D. Project Description: Acquiring airborne LiDAR data to study hydrologic, geomorphologic, and geochemical processes at three Critical Zone Observatories (CZOs) 1. Introduction The critical zone is defined as the external terrestrial layer extending from the top of the atmospheric boundary layer down through groundwater. The Critical Zone Observatories (CZOs) provide important platforms for studying the processes occurring in the Critical Zone. The goal of CZOs is to build a network to advance interdisciplinary studies of Earth surface processes as well as foster collaboration among scientists and engineers from different disciplines. In 2007 the NSF has funded three observatories: Boulder Creek, Southern Sierra, and Shale Hills (Figure 1). Research at the CZOs involves a growing list of Earth scientists doing integrated science and pursuing both disciplinary and interdisciplinary research (Anderson et al., 2008). At an all-CZO meeting in fall 2008, it was determined that having airborne LiDAR (Light Detection and Ranging) for the three CZOs would significantly advance research, both within and across the CZOs. This under this project we propose to acquire airborne LiDAR data for the study of hydrologic, geomorphologic, and geochemical processes at three CZOs. LiDAR is an optical remote sensing technology that measures properties of scattered light to find range and/or other information of a distant object. The range to an object is calculated by measuring the time delay between transmission of a laser pulse and detection of the reflected signal. Due to its ability to generate 3-dimensional data with high spatial resolution and accuracy, LiDAR technology is being increasingly used in ecology, geography, geology, geomorphology, seismology, remote sensing and atmospheric physics (Brandtberg and Warner, 2003; Gaveau and Hill, 2003; Hopkinson et al, 2004; Figure. 1. Location of three Critical Zone Observatories. SS: Southern Sierra CZO; BC: Boulder Creek CZO; SH: Shale Hills CZO. Lefsky et al, 2002). 2. Existing spatial data and scientific justifications of LiDAR data for three CZO sites: 2.1. Limitations of existing spatial data. We have partial spatial data for three CZO sites; however, none of the existing data meet the research needs of our CZO teams. There are four main problems with the existing data. 1. Incomplete existing data. LiDAR data are available for the headwaters of Boulder Creek CZO, but not for two of three focus subcatchments or for Boulder Canyon and its knick zone. For the Southern Sierra CZO, the LiDAR data only cover part of the study area. 2. Insufficient spatial resolution. For part of the Shale Hills CZO, we have 1.4-m spatial resolution LiDAR data, however at this resolution, many features cannot be resolved. For example, small headwater streams that are smaller than 1 meter in width are numerous in the study area (details see below session 2.2.1). In this proposal, we propose to acquire LiDAR data with ~10 point per square meter, which will result in 25-cm spatial resolution DEM. For the Southern Sierra CZO site, we have LiDAR data with >15-m spatial resolution, which is insufficient for many geomorphologic and hydrologic research questions that we plan to address. 1 3. Outdated spatial data. The partial LiDAR data that are available for Southern Sierra and Boulder Creek CZOs were acquired in 2001 and 2005 respectively. Hydrologic and vegetation products derived from these LiDAR data will be outdated and problematic in hydrologic and biogeochemical modeling. 4. Insufficient temporal scale. For the Shale Hills and Southern Sierra CZO sites, we propose to acquire two LiDAR flights per site (leaf on/off, and snow on/off) to estimate the change of vegetation structure and snow depth. None of current existing data could meet our objective. Boulder would like snow on/off as well. This was our request for the 2005 acquisition, but only the snow off flight was done. 2.2 Science questions and hypothesis that rely on the proposed LiDAR acquisition The following science questions are arranged by CZO, however it should be recognized that although articulated by groups from each CZO, most apply across the network of CZOs and when addressed will stimulate cross-CZO research. That is, questions common to all three CZO’s are not duplicated under each heading. 2.2.1. Boulder Creek CZO Geomorphic models. Recent models of regolith generation from bedrock and downslope transport predict that an inverse relationship should exist between regolith thickness and hillslope curvature. Curvature is highly sensitive to noise in topographic data, and cannot be reliably measured at typical DEM scales of 10-30 meters, but it can and has been measured reliably using bare-ground LiDAR data. Thus, only with bare-ground LiDAR data can this prediction be tested. Geophysics. LiDAR data can speed up the data collection routine especially in areas with steep slopes and in harsh environments. Collecting the start and end point of each line with a good GPS and adding topographic data by the LiDAR dataset afterwards could be a real advantage especially at Green Lakes Valley (short survey times due to thunderstorms etc.) . Snow. Snowmelt runoff provides up to 80% of all water availability in 11 western states for municipal, industrial, and agricultural uses. Snow deposition is heterogeneous, a function of redistribution by wind, avalanches, and sloughing. To understand, quantify, and model runoff, it is essential to account for the spatial variation in snow accumulation. The primary deterrent to advancing our understanding of the processes that control the spatial distribution of snow properties is the lack of spatially distributed measurements of these snow properties. LiDAR data from snow on-snow off time periods will allow us to evaluate: (i the predictive capability of various topographic parameters that may control the spatial distribution of snow, (ii the change in relative importance of each of these parameters over time, (iii the importance of nonlinear interactions between these parameters, and (iv the use of a single point index of total precipitation (such as SNOTEL) to improve models of the spatial distribution of snow depth. Hydrology and hydrochemistry. We propose merging ground-based and remotely sensed snow observations along with LiDAR within a spatially distributed, physically based snowmelt model to estimate the spatial and temporal distribution of SWE and snowmelt and to explicitly calculate the seasonal water balance. These estimates are then coupled to the Alpine Hydrochemical Model (AHM) to simulate discharge and hydrochemical fluxes at the catchment scale. This approach is used to address the following questions. i) What is the spatial and temporal distribution of snow water equivalent and snowmelt over different land cover types in the Green Lake 4 catchment? ii) Does explicit representation of this variability improve simulations of hydrochemical fluxes? 