增強(qiáng)現(xiàn)實(shí)技術(shù)于洪水風(fēng)險(xiǎn)溝通中的應(yīng)用

著：（英）亞當(dāng)·湯姆金斯 （德）埃卡特·蘭格 譯：陳琦

洪水所引發(fā)的問題日趨嚴(yán)重，尤其對三角洲地區(qū)的城市來說，它們正面臨愈加嚴(yán)峻的挑戰(zhàn)。如珠江三角洲就受到了諸如氣候因素造成的海平面上升[1]和快速城市化而引發(fā)的極端降雨增多等問題的影響[2]。因此，世界各地的三角洲地區(qū)城市政府正在積極制定各種全新的空間發(fā)展戰(zhàn)略，以應(yīng)對氣候變化對易澇的三角洲城市帶來的風(fēng)險(xiǎn)[3]。其中便包括歐盟洪水指令（2007/60/EC），著重強(qiáng)調(diào)了風(fēng)險(xiǎn)溝通的重要性。而作為洪水風(fēng)險(xiǎn)溝中的重要一環(huán)，洪水的可視化在其中扮演著核心角色[4]。洪水的可視化是一種經(jīng)過檢驗(yàn)的強(qiáng)大工具，它可以通過促進(jìn)與利益相關(guān)者的互動來提升人們對不同控制手段下的洪水殘留風(fēng)險(xiǎn)與未來風(fēng)險(xiǎn)的了解，改變?nèi)藗儗樗L(fēng)險(xiǎn)的認(rèn)知[5]。

事實(shí)證明，在規(guī)劃與設(shè)計(jì)流程中運(yùn)用增強(qiáng)現(xiàn)實(shí)技術(shù)來豐富可視化手段，能夠給使用者帶來一系列的益處[6]。增強(qiáng)現(xiàn)實(shí)技術(shù)（Augmented Reality，簡稱AR）自身的計(jì)算機(jī)屬性將其置于各種應(yīng)用形式的十字路口，例如，將這種技術(shù)作為3D可視化的手段，或通過該技術(shù)實(shí)現(xiàn)更加自然的交互方式，抑或利用其進(jìn)行實(shí)時(shí)模擬，等等。也正因?yàn)檫@樣，增強(qiáng)現(xiàn)實(shí)技術(shù)在洪水風(fēng)險(xiǎn)管理中開始扮演更多的角色。

基于增強(qiáng)現(xiàn)實(shí)技術(shù)在溝通方面所展現(xiàn)的各種新興能力，使用該技術(shù)時(shí)必須明確預(yù)期的溝通目標(biāo)。例如，為不同利益相關(guān)者之間架起對話的橋梁，促進(jìn)使用者相應(yīng)的行為改變，或單純地作為一種信息展示的手段。Demeritt等[7]提出一種風(fēng)險(xiǎn)溝通的二維框架，縱軸為互動程度，橫軸為溝通的歸因。

由此，也引出了風(fēng)險(xiǎn)溝通中4種宏觀方法。1）關(guān)于信息傳遞的風(fēng)險(xiǎn)消息模型：側(cè)重于將客觀數(shù)據(jù)進(jìn)行單向傳輸與展示；2）風(fēng)險(xiǎn)工具模型：旨在通過交互實(shí)現(xiàn)行為的改變；3）風(fēng)險(xiǎn)對話模型：通過參與式協(xié)商，進(jìn)行信息的雙向溝通；4）風(fēng)險(xiǎn)治理模型：旨在通過權(quán)威性指導(dǎo)促進(jìn)人們行為的改變。在接下來對增強(qiáng)現(xiàn)實(shí)相關(guān)應(yīng)用進(jìn)行討論的過程中，我們會將這些應(yīng)用場景歸置于Demeritt所提出的理論框架之內(nèi)。以此為指導(dǎo)，評價(jià)增強(qiáng)現(xiàn)實(shí)技術(shù)在各種洪水風(fēng)險(xiǎn)溝通手段中所具有的潛力。

1 方法

在洪水可視化和洪水風(fēng)險(xiǎn)溝通決策的相關(guān)實(shí)踐中，增強(qiáng)現(xiàn)實(shí)技術(shù)正逐漸成為風(fēng)景園林領(lǐng)域一種經(jīng)過檢驗(yàn)且日漸成熟的工具。其應(yīng)用改善了設(shè)計(jì)、開發(fā)和利益相關(guān)者的參與方式[8-12]?？萍嫉倪M(jìn)步降低了增強(qiáng)現(xiàn)實(shí)設(shè)備的成本，也使得該技術(shù)越來越多地被運(yùn)用于現(xiàn)場或?qū)嶒?yàn)室環(huán)境中，以更好地將方案的預(yù)期效果或不同干預(yù)方案的潛在影響展示給利益相關(guān)者[9,13-14]。增強(qiáng)現(xiàn)實(shí)可以通過位移的方式賦予參與者獲取空間線索的能力，使參與者了解動態(tài)的空間關(guān)系[15]，增強(qiáng)其空間認(rèn)知[16]。

