Exercise 基于分层抽样的估计地图精度，估计面积和置信区间 1. Accuracy estimation based on sample counting 根据样本计数计算简单的混淆矩阵和准确性估计。如果使用简单随机抽样，基于该方法的地图精度计算是正确的。换句话说，所有样本单元被纳入样本的概率相同 2. Area weighted accuracy estimation for stratified sampling Why is converting the map to an equal area projection a necessary step?等积投影 An accuracy assessment which includes area weights and inclusion probabilities based on the size of the surface, would be more accurate if the area size is preserved.
Exercise geolocated coherence backscatter images analyze and interpret the imagery. SNAP Preparation Coherence Estimation First, we will co-register the input images so that their pixels match and interferometry can be performed. Then, we will calculate the coherence, merge the bursts of the images, and geocode the product. Backscatter Estimation Pixel Spacing (是single-look slant range and azimuth resolutions) 𝛿𝑔𝑟 = 𝛿𝑠𝑟 / sin 查资料的已知S1 SLC data的slant range= 2.7m, azimuth resolution = 22.5m 又已知最终的地理编码相干和后向散射产品的像素间距(pixel spacing)为20米 又假设incidence angle 是34 degree
Exercise 1. Visualization 直方图的分布类型 正态分布（Normal Distribution）： 也称为钟形曲线，是一种对称的分布，具有单一的峰值。大多数观测值集中在均值附近。 均匀分布（Uniform Distribution）： 所有数值具有相同的概率分布，直方图呈现为一个水平线。 正偏分布（Positively Skewed Distribution）： 也称为右偏分布，具有一个较长的右尾。大多数观测值集中在左侧，而右侧有一些较大的异常值。 指数分布（Exponential Distribution）： 呈现指数衰减的形状，通常用于描述随机事件的时间间隔。 双峰分布（Bimodal Distribution）： 具有两个峰值的分布，但不一定非对称。两个峰值可能具有相似或不同的高度。 对数正态分布（Log-Normal Distribution）： 对数正态分布的直方图在普通尺度上可能呈现非对称形状，但在对数尺度上更接近正态分布。 混合分布（Mixture Distribution）： 结合两个或更多不同分布的特征，可能导致多个峰值或非对称形状。 Stretch Compare “Stretch: 100%” (full range of pixel values), “Stretch: 90%” (only the range between the 5th and 95th percentile), and “Stretch 1 σ” (range between mean - 1 std and mean +1 std). 哪个更好地区分了IF和F，heterogeneous 啥的，还有pepper and salt effect 2. HHHV backscatter ratio HH-HV By subtracting HV value from the HH, the difference between the bimodal peaks is removed, bringing the distributions together. 水是相对均匀的，所以水的标准差很小 3. Analysing and describing class specific backscatter characteristics Jefferies-Mathuis distance(JMD) High JMD values between two classes means these two classes are more separatable. And when JMD value close to 0 means that they have similar values, which makes them difficult to distinguish for one another. Euclidean distance 4. Class specific scatter mechanisms Water: surface scatter Forest: Volume scatter Non Forest :The rest of the scattering mechanisms, measured in the HH, have lower intensity in the forest, but higher in the non-forested areas, resulting in higher contrast between these classes. Inundated Forest: HH 疯狂double bounce, HV就跟Forest差不多The signal over inundated forest will be reflected perpendicularly from the smooth water surface to the tree trunks and from the tree trunks it would be sent back to the sensor. L-band 和C-band哪个更适合这个研究forest L!!! Non-forest is in this study area referring to the mines, which have rougher soil surfaces & second vegetation (grass), forest debris (fallen branches). Rougher surfaces would have more interaction with C-band signal, because L-band signal does not see the individual smaller rocks on the soil surface. Taking that into account, the C-band would have less contrast between the canopies and non-forested areas. PPT high backscatter» both VV and VH greener in tropical »more volume purple » high VV, low VH» high surface, low volume ![[27e4f309fb3e9211b0a65021e5cfcc5.png]] Radar Fundamentals all radar system is SAR»Side-looking airborne radar (SLAR) incidence angle will change with near range/far range VH == HV Backscatter often given in 𝜎0(sigma nought» surface) or γ0(gamma nought»stable) in dB scale(small) 不懂回去找 optical vs radar, Side-looking airborne radar, radar geometry, ascending and descending orbits, wavelength, polarisation, Radar measurement, gemetric effects(layover>shadow, foreshortening>weak return) speckle Scattering mechanisms of forest ==Backscattered== signal mainly results from: ● Volume scattering ● Surface scattering ● Single-bounce scattering ● Double-bounce scattering ==Relative contributions== depend on: ● Surface roughness ● Dielectric properties of the medium ==All of these factors== depend on: ● The radar wavelength ● The polarization ● The incidence angle Surface scattering The moisture content (dielectric constant) of the medium governs the strength of the backscatter Surface roughness Shell’s law 相对粗糙度: ==same surface: rougher in short wavelength, smoother in long wavelength== Dielectric Constant介电常数 Complex dielectric Constant (ε) ratio of the permittivity（介电常数） of a material to the permittivity of free space 自由空间的介电常数和材料的介电常数相比 In other words: indicator of the ==reflectivity of a material== ==increasing ε leads to increasing backscatter== 例子 介电常数的波动(fluctuation) Water is purple, because the wave of the water » wind Soil: large ε fluctuations (rain, dry) Forest: stable ε (saturated forest canopy) Ancient river network» dry sand/dry snow»Complex dielectric Constant (ε)=2-3» high permittivity Strong drop under freezing conditions» from water(80) to ice(3-4) Volume scattering Concept: Inhomogeneous medium (canopy): Each inhomogeneity (leaves, branches) scatters incident wave in different directions Multiple scattering leads to depolarisation去极化 and attenuation 衰退of the signal ==Main source of VH and HV scattering== Increasing backscatter with increasing dielectric constant Depolarization Change in the orientation of the electric field during a reflection with curved scatters’ such as disc or cylinders 圆柱体 弯曲散射体(如圆盘或圆柱体)反射时电场方向的变化。 Mainly occurs due to multiple scattering Single-bounce scattering Direct scattering from tree elements ==VV > HH== Some depolarisation, but very small (VH, HV) Double-bounce scattering Concept: Two smooth surfaces form a right angle facing the radar beam ==HH > VV== Some depolarisation; HV and VH backscatter > 0 (but very weak) Forest structure Penetration depth and key forest scatterers Understory » Volume Mangroves HH double bounce»Brightening of the forest during inundation
Exercise Preprocessing Creating a raster brick and cleaning the MODIS data using the reliability layer
SDG(Sustainable Development Goals) 11 5 main areas: Social development Economic development Environmental sustainability Peaceful, just and inclusive societies Partnership Make cities and human settlements inclusive, safe, resilient and sustainable urban : population/building density Urban areas defined Degree of urbanization based on geographical contiguity邻里临近度 and population density人口密度把地方行政单位划分为cities, towns and suburbs and rural areas 基于 1km^2 的population grid cells» 其用于分类行政» urban center, urban cluster, rural grid cells Urban Themes Urban land change, Urban Green Space, Urban Heat Island, Urban Nighttime Lights Urban Land Change How :Growing, shrinking, intensifying, restructuring Global products (4) Global Urban Footprint Global Rural-Urban Mapping Project GlobalLand30 Global Human Settlements Layer
Exercise 4 Why UAV thermal imaging can monitor vegetation stress? It sensitivity to disruptions in stomatal conductance Stage 1: Data exploration and emissivity correction Highest temperature and lowest(最高温最亮) Ground control points deployed to aid with the survey have the highest temperature. Artificial materials tend to be good absorbers, meaning they heat up quicker than natural surfaces. Areas with bare soil also have very high temperature values – soils have low relatively thermal inertia, warming up quickly during daytime. Vegetated areas have lowest temperatures in the image. This is because plants actively lower their temperature through transpiration. Surface temperature irrigated field » low temperature non-irrigated field » high temperature bare soil has low thermal inertia»heats up rapidly during the day Water has high thermal inertia» remain stable temperature during the day so, water is cooler than bare soil. 求 minimum and maximum emissivity value ɛ(λ) + ρ(λ) + τ(λ) = 1 Separate vegetated areas from bare soil NDVI thresholding, unsupervised/supervised classification, image segmentation. After emissivity correction Emissivity correction will increase recorded temperature values of both vegetation canopies and soil. 变化的裸土的温度差比冠层变化的差异大The impact of the correction will be more evident for soil (i.e. larger change in the retrieved temperatures)because of the lower emissivity value being used.