2.2.2. Shale Hills CZO Geomorphic models. A primary question is: what processes control the downslope transport and erosion of regolith, and can these processes be functionally related to regolith thickness in a deterministic fashion? in order to understand the feedbacks between regolith transport, thickness, and weathering, we need to i) characterize rates of erosion throughout the watershed and ii) relate these to 2 distributions of hillslope topography and regolith thickness through a landscape evolution model. The first task to address this will be to quantify rates of regolith transport and erosion. Central to the development of a model to predict regolith evolution in the Shale Hills CZO is the quantitative understanding of the rates and processes of soil transport on hillslopes. To characterize erosion rates we will exploit a high resolution data set of hillslope topography and soil depth, densifying these observations where necessary using a combination of hand-auger, soil pits, and ground penetrating radar. Moreover, understanding how the downslope transport of regolith depends on topographic characteristics such as hillslope gradient will provide the basis for a more process-based model of sediment transport. The second task is to develop and test competing models for soil transport. We propose to develop and test competing models for soil transport (linear creep, depth-dependent creep, disturbance-driven (tree-throw) transport) against the present-day distribution of soil depth, hillslope topography and erosion rate. In particular, observations of pit-and-mound topography in the watershed lead us to hypothesize that disturbance-driven transport (tree-throw -[24]) is a key component of sediment transport on hillslopes in the Shale Hills CZO. We will incorporate rules that describe various soil transport processes into a new landscape evolution model and test these against the distributions of topographic gradients, regolith depth, and lowering rates across the watershed. Vegetation. A second question is: what is the role of biological processes in the shale weathering engine and in particular, what roles do rooting and tree-throw play? Our hypothesis is that tree-throw [24] and the physical and chemical activity of roots strongly influence the evolution of regolith in the SHO by: i) moving regolith downslope; ii) mixing and chemically affecting bedrock and subsoil; iii) providing the principal pipes through the regolith mill for flow of water and solutes; and, perhaps, iv) dictating the microtopography of channelized flow onthe surface. We will deconvolve these effects by quantifying the role of tree-throw in regolith transport and landscape evolution. We will couple measurements of root-length density and soil strength as a function of regolith depth to develop a quantitative model for sediment transport and erosion by tree-throw that will then be incorporated into a new landscape evolution model. We will link this model to observations of erosion across the watershed and with historic records of climate (primarily, frequency of strong wind storms) to scale individual transport events to long-term records of denudation. Moreover, we think it is reasonable to assume that tree-throw swales impart a roughness to the surface that plays an important role in focusing overland flow and funneling surface water to the regolith-bedrock interface. We will map these surface features at the meter scale using a total station and terrestrial LiDAR. We will then conduct a series of numerical experiments using a new landscape evolution model subject to varying degrees of roughness based on our surveys. The experiments will test the hypothesis that regions of simultaneous tree-throw (e.g. during severe storms) determine the location of first-order channels by providing pathways for flow accumulation and saturation in the near subsurface. Hydrologic Modeling. A third question is: what are the processes and functional relationships interconnecting the products and throughput of the regolith mill with topography and local hydrology, and how have these changed in response to a past climatic perturbation? Here we wish to define the numerous interactions and feedbacks between the evolving three dimensional ground surface, the bedrock-regolith interface, regolith thickness, and the hydrologic flow field. Furthermore, the Shale Hills landscape has experienced repeated climate and vegetation changes from glacial to interglacial conditions as well as post-colonial land-use changes. We need to understand the extent to which the present system is a palimpsest. Our approach to obtaining both objectives is to develop a new landscape evolution model and conduct numerical experiments with it driven by fluctuating boundary conditions and calibrated against measurements of both short- and long-term rates of downslope regolith transport and isotopic measurement of long term regolith production rates. The main task involves coupling sediment transport modules to our hydrologic module. The landscape 3 model will be based on PIHM, the Penn State Integrated Hydrologic Model (REF). We will add functions describing conservation of bedrock and sediment mass, multi-grain size sediment routing, channel initiation and growth, hillslope sediment flux by three-throw, rilling after fires, gravity slides, and regolith-generating functions. This model will differ from present models such as SIBERIA [25], PRECIPTON [26], DRAINAL [27], Detachment-limited model [28], GOLEM [29], CASCADE [30], CAESAR [31], ZSCAPE [32], CHILD [33], and LAPSUS (LandscApe Process modelling at mUltidimensions and scaleS) [34] by containing full treatments of regolith hydrology and weathering rules, and will describe the evolution of ground and bedrock surfaces. 