在此，重點(diǎn)探討2種在洪水風(fēng)險(xiǎn)溝通中具有前景的增強(qiáng)現(xiàn)實(shí)應(yīng)用，并分別論述它們的優(yōu)勢與潛在機(jī)會。同時(shí)，批判性地評價(jià)它們當(dāng)下在洪水風(fēng)險(xiǎn)管理領(lǐng)域中，所存在的局限性和扮演的角色。

1.1 通過現(xiàn)場型增強(qiáng)現(xiàn)實(shí)技術(shù)提供優(yōu)化的決策支持

傳統(tǒng)洪水風(fēng)險(xiǎn)管理（flood risk management，簡稱FRM）系統(tǒng)往往基于實(shí)驗(yàn)室研究而非現(xiàn)場研究。然而，Haynes等[9]認(rèn)為，雖然基于現(xiàn)場的模擬仍存在許多問題，但它對于環(huán)境建模、規(guī)劃和設(shè)計(jì)都有著十分重要的意義。他們認(rèn)為基于實(shí)驗(yàn)室的研究很容易忽視一些現(xiàn)場所獨(dú)有的特征。因此，嘗試通過現(xiàn)場增強(qiáng)現(xiàn)實(shí)技術(shù)改善現(xiàn)有的FRM工具。 他們設(shè)計(jì)了一款洪水的可視化應(yīng)用，用以協(xié)助FRM流程。在其研究中，記錄了一種現(xiàn)場建模的方法，即通過比照真實(shí)的環(huán)境，手動創(chuàng)建幾何體，從而實(shí)現(xiàn)在沒有相關(guān)地形數(shù)據(jù)與定位標(biāo)識的區(qū)域進(jìn)行增強(qiáng)現(xiàn)實(shí)跟蹤。該應(yīng)用可以在現(xiàn)場對洪水進(jìn)行可視化，同時(shí)將這種簡易的可視化與現(xiàn)場體驗(yàn)相結(jié)合（圖1）。

圖1展示了該應(yīng)用對洪水水位所進(jìn)行的動態(tài)的可視化模擬，應(yīng)用中的洪水由一個(gè)簡單的平面表示。為了表現(xiàn)水流繞過環(huán)境障礙物的效果，使用者需要在現(xiàn)場比照真實(shí)環(huán)境，在軟件內(nèi)運(yùn)用簡單的幾何體搭建洪水以外的環(huán)境物體。運(yùn)行模擬時(shí)，代表洪水的平面會被這些手動搭建的幾何體所阻擋，從而給人一種水從這些環(huán)境物體旁繞過的可視化效果。然而，由于該應(yīng)用并不涉及任何當(dāng)?shù)鼐唧w的地形數(shù)據(jù)，且模擬的精度取決于現(xiàn)場的手動建模。當(dāng)需要考慮該區(qū)域以外洪水的傳播狀況及對應(yīng)的空間范圍時(shí)，便會凸顯該方法的局限性。

該應(yīng)用的優(yōu)點(diǎn)，是它提供了一種新穎且直觀的方法，在沒有增強(qiáng)現(xiàn)實(shí)追蹤標(biāo)識的環(huán)境中，可對洪水水位進(jìn)行動態(tài)可視化。此外，它還與WeSenseIt公民水文觀測站[16]進(jìn)行對接，以實(shí)時(shí)展示傳感器讀數(shù)、歷史信息和疏散路線的可視化。因此，該案例展現(xiàn)了增強(qiáng)現(xiàn)實(shí)技術(shù)在決策支持中一種適合的應(yīng)用場景。同時(shí)，其結(jié)果也證明，作為傳統(tǒng)FRM工具的一種增強(qiáng)手段，即使非專業(yè)人士也可以從中獲取實(shí)用且可靠的信息。 然而在該應(yīng)用中，洪水的建模較為簡單，相較之下，F(xiàn)RM所需的洪水動力學(xué)模擬需要更加復(fù)雜精細(xì)的模型作為支撐，因此該應(yīng)用并不適用于規(guī)劃目的。不過這種對于模型復(fù)雜程度的需求也引出增強(qiáng)現(xiàn)實(shí)技術(shù)的另一種可能性，即基于實(shí)驗(yàn)室的增強(qiáng)現(xiàn)實(shí)應(yīng)用。

為避免手動建模所帶來的各種問題，Mirauda等[11]提出對興趣點(diǎn)（points of interest，簡稱POI）進(jìn)行可視化的方法，即對洪泛區(qū)的重點(diǎn)區(qū)域，如水文站或重要的基礎(chǔ)設(shè)施等興趣點(diǎn)進(jìn)行可視化。

