and Southern Somalia (a region bounded by 4°S–4°N, 37°E–43°E, green polygon, Fig. E1A) once again contributed to crisis and emergency levels of food insecurity (FEWS NET 2012a). In some regions, rainfall deficits of more than 30% led to crop failures and poor pasture conditions, causing families in Kenya to move in search of work or take children out of school, and inhibiting Somalia’s recovery from the acute malnutrition and famine caused by the 2010–11 drought. While not particularly severe, the poor March–May 2012 rains added to climatic stresses associated with a series of March–May dry events occurring in 2007, 2008, 2009, and 2011. Figure E1b shows March–May (three month) Standardized Precipitation Index (SPI; McKee et al. 1993) values, based on 1981–2012 FEWS NET precipitation data (see Supplemental Material for a brief description). Dry events, defined as March–May seasons with SPI values of less than -0.5, are shown in orange. In fragile food economies, these repetitive dry events can lower resilience, disrupt development, and require large infusions of emergency assistance. It is not the climate alone that creates these outcomes, but rather the climate’s interaction with extreme poverty, high-endemic rates of malnutrition, limited or nonexistent governmental safety nets, and poor governance. In 2011, for example, the worst drought in 60 years combined with chronic food insecurity, high global food prices, and the actions of Somali terrorists produced an estimated 258 000 deaths in Somalia (FEWS NET, 2013). In this study, we examine the question of whether sea surface temperatures (SSTs) caused the poor 2012 March–May eastern East African rains and increased the frequency of dry events over the past decade (2003–12), using two new Global Forecast System E. ATTRIBUTION OF 2012 AND 2003–12 RAINFALL DEFICITS IN EASTERN KENYA AND SOUTHERN SOMALIAThe European summer of 2012 was marked by strongly contrasting rainfall anomalies, which led to flooding in northern Europe and droughts and wildfires in southern Europe. This season was not an isolated event, rather the latest in a string of summers characterized by a southward shifted Atlantic storm track as described by the negative phase of the SNAO. The degree of decadal variability in these features suggests a role for forcing from outside the dynamical atmosphere, and preliminary numerical experiments suggest that the global SST and low Arctic sea ice extent anomalies are likely to have played a role and that warm North Atlantic SSTs were a particular contributing factor. The direct effects of changes in radiative forcing from greenhouse gas and aerosol forcing are not included in these experiments, but both anthropogenic forcing and natural variability may have influenced the SST and sea ice changes................................................................................................................................................................... iv

dendrohydrology and long‐term hydrologic phenomena

Dendrohydrology provides accurate methods for studying long-term hydrologic variability at regional scales. A substantial literature and body of knowledge exists, attesting to the value of tree ring based hydrologic reconstructions to discern patterns of long-term hydrologic variability. Application studies encompass drought analysis, analysis of extremes, periodicity of rare hydrologic phenomena, regional interdependence of surface moisture conditions, and, in general, the probabilistic analysis of key hydroclimatic variables such as runoff, precipitation, and temperature. The probabilistic analysis includes distributional properties, frequency duration analysis, severity of events, and spatial variability of hydrologic indicators. This paper reviews some fundamental aspects of dendrohydrology, with a perspective on its value to hydrologists in pursuit of an understanding of long-term hydrologic spatial-temporal behavior and provides also a selective citation of previous work conducted within the dendrohydrologic discipline.