2.2.3. Southern Sierra CZO Snow. A primary Question is: what is the control of vegetation characteristics and topography in the spatial distribution of snow, and how such control influences melting patterns and runoff generation? We have two hypotheses: i) In mixed-conifer, Southern Sierra Nevada forests, vegetation exerts a controlling influence on snow distribution and melt, in patterns that are reproducible across the landscape. That control comes about mainly due to how vegetation influences patterns of solar and longwave radiation, and ii) the spatial patterns of snow distribution and melt across the landscape vary in a consistent manner with vegetation (tree) patterns. Understanding of the interactions between snow distribution and vegetation, topography, and other meteorological factors has been limited by the lack of spatially distributed and high-definition data. LiDAR measurements of snow depth provide information at spatial scales that are unmatched by any other existing technology. The snow depth exhibits significant spatial variability induced by wind redistribution of snow, and by the interactions of the snow with topographic features such as ridges and depressions, as well as the location of trees and tree clusters [Trujillo et al., 2007]. A spatially distributed and high-definition LIDAR dataset can be combined with the time series of meteorological and snow variables obtained at the instrument clusters to estimate expected spatial patterns of melting, and consequently runoff generation. In addition, these data will complement hyperspectral airborne data leveraged through collaboration with Noah Molotch at the NASA Jet Propulsion Laboratory. The combination of these two airborne data sets will provide unprecedented information regarding vegetation structure, phenology and interactions with the distribution of snow and snow-surface microstructure. Soil. Downscaling or soil properties and processes can be accomplished through the fusion of existing soil survey observations with high-resolution, raster-based, sources of terrain and micro-climate data to model near surface processes; resulting in a new raster-based description of soil properties at a finer scale than that of the existing soil survey, which depicts general soil groupings. Five specific hypotheses of this approach are as follows. First, soil bodies are not mapped in most traditional soil surveys, soil assemblages are depicted within map units. Existing soil survey datasets can be downscaled through the identification of geomorphic surfaces, microclimate and terrain attributes most correlated to the unmapped locations of soil types within a map unit. LiDAR can be used to derived very precise measures of landform shape, thus leading to much more accurate identification of geomorphic surfaces (as compared to photogrammetrically-derived elevation data). Second, continuous maps of specific soil properties (and confidence intervals) can be developed by statistically modeling landscape-scale pedogenic processes with soil characterization data and appropriate environmental raster-based data (i.e. LiDAR derived terrain attributes and indices of microclimate). Third, hydrologic flowpaths can be derived from a combination of downscaled soil survey information and surface topography. Field studies have shown that flowpaths derived from conventional DEM data (photogrammetric sources) deviate considerably from reality at the landscape to watershed scales. Fourth, the size of wetland meadows and the extent of hydric soils within these environments are a result of the contributing area and connectivity of hydrologic flowpaths to meadows in surrounding terrain. Field studies have shown topographic features (at the landscape to watershed 4 scale) such as low-lying areas or swales are often poorly represented in conventional DEM data, and connectivity between such features does not match reality. Fifth, the degree of conifer encroachment into meadows is a function of the extent of hydric soils, and the extent of saturated conditions. Conversely, pulses of fresh sediment from surrounding terrain raise the elevation of meadows resulting in microtopographic highs that promote conifer encroachment. Microtopographic features are not discernible in conventional DEM data. The ground-based characterization data and distributed measurements needed to address these hypotheses are in place or underway. 2.3. Existing spatial data 2.3.1. Boulder Creek CZO Our focus is on three subcatchments within the mountainous region of Boulder Creek watershed. We are also interested in topography of Boulder Canyon in a swath from the mountain front to the Continental Divide. Currently we have available: 1/3 arc second (approx 10 m) DEM of the full catchment (from USGS NED) 11-m LiDAR-based DEM of a 40-km2 area in the Boulder Creek headwaters, including our uppermost focus subcatchment, Green Lakes Valley. These data were acquired in fall of 2005 and are available in a “filtered” (bare earth) and “unfiltered” version. The data were collected by NCALM and they did the processing as well. This data can be downloaded at http://culter.colorado.edu/exec/Database/gis_layer_query.cgi There is some information on vegetation in our uppermost watershed (Green Lakes Valley), available from the url given above for the LiDAR data. No high-spatial-resolution vegetation data are available for the rest of the watersheds 2.3.2. Shale Hills CZO Our focus is on the Shavers Creek watershed and in particular the Shale Hills subcatchment. The larger Shavers Creek watershed will best suit the nested nature of the project by including 3 watersheds northwest of Shale Hills that will contribute their 50 year continuous data histories, including National Atmospheric Deposition Program (NADP) records since 1979. Currently, the entire Shavers Creek watershed has the following spatial data available 1/3 arc second (approx 