該應(yīng)用通過移動設(shè)備，將POI的位置信息疊加在真實(shí)環(huán)境之上，從而提供一種基于現(xiàn)場環(huán)境的空間視圖。其中，存在洪水風(fēng)險(xiǎn)的點(diǎn)會被突出顯示。類似于Haynes等[9]將可視化與WeSenseIt進(jìn)行整合的做法，Mirauda等[11]開發(fā)的應(yīng)用基于客戶端-服務(wù)器的網(wǎng)絡(luò)架構(gòu)。應(yīng)用程序會從水文站獲取最新的信息用于決策支持，并將實(shí)時(shí)的水文威脅等級告知使用者。該可視化程序僅采用了簡化的增強(qiáng)現(xiàn)實(shí)技術(shù)手段，將當(dāng)前的洪水威脅與潛在洪水威脅的方位與程度通過簡單的2D圖標(biāo)進(jìn)行標(biāo)示，以此為專家使用者提供參考。

1.2 利用增強(qiáng)現(xiàn)實(shí)技術(shù)進(jìn)行大尺度的設(shè)計(jì)與評估

傳統(tǒng)的FRM工具多為基于實(shí)驗(yàn)室的研究，且通過臺式計(jì)算機(jī)實(shí)現(xiàn)。然而筆者認(rèn)為，基于實(shí)驗(yàn)室的增強(qiáng)現(xiàn)實(shí)解決方案也應(yīng)在洪水風(fēng)險(xiǎn)溝通戰(zhàn)略中扮演相應(yīng)角色。因?yàn)樗粌H具有闡明洪水模型原理的能力，同時(shí)也可以用來激發(fā)創(chuàng)新性解決方案的設(shè)計(jì)思路。除此之外它也可以展示不同設(shè)計(jì)干預(yù)所帶來的效果，從而明確不同方案間的權(quán)衡取舍。AR沙盒（AR Sand Box）概念便是一種基于實(shí)驗(yàn)室的增強(qiáng)現(xiàn)實(shí)手段[17-18]。AR沙盒作為一種工具，為規(guī)劃與遠(yuǎn)景目標(biāo)構(gòu)想流程提供了支持。它旨在幫助人們在基于地點(diǎn)的設(shè)計(jì)流程中，更好地對周圍環(huán)境進(jìn)行理解與交互。AR沙盒作為一種設(shè)計(jì)和交流的工具正獲得越來越多的關(guān)注，超過150個(gè)實(shí)驗(yàn)室已將該技術(shù)作為實(shí)踐和教學(xué)的工具[19]。

1 現(xiàn)場洪水水位的可視化[9]On-site flood level visualisation[9]1-1 應(yīng)用的使用場景The application in use1-2 應(yīng)用的界面，展示了模擬的洪水事件The application interface showing a simulated flood event

2 使用中的增強(qiáng)現(xiàn)實(shí)洪水模擬器及一張用作增強(qiáng)現(xiàn)實(shí)配準(zhǔn)標(biāo)識物的地形圖片An Augmented Reality flood simulator in use, with a terrain representation acting as a tracked fiducial marker.

AR沙盒最初由加利福尼亞大學(xué)戴維斯分校（UC Davis）開發(fā)，被用于地球科學(xué)概論的教學(xué)。AR沙盒包含四大部分：一個(gè)含有沙子的盒子，用以直觀地塑造地形；一臺微軟Kinect 3D攝像頭用于掃描經(jīng)過塑造的地形；一臺計(jì)算機(jī)，用以分析地形并計(jì)算相應(yīng)的視覺輸出，例如等高線圖和彩色高程圖；一臺投影儀，用于將該可視化內(nèi)容投影在沙質(zhì)地形上。AR沙盒已經(jīng)運(yùn)用在不同主題的可視化與分析之中，例如洪水、水文、水土流失、流域開發(fā)、海岸建模、地球科學(xué)等[18-19]。

這種由人為改變地形作為輸入，由數(shù)字化手段進(jìn)行輸出的不斷循環(huán)，讓使用者在改變地形的同時(shí)，能夠?qū)崟r(shí)觀察到高程圖的變化。讓實(shí)體模型也具有數(shù)字化的實(shí)時(shí)更新能力，其中包括對水進(jìn)行實(shí)時(shí)模擬的能力[19]。AR沙盒的優(yōu)勢在于它允許使用者以更加直觀的形式進(jìn)行交互，因此十分適合工作室環(huán)境下的協(xié)同設(shè)計(jì)。然而作為一種媒介，AR 沙盒也存在一些問題，例如在實(shí)際的洪水風(fēng)險(xiǎn)溝通中，需要使用更加精確的地形模型。

