Detecting snow under and within trees with satellite lidar for improved climate and weather modelling

雪是地表最大的瞬态特征。它为很大一部分人口提供饮用水，影响天气，并通过水的可用性控制植物生长和野火。积雪范围的地图是由一系列卫星绘制的。它们用于驱动天气和水文预报，并测试气候模型；降雪范围随温度变化是气候模式敏感性准确性的关键指标（Mudryk et al 2020）。目前这些地图是通过被动遥感生成的。由于来自地面和植物的能量混合在一起，这些方法往往低估了森林地区的积雪程度，也不容易探测到树木内的积雪。被树木捕获的雪可以升华到大气中，而树下的雪可以遮挡阳光，改变水文，因此这些过程对于准确预测很重要（Ly等人，2019）。新一代激光雷达（激光测距）卫星可以分离出来自地面和冠层的信号（Armston et al . 2013）。这有可能绘制树下的雪图，并估计树内的雪量（Russell et al 2020）。这些新地图可以在气候和天气预测中逐步改变雪的准确性。树上的雪是基于非常有限的数据来建模的，所以有了大尺度的地图，就可以第一次详细测试树上的雪对天气和水文的影响。准确的树下积雪地图将允许对天气模型进行大规模测试，这是目前天气和气候预测模型的一个很大的不确定性。实验计划在NASA ICEsat-2和GEDI卫星激光雷达（与NASA ICEsat-2植被产品领导合作）确定地面和植被冠层反射率的基础上，第一步将是确定卫星激光雷达是否能够足够准确地测量地面反射率以预测冠层积雪。这里的主要误差是地面定位，因此任何新的发现都可以用来改进所有其他激光雷达数据产品，包括高度和生物量。这将与地面摄像机和高分辨率卫星图像进行比较。一种精确的方法将允许ICESat-2数据绘制地球大片地区的冠层积雪图。可以研究树冠反射率，以确定它是否可以用于测量树木内的雪量。这是目前积雪建模中未知的，可能会对水平衡造成很大的偏差（Russell et al 2021）。任何大规模的观测都有助于改进预报。这可以包括对受雪影响的森林（斯堪的纳维亚或北美）进行实地考察，利用地面激光扫描和雪量测量来监测降雪和被树木捕获。与被动卫星相比，激光雷达具有稀疏的时间覆盖范围，因此ICEsat-2将不适合在季节性时间分辨率下测试模型。为了实现这一目标，ICESat-2数据可用于校准被动光学和微波卫星，通过机器学习技术估算冠下积雪，从而实现高时间分辨率（每月至每日）的大规模制图。这些更新的地图可用于测试受雪影响森林的天气和气候模型，从而允许在气候模型、水文预报和野火估计中应用。最终申请的选择可由博士生在指导团队的支持下决定。

BackgroundSnow is the largest transient feature of the land surface. It provides drinking water to a significant fraction of the population, affects the weather, and controls plant growth and wildfires fire through water availability. Maps of snow extent are produced by a range of satellites. These are used to drive weather and hydrological forecasting and to test climate models; changing snow extent with temperature is a key metric of the accuracy of climate models' sensitivity (Mudryk et al 2020). Currently these maps are generated by passive remote sensing. Due to the mixing of energy from the ground and plants, these tend to underestimate the extent of snow in forested areas and cannot easily detect snow within trees. Snow that is caught in trees can sublime into the atmosphere, whilst snow under trees is shaded from the sun, changing the hydrology and so these processes are important for accurate forecasts (Ly et al 2019).A new generation of lidar (laser ranging) satellites can separate out signals from the ground and canopy (Armston et al 2013). This holds the potential to map snow under trees and to estimate how much is held within trees (Russell et al 2020). These new maps could allow step changes in the accuracy of snow in climate and weather predictions. The snow caught in trees is modelled based on very limited data, so having large-scale maps would allow the first detailed test of the impact of snow in trees on weather and hydrology. Accurate maps of snow under trees would allow large scale testing of weather models, which is currently a large uncertainty in weather and climate forecast models.Experimental planBuilding on work to determine ground and vegetation canopy reflectance from NASA's ICEsat-2 and GEDI satellite lidars (working with the NASA ICESat-2 vegetation product lead), the first step would be to determine whether the satellite lidars can measure ground reflectance accurately enough to predict sub-canopy snow cover. The primary error here is ground finding, and so any novel findings could be used to improve all other lidar data products, including height and biomass. This will be compared against ground cameras and high-resolution satellite images. An accurate method will allow ICESat-2 data to map sub-canopy snow over large areas of the Earth.The canopy reflectance can be investigated to determine whether it can be used to measure the amount of snow held within trees. This is currently an unknown in snow modelling and may be causing large biases in the water balance (Russell et al 2021). Any large-scale observations would help improve forecasts. This can involve fieldwork to snow affected forests (Scandinavia or North America), making use of terrestrial laser scanning and snow mass measurements to monitor snow falling and being caught within trees.Lidar has sparse temporal coverage compared to passive satellites and so ICEsat-2 will not be suitable for testing models at seasonal temporal resolutions. To achieve that, ICESat-2 data can be used to calibrate passive optical and microwave satellites to estimate sub-canopy snow through machine learning techniques, allowing large-scale mapping at high-temporal resolution (monthly to daily).These updated maps can be used to test weather and climate models in snow-affected forests, allowing applications in climate models, hydrological forecasting and wildfire estimation. The choice of which final applications to pursue can be determined by the PhD student, with the support of the supervision team.