10 m) DEM of the full catchment (from USGS NED) 1.4 meter average point spacing operational LiDAR, with vertical accuracy of 18.5 cm RMSE. These data are being collected by the PAMAP program (http://www.dcnr.state.pa.us/topogeo/pamap) 1-m Digital Ortho Photos (www.pasda.psu.edu) An extensive geodatabase of physical land properties has been collated for the entire Susquehanna River Basin. Examples of data include soils, topography, geology, hydrology, etc. The data are available upon request. Spatial vegetation information currently being used includes the National Land Cover Data NLCD, 2001, for the greater Shavers Creek area and initial evaluation of the operational LiDAR (not expected to produce high resolution information of canopy structure). On a smaller scale, an extensive tree survey was conducted at the 20-acre Shale Hills catchment, in which all trees over 18 cm diameter at breast height (DBH) were identified, inventoried and recorded geospatially (~2100 trees). 2.3.3. Southern Sierra CZO Our main focus is on the Kings River Experimental Watershed (KREW) catchments, which cross the rain-snow transition elevations in the southern Sierra Nevada. LiDAR data are needed for the nested Providence catchments, which are the focus of our ground instrumentation, and for comparison with the higher-elevation snow zone, the Bull, Teakettle and Wolverton catchments. All of the KREW catchments, which are on the order of 1 km2 (5-10 km2 for integrating catchments), are instrumented and have ground data on: stream profiles and cross sections, ii) stream conditions, iii) geology and soils, and iv) litter, 5 riparian and upland vegetation. In addition, we have: LiDAR data acquired in 2001 that cover part of our study area, with >15-m spatial resolution. For the rest of watersheds, we have a 30-m spatial resolution DEM from USGS NED. We have developed a DEM from existing LIDAR data, but numerous problems limit its use (Figure 2). The vegetation data are from the USFS standard vegetation product, which were based on 2001 satellite images. The spatial resolution of the vegetation data is 100 m. Only selected areas are covered by high-resolution aerial photos at this site. An extensive database of physical land properties has been collated for the entire Sierra National Forest and Sequoia-Kings Canyon National Park. Examples of data include soil surveys, topography, geology, hydrology, habitat, etc. 2.4. Science justifications for LiDAR data 2.4.1. Boulder Creek CZO Figure 2. Hillshade of Southern Sierra CZO site, which is 10m x 10m and created from the existing LIDAR data. Watershed boundaries are overlain on the hillshade. The LIDAR data points are in a very obvious pattern as NE-SW strips, which are also generally noticeable in the hillshade. Those areas in the red circle are especially apparent. Understanding geomorphic processes. High-resolution topographic data are needed to characterize landscapes and test hypotheses about the geomorphic processes that have generated these landscapes. Three elements of the landscapes of interest are i) bedrock tors (knobs) that range in scale from a few m to 10s of m in width and height, ii) colluvial deposits that swamp bedrock topography in some locations, and should be identifiable by a smooth bare-ground surface, and iii) gullies cut into colluvium. A LiDAR-generated bare-ground 1-m DEM would have the spatial resolution needed to identify bedrock tors and gullies and to test models of sediment transport based on topographic gradient, curvature, and upslope drainage area. Some models of hillslope creep depend non-linearly on slope (topographic gradient), and hence require accurate characterization of slopes. Similarly, since curvature is a second derivative, the topographic data must be high resolution to produce meaningful results. Slope aspect controls on the critical zone. A specific question that has emerged from inspection of Boulder Creek CZO catchments is the role of aspect in geomorphic and hydrologic behavior. We can see qualitatively that north- and south-facing slopes differ strongly in tree density, and suspect that these differences carry over into hillslope morphology and hydrologic function. Testing hypotheses about the role of aspect will require first characterizing differences in bare-earth topography and roughness as a function of aspect. Understanding the role of vegetation could be approached by measuring canopy height and its variance, as well as tree density, as a function of aspect. From this, one could perhaps estimate canopy interception losses and maybe even root density (if one did some field studies of root density and sought correlation between this and canopy properties). Finally, one could quantify the frequency / extent of bare bedrock as a function of aspect. Geomorphic process rates. High-resolution topography will be useful for interpreting alluvial cosmogenic nuclide concentrations, particularly in comparing results between catchments. Significant differences in apparent erosion rate determined from cosmogenic nuclides could be interpreted from detailed 6 measures of slope, aspect, vegetation cover, channel slope, etc. Standard 10-m DEMs are so noisy that it would be hard to resolve differences in slope, stream power and other measures. Hydrologic simulation. The best high-resolution topography is needed for physics-based simulation of surface/near-surface hydrologic response. Using lower resolution data will be of much less use in concept development (understanding how the system works). Data on canopy structure may be useful for constraining the evapotranspiration (ET) fluxes in such a simulation. LiDAR data to simplify interpretation of geophysical data. Topography must be accounted for in interpretation of GPR, seismic refraction, and resistivity. Figure 3 shows tomographic output at two of the Boulder Creek CZO sites based on seismic refraction with and without LiDAR DEM data. Topographic variations across the entire watershed. We have requested a 10-km wide swath of LiDAR along Boulder Canyon from the mountain front to the headwaters at the Continental Divide. This will allow testing of any contrasts in vegetation structure, hillslope topography, north slope vs. south slope differences, and channel topography across a prominent knickpoint. We can begin to develop models of hillslope response to baselevel lowering. Vegetation. We have very strong differences in tree density on north and south facing slopes. There is interest in understanding the influence of slope aspect on soil moisture, sediment movement, hillslope morphology (slope, roughness, curvature, tor frequency). with the high resolution vegetation structure and DEM data derived from LiDAR, we can address many interesting ecological questions: i) to study how higher tree density on N facing slopes affects moisture and recharge, ii) to pinpoint a priori sties that have more or less light reaching the stream, and study the impact of lighting on stream ecology, and iii) to provide the baseline forest canopy before the pine beetle infection, and understand what happens in the aftermath. Figure. 