AR沙盒是邁向成熟的增強(qiáng)現(xiàn)實(shí)平臺的重要一步。然而作為風(fēng)險(xiǎn)溝通的手段，它的有效性仍然有限。首先沙子作為媒介缺乏可塑性，無法進(jìn)行精確的造型，從而讓整個(gè)使用流程被平滑的地形所限制。這樣的應(yīng)用形式雖然適合大尺度的模擬，但對于洪水的可視化，特別是對于水動力學(xué)的模擬來說，模型精度上的欠缺會帶來許多問題，例如類似圖1中展示的情況。其次，運(yùn)用沙子進(jìn)行地形建模是一種建造—破壞—再建造的過程，對模型的改動無法進(jìn)行精確的撤銷還原，影響模型的可靠性。所以使用該方法進(jìn)行方案之間的對比會產(chǎn)生很多問題。理想的方法是將一系列解決方案同時(shí)進(jìn)行可視化，從而為使用者提供一種更加有效的溝通手段，方便其直觀地對比不同方案下的風(fēng)險(xiǎn)與權(quán)衡取舍，同時(shí)也為設(shè)計(jì)與分析的迭代優(yōu)化提供支持。

Afrooz等[17]表示，AR沙盒的主要優(yōu)勢在于它能針對模型的變化進(jìn)行相應(yīng)的快速可視化。受其啟發(fā)，筆者開發(fā)了一種創(chuàng)新性的增強(qiáng)現(xiàn)實(shí)應(yīng)用。該應(yīng)用能夠進(jìn)行交互式的地形建模和洪水的可視化。這項(xiàng)創(chuàng)新旨在改善Haynes等[9]的AR現(xiàn)場洪水可視化應(yīng)用程序與AR沙盒中所存在的不足。

2 成果展示

在此，展示了一種創(chuàng)新型的增強(qiáng)現(xiàn)實(shí)應(yīng)用。它能夠?qū)?shí)時(shí)反饋的洪水模擬與可視化進(jìn)行整合。該應(yīng)用通過一個(gè)3D地形網(wǎng)格對數(shù)字高程模型（digital elevation models，簡稱DEM）進(jìn)行讀取和編輯。圖2展示了通過移動設(shè)備，將虛擬的地形模型置于真實(shí)世界中的情景。參考AR沙盒可塑性的特征，該應(yīng)用允許使用者對模型進(jìn)行數(shù)字化的建模與造型，方便使用者在更為逼真的景觀場景中進(jìn)行相應(yīng)的調(diào)整和測試，并通過與3D景觀的不斷交互來評估與計(jì)算洪水風(fēng)險(xiǎn)的潛在影響。

筆者希望通過這種探索性的應(yīng)用，鼓勵(lì)人們更多地將動態(tài)的可視化手段運(yùn)用于洪水的風(fēng)險(xiǎn)溝通之中，從而協(xié)助使用者設(shè)計(jì)、理解和評估景觀方案與相應(yīng)的干預(yù)措施。在此次應(yīng)用設(shè)計(jì)中，我們使用了在風(fēng)景園林領(lǐng)域被廣泛使用的Unity 3D游戲引擎（Unity Game Engine）[8-9,12,20]作為開發(fā)工具。

該應(yīng)用中的地形基于底層的DEM而創(chuàng)建，使用Unity 地形引擎（Unity Terrain Engine）進(jìn)行可視化，并使用自定義的地形著色器進(jìn)行增強(qiáng)。在此基礎(chǔ)上，將各種2D信息進(jìn)行疊加，例如衛(wèi)星圖像、坡度信息或者更加精細(xì)的數(shù)字表面模型（digital surface models，簡稱DSM，圖3）。

本研究目的是細(xì)化增強(qiáng)現(xiàn)實(shí)技術(shù)在洪水風(fēng)險(xiǎn)溝通中的潛在作用。所以，必須從模型精度的需求、運(yùn)算時(shí)間的限制以及展示的直觀性這3個(gè)方面進(jìn)行考慮，并在其中找到平衡點(diǎn)[22]。為此，采用概念化的洪水傳播模型來有效控制運(yùn)算時(shí)間，以實(shí)現(xiàn)可視化的快速反饋（圖4）。

該應(yīng)用的算法基于“浴盆法”（bathtub method）。出于保持運(yùn)算效率的考慮，并未引入流體力學(xué)進(jìn)行模擬，因此，該應(yīng)用更加適合洪水風(fēng)險(xiǎn)評估與場景建模[22]。在應(yīng)用程序中，為呈現(xiàn)洪水的傳播與對應(yīng)的洪水范圍，將一系列等間隔的平面與高分辨率DEM相交，從而建立洪水范圍與水位之間的聯(lián)系。該方法改善了前文提到的簡易AR可視化中[8-9]不能模擬洪水漸進(jìn)式傳播的限制。同時(shí)，模擬在運(yùn)算速率上保持了高效的可處理性，以便在設(shè)計(jì)工作中為用戶提供快速的反饋。

基于2010年里約熱內(nèi)盧普爾科河漫灘洪水動力學(xué)研究（2010 Rio Puesce River Study）的DEM數(shù)據(jù)[23]，我們模擬了洪水在地形中逐級發(fā)展的場景（圖4）。使用者可以與洪水水位進(jìn)行交互，指定洪水的起始點(diǎn)，并測試地形的改變?nèi)绾斡绊懞樗膫鞑?。該可視化證明，當(dāng)前的AR技術(shù)能夠以簡明且具有互動性的形式提供洪水風(fēng)險(xiǎn)的相關(guān)信息。