Exercise 一. SfM photogrammetry of UAV data in Agisoft Metashape 1. Import photos 为什么不用起飞和降落时候的照片？ The photos from take off and landing have already been removed, so all photos are at the same height level. ==However, the lines are not really parallel, but they are a bit skewed==. Dependent on the field of view of the camera this may cause issues with the overlap and with having enough different viewing angles over the entire scene. 2. Project with UTM projection Geographic data is often presented in a projected coordinate system, in order to visualize 3D data in two dimensions (on your screen). In this process longitude/latitude is converted to (x,y) coordinates.
Exercise 2 Part 1 Terrestrial Laser Scanning Dataset TLS: originally consisting of 25 scan positions, linked using reflectors. We have two datasets MLS: Clip from scanned area using simultaneous localization and mapping (SLAM).
Physical foundations (a) Draw a (triangular) prism and then draw the path of a beam of light that enters at an incidence angle of 30 degrees, travels through the prism and leaves it at the opposite wall. (b) Does red and blue light follow the same path? What is the reason for your answer?
Exercise 1 Goal 哪些生物物理参数对光谱特征影响最大，在光谱的哪一部分。 PORSAIL = PROSPECT(leaf) + SAIL(canopy) Change soil brightness factor(moisture» brighter= lower) Increasing the soil brightness factor will increase the reflectance (soil looks brighter, probably because the moisture content is lower). 增加土壤亮度系数会增加反射率(土壤看起来更亮，可能是因为含水量更低)。 Cab the leaf chlorophyll content »> BRF Cw the leaf water content »> BRF N internal structure parameter »> BRF LAI ,check the changing in NIR /VIS LIDF leaf inclination angle distribution functions»> canopy reflectance spectra Part 1(Leaf Parameters) Cab: BRF is higher for lower chlorophyll concentrations. The more chlorophyll, the more light is used for photosynthesis, resulting in less reflection in Blue and Red. Therefore, lower BRF values can be observed for higher chlorophyll concentrations. For Green (550 nm), there is a maximum difference between the Cab quantities. Cw: the near infra-red and middle infra-red part of the spectrum (800 nm to 2500 nm) is sensitive to changes to this parameter. The larger the water content, the more light is absorbed, resulting in lower BFR values. N: 如果N大，随着波长的变化更剧烈，但不能说 the more/less leaf layers, the higher /lower BRF value Sensitivities are more pronounced at ==high LAI== because there simply is more leaf mass present. ==答案版：== 叶片叶绿素含量对可见光部分的影响可达750 nm左右。 叶片含水量(Cw)在1000nm以上的红外部分有影响。 叶片结构参数(N)对整个光谱都有影响，但其影响不如其他两个参数那么明显(至少对于所选的设置)。 The leaf chlorophyll content has an effect in the visible part up to about 750 nm. The leaf water content (Cw) has influence in the infrared part beyond 1000 nm. The leaf structure parameter (N) has an effect over the whole spectrum, but its influenced is less pronounced as that of the other two parameters (at least for the settings chosen). Part 2 (Canopy Level) 不同的波长上 In the ==visible== part of the spectrum the highest reflectance is obtained at low LAI because most soil still is visible and the ==soil is brighter than the green leaves.== In the ==NIR== part of the spectrum ==lowest reflectance is obtained at low LAI== (except for the erectophile LIDF) because leaves reflect more than the soil and this effect is enlarged due to ==multiple reflectance== between the leaf layers (leaves are partly transparent for NIR radiation). For the very vertical leaves (erectophile LIDF) the NIR reflectance at low and high LAI is quite similar (there is a cross-over point at about 1000 nm). In the ==MIR== part the effect of absorption of water increases. As a result, ==the low LAI (less leaves, less water absorption) gives the highest reflectance== like in the visible part (although in the visible part it is the chlorophyll that is important). 