3. Tomographic output at two of the Boulder Creek CZO sites based on seismic refraction with and without LiDAR DEM data. Left (without LiDAR): 10-m DEM. Right (with LiDAR): 1-m DEM. 2.3.2. Shale Hills CZO Geomorphology. The current product available for the area is above mentioned PAMAP LiDAR. At the estimated 1-m resolution, many features cannot be resolved. For example, small headwater streams which are smaller than 1-m in width and numerous in the study area (e.g. the Shale Hills subcatchment stream that is ephemeral in the upper locations of the watershed, especially during the current drought). Hillslope evolution can occur on areas as small as sub-meter, due to acute events related to voids created by tree throw, a major activity in the basin. The void scars change in shape and sediment migrates during large rain events and drought, for that matter. Small landforms such 7 as concave and convex slopes can be from sub-meter to 10s of meter in size, necessitating the highest resolution as possible. Further, the connection of hillslope curvature and soil depth and how that relates to the occurrence of treethrow will lead to a better model of hillslope erosion from this phenomenon. With regard to cosmogenics, the measurements of soil depth (already collected), coupled with higher resolution topography will lead to better estimate of soil flux rates. Comparison of soil flux rates among different subcatchments will aid in evaluating this technique at a larger scale Vegetation. Treethrow is an important physical phenomenon in the subcatchments. Large amounts of sediment and fractured bedrock are displaced during these events. Coupling the high resolution LiDAR with the existing tree survey will aid in developing a model for species-specific effect on erosion rate related to this phenomenon. Additionally, we would be able to assess LiDAR as a technique for tree inventory at subcatchment and even watershed scale. Current ground-based DHB measurements compared with LiDAR estimates of DBH will inform that technique. Hydrologic modeling: High-resolution topography is needed for physically based simulation of coupled surface/ground water. Previous studies using PIHM indicate that patterns evolving at the sub-meter scale are important for hydraulic connections across the hillslope that lead to increased macropore flow. Current LiDAR may illuminate these processes, however, the better sub-meter resolution will continue to advance the realization of these processes at the smallest possible scale. Estimates of streambed geometry and channel roughness in the ephemeral channel will aid in better estimates of the timing and magnitude of streamflow (Cobby et al., 2001). Currently those parameters are estimated in the model and not measured. Due to the prevalence of small headwater streams in this area, the higher resolution LiDAR will be more adequate in producing these measurements. Further, sediment transport processes associated with treethrow will be better resolved (mentioned above). Evapotranspiration is largely unmeasured in hydrologic simulations and is estimated. Canopy structure data correlated with species-specific DBH sapflow and root density (both currently being instrumented) data will be useful in estimating ET or at least in constraining the simulated values. The current PAMAP LiDAR does not provide research grade information on canopy structure Geophysics. Ground-penetrating radar, gravity and electrical resistance tomography studies have been conducted in the Shale Hills catchment and are planned at the larger scale. LiDAR data are expected to reveal micro-topography that will aid in interpretation of these studies. Hydropedology. Similar to the discussion in the geomorphology section, high-resolution estimates of bare earth topography and roughness on hillslopes will aid in developing models for soil depth and hydrologic behavior related to slope, aspect and landscape position. The current data set for soil depth, type, and moisture will be used in concert with the LiDAR to inform these models. 2.3.3. Southern Sierra CZO Geomorphology. High-resolution topographic data are needed to help test hypotheses about the geomorphic and geochemical processes responsible for shaping slopes and stream profiles within the Southern Sierra Nevada CZO. Of particular interest is the so-called "stepped topography," which is manifested at multiple scales in the CZO area and adjacent catchments in the Southern Sierra Nevada. The stepped topography, characterized by sequences of steep "steps" and gently sloping "treads," has previously been described in only general, qualitative terms (e.g. Wahrhaftig, 1965; Clark et al., 2005; Chartrand and Whiting, 2000). It is thought to reflect hydrologic control of weathering and erosion in the critical zone. To test this hypothesis, it will be crucial to first quantify the extent and scale of the steps; a bare-ground 1-m DEM from LIDAR would have ideal spatial resolution for these purposes. A LIDAR-based DEM would also help characterize the geometry of slopes (gradient, curvature, and aspect) and channels (width, gradient, and area) at scales that are appropriate to analysis of functional controls on rates of erosion an weathering---which are being quantified at a range of spatial and 8 temporal scales across the