受到AR沙盒在洪水風(fēng)險(xiǎn)對話模型中所具潛力的啟發(fā)[17]，我們的應(yīng)用成功實(shí)現(xiàn)了不同設(shè)計(jì)干預(yù)效果之間的直接對比。以防洪為目的，我們將左邊模型中海岸和內(nèi)陸的地形進(jìn)行了一些改動（圖5）。通過可視化展現(xiàn)了原始地形與調(diào)整后地形的對比。兩邊景觀中的水位都處于同樣的高度，但可以看出，左邊地形中的洪水風(fēng)險(xiǎn)有所降低。其中綠色高亮的區(qū)域通過實(shí)時(shí)運(yùn)算，動態(tài)地展示了洪水風(fēng)險(xiǎn)程度降低的區(qū)域。右邊的地形為未經(jīng)改動的原始地形，用以進(jìn)行洪水水位的對比。

3 討論

通過討論增強(qiáng)現(xiàn)實(shí)技術(shù)在各種溝通模型中的應(yīng)用發(fā)現(xiàn)，應(yīng)用之間最大的區(qū)別在于互動程度的目標(biāo)期望不同。Mirauda等[11]與Haynes等[9]的研究成果類似，他們應(yīng)用的主要特征都與風(fēng)險(xiǎn)治理模型相契合。這2種應(yīng)用都是以傳達(dá)相關(guān)信息的形式，嘗試影響專家使用者的行為。然而它們均沒有為使用者提供對話機(jī)制，使可視化僅被用作展示權(quán)威信息來源的手段。

相比之下，AR沙盒采用風(fēng)險(xiǎn)對話模型，遵循迭代設(shè)計(jì)和地理設(shè)計(jì)原則（GeoDesign），其自由靈活的使用流程為創(chuàng)造性的解決方案提供了支持。該應(yīng)用中的信息不包含任何建議和命令，即不具有任何原生的權(quán)威性。在不對使用者的行為目的進(jìn)行指導(dǎo)的前提下，促使參與者通過數(shù)字化的分析與反饋，主動思考并進(jìn)行協(xié)同決策。在此過程中，使用者之間會產(chǎn)生強(qiáng)大的規(guī)范性影響，進(jìn)而讓使用者自行決定相應(yīng)的行為。

基于迭代設(shè)計(jì)，筆者開發(fā)了一種應(yīng)用程序的原型，用以解決以往方法中存在的問題，例如上文提到的模擬過于簡單化，以及在比較不同場景時(shí)溝通能力上的不足。與AR 沙盒的方法不同，該軟件能夠根據(jù)具體的、精確的地形數(shù)據(jù)對洪水風(fēng)險(xiǎn)進(jìn)行評估，并結(jié)合動態(tài)的可視化，展現(xiàn)景觀設(shè)計(jì)干預(yù)所帶來的益處。這些特點(diǎn)凸顯了增強(qiáng)現(xiàn)實(shí)作為一種日漸成熟的溝通工具，在改變用戶態(tài)度和行為方面所具備的能力。

筆者所開發(fā)的原型雖然在技術(shù)特征上與之前的應(yīng)用有著諸多相同之處，但與之不同的是，在風(fēng)險(xiǎn)溝通方面采用風(fēng)險(xiǎn)工具模型，旨在通過雙向的參與模式促進(jìn)行為的改變。

本文討論了增強(qiáng)現(xiàn)實(shí)技術(shù)作為一種媒介，在優(yōu)化傳統(tǒng)FRM工具方面，特別是在洪水風(fēng)險(xiǎn)溝通領(lǐng)域逐漸展現(xiàn)出的能力與適用性。然目前，由于運(yùn)算能力的限制，增強(qiáng)現(xiàn)實(shí)技術(shù)只能進(jìn)行簡化的水文模型模擬。這也阻礙了增強(qiáng)現(xiàn)實(shí)技術(shù)在FRM中的應(yīng)用。未來的研究需要將重點(diǎn)更多地放在如何提高水文模型的復(fù)雜程度上，無論是基于本地設(shè)備還是通過客戶端—服務(wù)器接口的方式。

4 結(jié)論

綜上所述，我們在文中展示了增強(qiáng)現(xiàn)實(shí)技術(shù)在風(fēng)險(xiǎn)溝通實(shí)踐活動方面所展現(xiàn)的潛力。討論了沉浸式的科技作為一種強(qiáng)大的工具，如何透過不同的溝通模型為氣候變化與日益嚴(yán)重的洪水問題，提供風(fēng)險(xiǎn)溝通的手段。無論對于專業(yè)人士還是非專業(yè)人士來說，該技術(shù)都展現(xiàn)了它在溝通方面所具備的潛力。除此之外，它也為現(xiàn)場決策與工作室環(huán)境的協(xié)同合作提供了支持。在此需要強(qiáng)調(diào)，在對該領(lǐng)域未來的研究與相關(guān)應(yīng)用進(jìn)行評估時(shí)，應(yīng)清楚不同的溝通模型及其作用，因?yàn)椴煌臏贤Ｐ蜁绊憫?yīng)用的設(shè)計(jì)和實(shí)施。