不同的LIDF 水平A planophile LIDF means leaves are oriented more horizontally. So, you observe relatively more leaf area from nadir and the above effects are very pronounced. ==The reflectance in the NIR also is highest.== 垂直An erectophile LIDF means leaves are oriented more vertically, and the influence of the soil is larger at the lower LAI values. Also more shadows will be visible in the canopy. ==Overall the reflectance in the NIR is lower.== A spherical LIDF is in-between the planophile and erectophile LIDF. 您能否通过比较LAI从低到中与从中到高的阶跃，得出LAI在不同光谱区域灵敏度的线性关系? For all three LIDFs the difference between LAI 0.5 and 2.5 is larger than the difference between LAI 2.5 and 4.5. This underlines the asymptotic nature of the LAI effect on reflectance. This is most pronounced in the visible and middle infrared part of the spectrum. Part 3( Soil background) LAI 的数值影响土壤对LAI的影响 At an LAI of 0.5 there is a clear effect of soil brightness/moisture, meaning that the soil influence still is significant. A lower brightness reduces the reflectance due to increased absorption. At an LAI of 4.0 this effect has nearly completely disappeared, meaning that the soil background has no effect anymore. 基础 LAI lower, the more visible soil , the more mix in the spectrum the more moisture , the darker soil C is defining the slope of the soil line Using the simulated spectrum with LAI = 0.0 and soil brightness = 0.5, assuming that 0.5 is the average across all soil brightness values. The slope is the value of NIR (870 nm) divided by the value of Red (670 nm). WDVI vs NIR WDVI is meant to extract vegetation and ==reduce influence of the soil.== 就是叶片层数越多的时候，土壤越不可见，所以WDVI的作用再减小。 The influence of soil brightness is noticeable in the NIR region because of slightly higher Std Dev for both LAIs. WDVI values are lower for LAI = 0.5 compared to LAI = 4.0, because higher values of WDVI equal to denser vegetation。 The WDVI value increases with greater values for soil brightness, because the dry soil reflects higher the NIR spectrum. Part 4(BRDF双向反射分布函数) 上面的大图 On top we see a so-called polar plot, showing the BRF for all combinations of ==view zenith and azimuth angles== at the set solar zenith angle. Bottom left we see a BRF profile in the ==principal plane== (psi = 0º), and bottom right we see a BRF profile in the ==cross-principal plane== (psi = 90º or -90º). 下面的两个小图 On the x-axis the viewing angle is plotted. The reflectance show a narrow peak in reflectance at -30º. With a solar zenith angle of 30º, this is the backscatter observation position (looking with the sun at our back). ==At this position we do not see any shadows, and the reflectance has a maximum.==也就是： With a solar zenith angle of 10 degrees a hot-spot at 10 degrees viewing angle in the backward direction is observed. Wavelength的变化，引起的BRF的变化 Towards the periphery边缘 of the circle, the value of BRF gets higher. For the Green and NIR bands, the changes are relatively obvious, while for the red band, the changes are not significant. hot-spot parameter (hspot) 的变小 Moving out of the principal plane, the hot-spot effect becomes less pronounced. In the cross-principal plane no hot-spot can be observed anymore and the BRF is mirrored around nadir viewing. Hot-spot parameter设成0和0.5 A hot-spot parameter of 0 means no hot-spot effect is included in the model and the peak has disappeared. However, the angular dependency (anisotropy) still is visible.