site as part of the CZO study. By quantifying functional controls on rates of erosion and weathering, the LIDAR data would contribute to improved understanding of how (and why) sediment and solutes move across (and through) the landscape. This type of analysis is crucial to the development of geomorphic transport laws, which in turn are fundamental to understanding landscape evolution across a range of scales. A 1-m DEM from LIDAR would also be useful for identifying individual erosional processes and their pervasiveness across the landscape. For example, stream bank and headcut erosion are thought to be important contributors to sediment delivery in the catchments, and the LIDAR data will help in quantitatively mapping it. This should help in developing realistic budgets for decadal-scale sediment supply and delivery from the catchments Hydrologic modeling. High-resolution surface topographic data are also needed to i) delineate channel geometry and ii) quantify fine-scale heterogeneity in surface form for the high-resolution hydrologic modeling that we are setting up for the CZO catchments. Channel geometry will be used in-stream flow routing in models. Sub-catchments within the CZO differ in channel form and improved characterization of these differences will contribute to hydrologic model estimates of snowmelt and winter-precipitation peak flows. Fine-scale surface heterogeneity contributes to surface detention storage and hydrologic controls on soil biogeochemical cycling. LiDAR estimates of fine-scale (sub-meter to meter) topography will be used to parameterize within-unit heterogeneity in a spatially distributed hydrologic model (Kenward et al., 2000) and support model sensitivity analysis that is designed to explore the role of critical-zone heterogeneity in watershed-scale responses (e.g. French, 2003). Estimates of vegetation parameters can also be linked to LiDAR measurements (Entekhabi et al., 1999; Schmugge et al., 2002). Biogeochemical cycling. Micro-topography gives rise to “hot” and “cold” spots of soil biogeochemical cycling generated by preferential flow of nutrient-rich litter leachate into mineral soils. LiDAR based estimates of micro-topography will help identify potential sampling locations along gradients of water and biological availability, guiding locations for sampling litter and soil nutrient concentrations. Biogeochemical hot spots may also be created from greater snow retention, resulting in locales of elevated soil water content during the growing season. Hence, identification of spatial variation in snow depth using LiDAR (see below) should also enhance our ability to track spatial patterns in biogeochemical fluxes resulting from snow-melt patterns. Snow hydrology. Winter LiDAR data are needed to provide high-resolution estimates of spatial variation in snow depth (Fassnacht and Deems, 2006). One of the central questions in the Southern Sierra CZO is the role of forest canopy as a control on snow processes. Snapshots of spatial co-variation in snow depth, surface topography and forest cover will complement plot level studies of snow accumulation and melt. Plot level instrumentation of snow accumulation and melt along transects from under canopy to open can be compared with LiDAR-based estimates of spatial patterns of snow depth and LiDAR used to extend the spatial coverage of these patterns. Methods for estimating the distributions of snow depth have been developed at another Sierra Nevada site, and can be applied to the CZO. Forest structure. Summer LiDAR data are needed to fully characterize spatial patterns of open vs. forest cover and canopy structure. Significant portions of the KREW site have relatively low canopy density, while others are very dense. LiDAR scenes will contribute to estimates of this vegetation structure that are used in coupled hydro-ecologic model analysis and scaling of evapotranspiration and carbon flux. Coupled eco-hydrologic model estimates of evapotranspiration and carbon flux will be compared at the plot level with eddy-flux tower data and sap-flow measurements. LiDAR scenes support parameterization of this model at the watershed scale and contribute to spatially distributed model estimates of ET and carbon flux under different land cover and climate warming scenarios. 3. Field work and LiDAR data-processing methods 9 We propose to have two LiDAR flights per site. Snow on/off for the Boulder Site and Southern Sierra Site, and leaf on/off for the Shale Hills site. We will use physiographic data already available from the CZO science teams and from other sources to derive the LiDAR products. In addition, we will visit each CZO site to collect ground-truth data for evaluate the products, i.e. collect accurate references points, elevation, tree heights, leaf area index, DBH; and systematically validate and calibrate the LiDAR derived products. A team of 3 people will visit each site to conduct the field work. The field data will be used to evaluate the LiDAR product or correlate with LiDAR product to upscale site measurements to the landscape level. LiDAR data processing methods on DEM, vegetation parameters and snow depth are described as below: 3.1. DEM Although LiDAR data can generate a DEM with high accuracy and resolution, appropriate interpolation methods with suitable parameters should be chosen based on the characteristics of data sets. There are a variety of methods to derive DEM from point data, and each method has its own advantages and disadvantages, depending on the characteristics of the data sets. One of our current researches focuses on identifying suitable approaches for deriving DEM from LiDAR data in different situations, and providing guidance in choosing the appropriate interpolation method that best suits the datasets. Factors that are used to describe the characteristics of different data sets include surface roughness, elevation slope, and LiDAR point density. For example, Inverse distance weighted (IDW) interpolation methods have been commonly used to generate DEM from LiDAR data (Hopkinson, 2004; Liu, 2007), but our preliminary results show that IDW performs poorly in high slope areas. Figure 4 shows the Root Mean Square Error (RMSE) of generating DEM by different interpolation methods. We plan to evaluate different interpolation methods with consideration of the influencing factors mentioned above, and produce the optimal DEM data with the suitable parameters based on the characteristics of datasets. 