致謝：

本研究由牛頓基金（Newton Fund）、英國工程與自然科學(xué)研究理事會（EPSRC）支持，是可持續(xù)三角洲獎（Sustainable Deltas award）的一部分（編號 P/R024979/1）。本研究為我方與華南理工大學(xué)、代爾夫特理工大學(xué)（TU Delft）的共同合作項(xiàng)目。該研究為我方與華南理工大學(xué)、代爾夫特理工大學(xué)的合作項(xiàng)目。

圖片來源：

圖1引自參考文獻(xiàn)[9]；圖2、3、5的底層DEM數(shù)據(jù)來自參考文獻(xiàn)[21]；圖4的底層DEM數(shù)據(jù)來自參考文獻(xiàn)[23]。

Flooding is becoming increasingly problematic, with cities in delta regions such as the Pearl River Delta particularly affected due to both climate-driven sea level rise[1]and increased extreme rainfall events due to rapid urbanization[2]. Therefore, city governments in delta regions are developing a variety of new spatial development strategies to cope with the risks of climate change to flood-prone delta cities[3],with the European Union (EU) Floods Directive(2007/60/EC) emphasising the importance of risk communication. Flood visualisation plays a central role in flood risk communications[4], and has proved to be a powerful tool for changing perceptions of flood risk by engaging with local stakeholders to raise awareness of residual and futureflood risk[5].

Enriching visualisations through Augmented Reality (AR) has been shown to have a range of bene fits in the planning and design process[6]. Due to its inherently computational nature, Augmented Reality, stands at the crossroads of 3D visualisation,natural interaction and real-time simulation, and as such, it has begun to play an increasing role in flood risk communication.

When considering the burgeoning role of this new technology in a communicative capacity, the intended communication goal of its application must be defined, be that to encourage a behavioural change,open a dialogue between stakeholders, or act only as an informative display. Demeritt et al.[7]propose a two-dimensional framework of risk communication,organised by level of engagement on the vertical axis, and the communication rational on the horizontal axis.

This leads to the four broad approaches to risk communication: a risk message model of information transfer focused on a one-way delivery of unbiased data; a risk instrument model focused on achieving behavioural change through engagement; a risk dialogue model of participatory deliberation focused on two-way communication, and a risk government model focused on achieving change though mandate.As we examine the current and potential role of Augmented Reality with respect to flood risk communication, we will take care to situate each application of the technology within the Demeritt framework, as a guide to evaluate the potential of the technology for different flood risk communication methodologies.

1 Methods

Augmented Reality is becoming a tried and tested tool in Landscape Architecture, with applications enriching design, development and stakeholder participation, with recent developments in decision support for flood visualisation and flood risk communication[8-12]. Due to technological advances, affordable Augmented Reality devices have begun to be used to better inform stakeholders of the desired outcomes, and potential consequences of proposed design interventions in both onsite and lab environments[9,13-14]. Augmented Reality can help participants learn dynamic spatial relationships[15], and by engaging a participant’s spatial cues through locomotion, it can improve spatial cognition[16].

Here, we delve into two promising applications of Augmented Reality in flood risk communication,and discuss their strengths and opportunities, with a critical appraisal of their current limitations and roles within the larger flood risk communication landscape.

1.1 Enhancing Decision Support in On-Site Augmented Reality

Traditional flood risk management (FRM)systems tend to be lab-based. However, Haynes et al.[9],argue that on-site modelling, while a difficult problem to solve, is important to environmental modelling,planning and design, arguing that on-site features could be easily overlooked in a lab setting. For this reason, Haynes et al.[9]sought to enhance existing FRM tools by using on-site Augmented Reality technology. They built a flood visualisation tool designed to assist the FRM process. They document a method of manual geometry creation,to support tracking in an unspecified area. The application can visualise environmental flooding in-situ, linking rudimentary simulation to the onsite experience (Fig. 1).

3 該應(yīng)用將地形可視化與洪水模擬進(jìn)行結(jié)合，通過使用自定義的地形著色器，將洪水可視化與其他地形表現(xiàn)形式無縫銜接[21]Combining Topographic visualisation with Flood Simulation. Custom terrain shaders enable the seamless interaction between flood visualisation and a variety of terrain representations[21]

4 基于河床的數(shù)字高程模型（DEM）所進(jìn)行的洪水可視化[23]The flood visualisation in a river bed Digital Elevation Model[23]

5 經(jīng)過改良的地形與原地形之間的對比，以及干預(yù)方案對洪水的影響。圖中綠色高亮區(qū)域經(jīng)過實(shí)時(shí)計(jì)算，動態(tài)地標(biāo)示出了因模型變化而使洪水風(fēng)險(xiǎn)降低的區(qū)域A side-by-side comparison of terrain modification, showing differential flooding severity due to proposed interventions. Green highlights are dynamically calculated to highlight areas with decreased flood risk due to model changes

Fig. 1 shows a dynamically interpolated flood plain visualisation, occluded by the constructed environmental geometry, giving a reasonable visualisation of the water “flowing” around the constructed geometry. However, there is no implicit local terrain knowledge and as such accuracy is limited to the geometry created through primitive shapes, which bring limitations when considering the spatial extent of flood progression outside of the immediate area.