3.2. Vegetation parameters RMSE(m) Compared to traditional methods for assessing forest structure like field inventories and aerial photo interpretation, which are limited in their capacity to make detailed, spatially explicit measurements (Lang et al., 2006), LiDAR has demonstrated the ability to 0.25 generate 3-D data with high spatial resolution and accuracy in forest 0.20 measurements (Figure 5). Methods to extract some vegetation parameters are described as 0.15 follows: IDW Vegetation height. Appropriate 0.10 Neares Neighbor classification procedure should Spline be applied and pre-classify the Ordinary Kriging 0.05 LiDAR data into ground and Universal Kriging non-ground returns. The 0.00 ground returns and canopy 1 2 3 4 5 6 7 8 9 returns can be used to generate Tile number a DEM and digital surface model (DSM,) respectively, with Figure 4. Root Mean Square Error (RMSE) of DEM by different interpolation methods with increasing slope from Tile 1 (5 degree) appropriate interpolation to Tile 9 (25 degree) (IDW: Inverse Distance Weighted, Nearest methods. Subtracting DEM Neighbor, Spline, Ordinary Kriging, and Universal Kriging) 10 from DSM thereby creates a digital canopy model (DCM), a canopy surface with height values recorded in square pixels or cells (Clark et al., 2004). With fine spatial resolution DCMs, individual tree crowns can be detected by appropriate segmentation technique (Brandtberg et al., 2003; Persson et al., 2002). The height and structure of individual crowns can then be estimated by applying metrics to either DCM cells or footprint heights from within crown segments (Brandtberg et al., 2003; Gaveau & Hill, 2003; Persson et al., 2002). Figure 5. 3D vegetation structure visualization from LiDAR data for Fish Camp, CA (a study area from the PI's Sierra Nevada Adaptive Management Project; acquisition date: Sept 2007; spatial resolution: average 9 points/meter.) Crown cover (CC). Crown cover is defined as the percentage of an area within the vertical projection of the periphery of crowns (McDonald et al., 1998). To generate LiDAR-derived equivalent CC estimates, returns greater than a height threshold are considered as tree crown elements which will be interpolated using the highest returns within a proper search radius into a top of canopy surface with appropriate pixel size (Lucas et. al, 2006). The height threshold should be chosen to conform to definitions of forest cover. Therefore, all pixels above the threshold would be coded as 1 or 0 otherwise. As a result, canopy cover percent (CC%) could then be estimated as the sum of all cells with a value of 1 as a percentage of the total (Lucas et. al, 2006). Leaf Area Index (LAI). Leaf Area Index, defined as the single-sided, green leaf area per unit area of ground surface, is a critical variable in ecosystem modeling because of its influence on bio-geochemical cycles (Houldcroft et al., 2005; Gower et al., 1999). LAI can be estimated by measuring both incident (I0) and below-canopy radiation (I) with following equation (Monsi and Saeki, 1953): LAI = −1/ k ln (I / I0 ) where k is the extinction coefficient. The total emitted laser point from an aircraft could be considered as the total amount of sunlight, and the total intercepted or penetrated laser point through the canopy could be regarded as the total amount of blocked or incident sunlight(Kwak et al, 2007). To generate LiDAR-derived LAI, raw LiDAR points are classified into four groups: Ground Return (GR), Low Vegetation Return (LVR), Medium Vegetation Return (MVR) and High Vegetation Return (HVR) (Lim et al., 2001). The Laser Penetration Index (LPI) can be estimated by the HVR and GR, and the Laser Intercept Index (LII) can be calculated from the GR and all point data. Thus, instead of the radiation which reaches from the sun to the ground and vegetation, LAI can be analyzed with the number of ground returns and vegetation returns which reaches from an aircraft (Kwak et al, 2007). In addition, tree Diameter at Breast Height (DBH) and individual tree distribution will also be retrieved from the LiDAR. Detailed methods could refer to (Chen et al, 2006; Popescu, 2007) 11 Snow depth. Two LiDAR datasets are required for this study—both snow on and off for the Boulder Site and Southern Sierra Site. The raw LiDAR data are processed to generate XYZ data files of first and last returns. The last return data from the snow-off survey provides ground surface elevations, and the last return data from the snow-on survey provides snow surface elevations. The bare earth point data would be interpolated into DEM with appropriate pixel size using appropriate interpolation method. The DEM elevations were then subtracted from the snow surface elevation points, producing a dataset of snow depth point estimates. Both survey datasets should be pre-classified into ground or vegetation returns in order to remove the influence of vegetation so that the snow-on ground surface could be compared directly with the snow-off surface (Hopkinson et al., 2004). 