This application presents a novel and intuitive way to dynamically visualise flood levels for any untracked environment in the field. Further to this, the application interfaces with the WeSenseIt citizen water observatory[16], to provide realtime sensor readings, historical information and visualisation of evacuation routes, highlighting a particularly suitable use case for Augmented Reality in decision support. As an augmentation of traditional FRM tools, the results suggest that lay users found the application to be both useful and plausible as a flood plain visualisation, however, due to the simplicity of the flood modelling, left it not suitable in planning due to the more complex flood simulation dynamics required in FRM practice.The requirement for greater geometric complexity suggests a potential role for Augmented Reality in a lab-based situation.

To avoid the issue of content authoring,Mirauda et al.[11], propose an Augmented Reality approach to visualise flood risk using digital Points of Interest (POI), such as hydrologic stations or critical infrastructure in flood plains.

Using a mobile device, POI positional information is overlaid onto the environment,giving a spatial view of salient flood risk points in the field. Similar to the WeSenseIt integration of Haynes et al.[9], Mirauda et al.[11], use a clientserver infrastructure to retrieve up-to-date decision support information from hydrologic stations.These data are used in the application to inform users of real-time hydrologic threat levels. This application takes a simplified approach to AR visualisations, using only simple billboard overlays to instruct experts on the direction and criticality of current and potential threats.

1.2 Large Scale Design and Evaluation in Augmented Reality

Traditional FRM tools tend to be lab-based and desktop-driven. However, lab-based Augmented Reality solutions have a role to play in flood risk communication strategies due to their ability to illustrate flood modelling principles, encourage creative design solutions and elucidate the tradeoffs between competing design interventions.One approach to lab-based Augmented Reality, is the AR Sand Box concept[17-18]. The AR Sandbox is a tool that provides support to planning and visioning processes, designed to help people better understand and engage in place-based design. AR Sandboxes are gaining traction as a design and communication tool, with over 150 labs adopting the technology as both an in practice and as an educational tool[19].

First developed by UC Davis, California, with the aim of teaching earth science concepts, The AR sandbox comprises of 4 main components; A box containing sand to facilitate intuitive terrain shaping, a Microsoft Kinect 3D camera to scan the topography of the moulded terrain, a computer system to analyse the topography and calculate a corresponding visual output such as contour maps and colour elevation, and finally,a projector to re-project the visualisation on top of the sand terrain. AR Sandboxes have been used to visualise and analyse different themes such as flooding, hydrology, soil erosion,watershed development, coastal modelling and geoscience[18-19].

This loop of human input and digital output enables a user to simultaneously manipulate the sand terrain and observe the real-time changes of the elevation map. This allows changes to a physical model to update a digital simulation in real time, including real-time water simulation[19]. The major strength of the AR Sandbox lies in it being a more intuitive approach participant engagement and is especially suited to collaborative design in a workshop setting.However, as a medium, the AR sandbox highlights the utility of accurate terrain models in practical flood risk communication.

The AR sandbox is an important step towards a mature Augmented Reality platform; however,it still suffers from some drawbacks for effective risk communication. Firstly, the adaptability of sand as a medium also limits the accuracy of the medium, restricting the process to generally smooth terrain models, which while suited to large scale simulations, poses a problem when it comes to visualising flood effects of water dynamics around detailed site modes, such as those seen in Figure 1. Secondly, the creative process of terrain modelling in sand is also inherently destructive, as changes to the physical model cannot be reliably undone. This poses a problem when it comes to comparing the effectiveness of two competing design interventions. Ideally, a range of proposed solutions could be visualised concurrently in order to better support iterative design and analysis, and more effectively communicate the risks and tradeoffs of different scenarios.

Afrooz et al.[17], report that the main benefit of the AR Sandbox was reported to be the quick visualisation of changes. Inspired by this work, we introduce a novel Augmented Reality approach for interactive terrain modelling and flood visualisation,which aims to address some of the limitations of both the on-site flood visualisation models of Haynes et al.[9], and the AR Sandbox models.

2 Results

Here we present a novel Augmented Reality application for loading and editing digital elevation models (DEM) using a 3D terrain mesh, integrating a reactive flood simulation and visualisation. Fig. 2 shows the application using mobile Augmented Reality to situate terrain models in the real world. Emulating the malleable terrain of the AR Sandbox, these digital terrain models can be moulded and sculpted digitally.This allows users to prototype landscape changes to a realistic landscape and evaluate and compute the potential impact on flood risks as they interact with the 3D landscape.