4. Broader impacts There are two main, direct broader impacts of this project. First, making LIDAR data available for the three CZOs will enhance their potential and use as community platforms for research. Both investigators currently engaged at the CZOs and others who would like to work at one or more CZO sites after LiDAR data are made available contributed to this proposal. A second broader impact of this proposal will be to make our LiDAR product easily understood and widely used by researchers working at CZOs and other similar study areas. We plan to organize a LiDAR data-education workshop to introduce our LiDAR products and methods we use to derive our products. We will take advantage of the annual CZO meeting at each CZO site, and schedule the workshop during the annual meeting to discuss with the CZO scientists on LiDAR data collection and processing as well as synergies among scientists to use these data. Meanwhile, we plan to work with other organizations (e.g. CUAHSI) to co-sponsor the LiDAR workshop and make the data more broadly accessible to researchers who are interested in working at the CZO sites. In terms of the data distribution plan, we will make the raw LiDAR data as well as the derived products available through our UC Merced digital library, which has already hosted the spatial and non-spatial data for the Southern Sierra CZO Site. We will also explore the opportunity to make the data more accessible via sharing the data with other organizations (e.g. CUAHSI). Another educational impact of this project is to help to build both graduate and undergraduate research at UC Merced, the PI’s home institution. UC Merced has a special commitment to diversity, and we will take advantage of emerging efforts that will help recruit under-represented groups, particularly in California’s San Joaquin Valley. Based on the undergraduate class, UC Merced is provisionally classified as a minority-serving institution. Of the undergraduate students (Fall 2006) 50% identified themselves as first-generation college students and 43.1% as first-generation college students who come from families earning less than $30,000 annually. Two undergraduate students and three graduate students are currently working in Q. Guo’s research group: both undergraduate students are of Hispanic origin from the San Joaquin Valley; one graduate student is Hispanic. Q. Guo will use results from this work in his classes (spatial analysis and modeling, and remote sensing of environment) at UC Merced, and will be involved in an REU program, which will reach out to undergraduate students from the Central Valley, California. 12 5. Schedule Time July - Sept 2009 Oct - Dec 2009 Jan - May 2010 June - Sep 2010 Oct - Dec 2010 Jan - May 2011 June 2011 Sept 2011 Task Contract with NCALM for LiDAR flight at the Shale Hill site (1st flight) Conduct the field work to collect high accurate localities and forest parameters (LAI, tree height, Crown cover, etc) Contract with NCALM for LiDAR flight at the Boulder Creek and Southern Sierra site (1st flight), and LiDAR flight for Shale Hill (2nd flight) Georeference field locality data & collecting relevant field data and geospatial data from each CZO site. Process LiDAR data evaluate different interpolation methods to generate the bare earth and digital surface models generate digital crown model and extracting forest parameters Do LiDAR flight at Boulder Creek and Southern Sierra sites (2nd flight) Conduct field work on these two sites Present the preliminary LiDAR results during the annual CZO meeting, and solicit comments and feedback Process LiDAR data Generate bare earth and digital surface models for Bounder Creak and Southern Sierra Sites Analyze snow off/on LiDAR data for these two sites Up-scaling field measurements to the landscape level Assess the accuracy of the LiDAR derived products (topographic and vegetation variables) Document the project and create all necessary metadata for the spatial data derived from this project Upload and share the data in the digital library Prepare the LiDAR workshop Invite scientists working at CZOs and other similar study areas Present the final results at the annual CZO meeting 6 Management team Q.Guo (PI) will have responsibility, management, and oversight of the project. The PI will contract with NCALM to acquire airborne LiDAR data, and design the overall sampling strategy to collect ground truth for accuracy assessment . A research scientist will conduct field work, and develop algorithms to process and upscale LiDAR data from the plot level to the landscape level. Two undergraduate assistants will help to collect field data & document the project. The PI and his team will work closely with the CZO PIs (Drs. Suzanne Anderson, Chris Duffy, and Roger Bales, see support letters) and other CZO scientists to acquire additional field data for accuracy assessment and upscalling. Meanwhile, the CZO PIs will also provide expert knowledge on earth science to review and interpret the LiDAR products. 13 David M. Cobby, David C. Mason, Ian J. Davenport, Image processing of airborne scanning laser altimetry data for improved river flood modelling, ISPRS Journal of Photogrammetry and Remote Sensing, Volume 56, Issue 2, December 2001, Pages 121-138, ISSN 0924-2716, DOI: 10.1016/S09242716(01)00039-9. Airborne LiDAR in support of geomorphological and hydraulic modelling French, JR Earth Surface Processes and Landforms [Earth Surf. Process. Landforms]. Vol. 28, no. 3, pp. 321-335. Mar 2003. GSA Bulletin; October 1965; v. 76; no. 10; p. 1165-1190; DOI: 10.1130/00167606(1965)76[1165:STOTSS]2.0.CO;2 © 1965 Geological Society of America, Stepped Topography of the Southern Sierra Nevada, California The non-equilibrium landscape of the southern Sierra Nevada, California. Marin K. Clark, Gweltaz Maheo, Jason Saleeby, and Kenneth A. Farley, GSA Today pp. 4–10, vol 15 2005. Shawn M. Chartrand, Peter J. Whiting TI: Alluvial architecture in headwater streams with special emphasis on step-pool topography SO: Earth Surface Processes and Landforms VL: 25 NO: 6 PG: 583-600 YR: 2000 Vol. 80, No. 10, October 1999 Bulletin of the American Meteorological Society Vol. 80, No. 10, October 1999 An Agenda for Land Surface Hydrology Research and a Call for the Second International Hydrological Decade Dara Entekhabi,* Ghassem R. Asrar,+ Alan K. Betts,# Keith J. Beven,& Rafael L. Bras,* Christopher J. Duffy,@ Thomas Dunne,** Randal D. Koster,++ Dennis P. Lettenmaier,## Dennis B. McLaughlin,* William J. Shuttleworth,&& Martinus T. van Genuchten,@@ Ming-Ying Wei,+ and Eric F. Wood*** Remote sensing in hydrology, Thomas J. Schmugge, a, William P. Kustasa, Jerry C. Ritchiea, Thomas J. Jacksona and Al Rangob, Advances in Water Resources Volume 25, Issues 8-12, August-December 2002, Pages 1367-1385 Effects of Digital Elevation Model Accuracy on Hydrologic Predictions Tracey Kenward,*Dennis P. Lettenmaier,* Eric F. Wood,andEric Fielding, REMOTE SENS. ENVIRON. 74:432– 444 (2000 S. R. Fassnacht, J. S. Deems TI: Measurement sampling and scaling for deep montane snow depth data 14 SO: VL: NO: PG: YR: Hydrological Processes 20 4 829-838 2006 15