Through this exploratory application, we aim to promote dynamic visualisation as a route to effectively communicate the current state-ofaffairs vis-a-vis flood risk and enable the design,comprehension and evaluation of proposed landscape interventions. We chose the Unity Game Engine to implement our application, due to its strong presence in the Landscape Architecture literature[8-9,12,20].

The terrain visualisation is powered by underlying DEM height maps, visualised with the Unity Terrain Engine, and augmented with custom terrain shaders. The terrain model can then be layered with 2D textures to display a range of information such as satellite imagery, slope information, or detailed digital surface models (Fig. 3).

Our aim for this paper is to detail the potential role of AR in flood risk communication,as such, as such we must balance the demands of model complexity, and computational time restraints and intuitive result representations[22].We employ a conceptual model of flood plain propagation in order to have tractable run times to support reactive visualisation (Fig. 4).

Our flood algorithm is based on the“bathtub method”, which is suitable for flood risk assessment and scenario modelling, while remaining computationally efficient, by removing flow dynamics[22]. The flood extent is propagated by intersecting a series of planes at intervals with a high-resolution DEM, to link the water stage with the flood extent. This approach improves on the flood plain representations in previous AR visualisations[8-9], which could not simulate the progressive flood propagation. Our simulation remains computationally tractable, to provide expedited user feedback during design exercises.

Figure 4 shows the results of progressive flooding on a terrain model based on DEM data from the 2010 Rio Puerco River Study of overbank flood dynamics[23]. Through these visualisations,we show that current AR technology can support relevant flood risk information in a succinct and interactive manner, with users able to interact with the flood levels, specify flood starting points,and test how the flood propagation is altered by changes in the terrain.

Inspired by the potential for dialoguedriven flood risk communication seen with AR Sandboxes[17], we have implemented the ability to directly compare the effect of multiple different design interventions on flood risks. Figure 5 shows the live comparison of a potential coastal and inland terrain alteration for improved flood defences. Both landscapes represent the same rise in water height, with the left terrain model showing the reduced flood risks. A green highlight is applied to the terrain in real time, to indicate where the severity of the flood risk is reduced. On the right is the original unchanged terrain model for a floodlevel comparison.

3 Discussion

When evaluating the communication models employed by Augmented Reality approaches,it emerges that the most differentiating factor between applications is the expected level of engagement. Primarily, the most prominent model,shared by the work of Mirauda et al.[11]and Haynes et al.[9]is the Risk Government model. Both applications seek to inform users of pertinent flood information, with the aim of influencing the behaviour of the expert users. In each of these cases, there is no dialogue with the respective user,as information is presented as an authoritative source.

As a counterpoint, the AR Sandboxes follow a Risk Dialogue model to enable a freeform creative process, following the GeoDesign principles of iterative design and creative problem-solving. Here,no information is inherently authoritative, although digital analysis and feedback provide a strong normative influence, reflecting the consequences of collaborative decision making, without instrumental guidance.

By way of iterative design, we develop a prototype application to address drawbacks of previous approaches, such as the unrealistically simplistic flood plain simulations, and an inability to compare different scenarios for more effective communication. Unlike the AR Sandbox approach,our software can evaluate flood risk upon existing,specific, terrain data, coupled with dynamic visualisations designed to highlight the benefit of landscape intervention. These features highlight the ability of Augmented Reality to become a conscious instrument for changing the attitudes and behaviours of the users. In contrast to the previous examples, our prototype, while sharing many of the same technological features, instead embodies the Risk Instrument model of risk communication,explicitly designed to encourage behaviour change while embracing two-way participation.

We have illustrated that current generation Augmented Reality technology is a developing medium suitable for augmenting traditional FRM tools, especially in the domain of flood risk communication. However, as it stands, current computational restrictions prevent Augmented Reality adoption as FRM tools, hampered by reduced complexity of hydrological modelling.Future work must focus on improving hydrological model complexity either on-device or through networked client-server interfaces.

4 Conclusion

In conclusion, we have shown that rapidly developing Augmented Reality technology has the potential to enhance how we conduct flood risk communication exercises. We discuss how immersive technology can be used as a powerful tool to communicate the impact of increasingly severe flood events due to climate change, with applications ranging across the spectrum of communication modes. The technology shows potential for communication with both experts and lay-people alike and also in supporting onsite decision support and collaborative workshop environments. We stress that care should be taken to evaluate future work in this field in line with the multitude of communication modes that could be exploited, affecting the design and implementation of future applications.

Acknowledgments:

This research is supported by the Newton Fund/Engineering and Physical Sciences Research Council (EPSRC) as part of the Sustainable Deltas award (No. P/R024979/1). The Aaptive Urban Transformation research project is a collaborative project with South China University of Technology and Delft University of Technology (TU Delft).

Sources of Figures:

Fig. 1 reproduced with permission from reference [9]; Fig. 2, 3, 5 underlying DEM data from reference [21]; Fig. 4 underlying DEM data from reference [23].